Release Notes

2026-07-22

• WIPO ST.26 Sequence Export Case: Converted exported amino acid sequences to lowercase in the WIPO ST.26 XML sequence listing generation to ensure strict compliance with the WIPO ST.26 standard (which requires all sequences in the XML data block to be lowercase).

• Germline Settings Legend Placement: Relocated the legend block in the Germline analysis workspace settings sidebar from under the Species selection to the very bottom of the sidebar. This prevents the legend from cluttering the species select settings.

• Mol* Residue Selection Highlight & Color Mapping for Constant/Leader Regions: Fixed a bug in the Germline analysis workspace where clicking columns in the Constant or Leader domains did not highlight or show the residue position in the Mol 3D structure viewer. This behavior is corrected by falling back to full_regions alignment mapping and fixing the colon-splitting selection parsing (e.g. for KC:1 formatted identifiers). Robust parsing and fallback support were also added to PFA and Humanization to maintain code parity. Additionally, corrected constant region color mapping in both getRegionColor (sequence tables) and the Mol structure cartoon renderer to resolve their specific KBC_COLORS values (CL, CH1, HINGE, CH2, CH3) instead of generic fallbacks.

• Solvent Exposures Mol* 3D Structure Viewer: Added an interactive Mol 3D structure viewer panel to the Solvent Exposures workspace matching the Humanness workspace layout. Users can click residues in the VL/VH sequence tables to highlight their positions as spacefill/ball-and-stick representations in 3D, and the structure is dynamically colored according to region schemes. Added a "Selected Residue Style" toggle control in the toolbar and "Deselect All" (clear-section-btn) buttons in the VL/VH header sections to quickly clear residue selections. Corrected constant domain color mapping on the 3D structures in the Mol viewer to resolve their specific individual KBC_COLORS values (CL, CH1, HINGE, CH2, CH3) instead of graying them out.

2026-07-20

• AbLang and IgBert NC Reporting for VHH/Single-Chain: Configured the antibody analysis reports to display "NC" (Not Calculated) instead of 0.00 for AbLang2 and IgBert scores on single-chain/VHH sequences where the paired-chain models cannot run. Added conditional formatting support in Excel exports so that "NC" and "N/A" cells bypass coloring.

• Non-Standard Residue Warning on Import: Added warning modals to both the project creation page and the "Add Fvs" project modal. When imported sequence files or pasted sequences contain non-standard amino acid residues (not in the standard 20 amino acids), users are shown a warning detail list of the affected sequences and residues, and can choose to either cancel the upload or proceed anyway.

• IgBert Model Integration: Restored the deep-learning IgBert antibody language model calculation in sequence analysis runs by correcting an internal initialization typo. This enables IgBert probability rows to render correctly in the PFA (Positional Frequency Analysis) view.

• Humanness Molstar Light Chain & Constant Domains Coloring: Fixed a bug where Kappa and Lambda light chains were not colored or highlighted correctly in the Mol* 3D structure viewer. The visualizer now correctly maps both light chain variants to chain ID 'L' in the structural model. Also restored 3D representation coloring for constant domains (CH1, CH2, CH3, Hinge, CL) by retaining constant region residues in the humanness payload.

2026-07-19

• Humanization Molstar Panel Display Alignment: Aligned the PDBe Molstar 3D viewer configuration in the Humanization workspace to be identical to the Germline workspace. Added layoutShowControls: false to start with a collapsed panel by default, and implemented a layout mutation observer that automatically opens the settings panel upon going to fullscreen/expanded view and collapses it when returning to standard view.

• Humanization Sidebar Scrolling: Added vertical scrolling support (overflow-y: auto) and customized scrollbar styling to the .config-content container in the Humanization workspace. This ensures settings and configuration options remain accessible and fully scrollable on smaller screens and restricted height viewports.

• Admin Company List Vertical Expansion: Refactored the "Existing Companies" scrolling container under Admin User Management to use CSS Flexbox (flex: 1) and increased its maximum height bounds. This permits the scroll container to scale and occupy the available vertical space within the parent card, preventing unnecessary vertical scrollbars while preserving the inputs' static widths for horizontal scrolling.

• Unified Click-Away Dropdown Handling: Added a robust, global click-away event listener to base.html that automatically closes open dropdown menus (including user profile, analysis, edit, and export menus) when clicking off of them. The handler dynamically attaches to all iframe documents (e.g. the sequence tables and results grids) so that clicks within these embedded frames correctly dismiss open dropdowns.

• RPEMHC Maximum Allele Score Update & Calibration: Changed the RPEMHC score calculation from a population average to the Maximum Allele Score (worst-case allele risk) to avoid immunogenicity risk dilution. Calibrated the formatting cutoffs to use identical thresholds for both VL and VH (\(\le 28.0\) Good, \(\le 35.0\) Warning) based on clinical reference controls (Pembrolizumab, Trastuzumab, Adalimumab).

• Clinical Comparison Benchmark Re-alignment: Recalculated the maximum-allele RPEMHC scores for all 588 aprobado human therapeutics and regenerated the reference cohort statistics (clinical_stats_data.py) to align the comparison bars with the new maximum-allele distribution.

• Germline Card Label Cleanup: Removed confusing "no data" warnings and human fallback labels from the Humanness scorecard headers. Since comparing non-human candidates to human databases is the standard expected behavior for a humanness tool, these alerts are now hidden and comparisons are carried out cleanly in the background.

• KBC Score Default Formatting Alignment: Aligned the KBC Score's default conditional formatting cutoffs with the updated scoring model that includes Sapiens and RPEMHC weights. By recalculating the entire clinical reference cohort, the thresholds were calibrated to match the new percentile distribution: Good (\(\le 16.4\)), Low/Warning (\(\le 25.2\)), and Medium (\(\le 34.0\)).

2026-07-18

• Responsive Toolbar Menus: Restructured the project results toolbar layout to support responsive design on narrow screens and mobile viewports. Replaced the right-side flex alignment with a margin-based push to prevent overflow clipping, and configured the toolbar to wrap elements dynamically into stacked, readable rows below 1024px, ensuring that no buttons or text are cut off or hidden.

• Clinical Comparison Humanness Sapiens and RPEMHC Ranges: Expanded the Clinical Comparison workspace by calculating Sapiens and RPEMHC immunogenicity ranges across the reference cohort of 584 human clinical antibodies (Thera-SAbDab and Jain et al.). These ranges are now fully integrated into the clinical comparison view and Excel exports, enabling automatic profiling of candidates against reference-stage therapeutics.

• Standardized Humanness Legends & Residue Colors: Added a dedicated RPEMHC severity legend card to the Humanness analysis workspace, visible alongside the OASign card at all times. Standardized the local residue risk colors for both OASign (non-human 9-mer counts) and RPEMHC (MHC-II 15-mer scores) to utilize the low, medium, and high severity colors from colors.py (var(--color-sev-low), var(--color-sev-med), and var(--color-sev-high)). The RPEMHC cutoffs are identical for VL and VH and have been calibrated using clinical controls (e.g. Pembrolizumab, Trastuzumab, Adalimumab) to distinguish low-risk from higher-risk candidates.

• RPEMHC Scoring & Formatting Preset Integration: Fully integrated RPEMHC columns (RPEMHC VL, RPEMHC VH, RPEMHC Fv) into the KBC Scoring Weights config rules (COLUMN_OPTIONS dropdown and default penalty parameters) and user Format Preferences styling, supporting comprehensive immunogenicity customization.

• Dynamic Scoring and Documentation Guides: Documented how sequence humanness (OASign/Sapiens) and MHC-II binding affinity (RPEMHC) act as complementary developability criteria.

• RPEMHC MHC-II Binding Prediction Integration: Integrated the pre-trained RPEMHC Major Histocompatibility Complex Class II (MHC-II) binding affinity predictor into the Humanness Analysis workspace. Added an HLA-II allele selector dropdown to the toolbar, allowing programmatic evaluation across four common human alleles (HLA-DRB10101, HLA-DRB10301, HLA-DRB10401, HLA-DRB11501). Added a dedicated MHC-II score column to the sequence tables, dynamically rendered and color-coded based on affinity risk thresholds to facilitate de-immunization engineering.

• RPEMHC Columns, Dynamic Tool Selector, & Fv-Only Display: Expanded RPEMHC integration by adding RPEMHC VL, RPEMHC VH, and RPEMHC Fv columns directly to the results dashboard with default conditional formatting rules. Added a mode dropdown to toggle the Humanness tool between OASign and RPEMHC (automatically switching score cards, table columns, and alignment grid highlights). Restrained the Humanness sequence tables and grids to show only Fv (variable region) residues by filtering out constant regions.

• Customizable Conditional Formatting (Format Preferences): Added a 'Format Preferences' page in the user settings menu, modeled after KBC Scoring Preferences. This allows users to configure custom conditional formatting ranges and named sets for standard results columns (e.g. SPH, Sapiens, pI, CDR Lengths, and Liability Counts). Also added a 'Format Preset' dropdown selector to the project toolbar, allowing project-specific formatting presets to be applied dynamically, which automatically updates background coloring in the results dashboard and Excel downloads.

• Humanness Real-Time Mutation Evaluation & In Silico Scanning: Added predictive de-immunization scanning and real-time live mutation preview capabilities to the Humanness tool. When opening the residue-level engineering modal, the system performs an in silico mutational scan of all 20 amino acids at that position, displaying HLA-DRB1 binding scores and germline positional frequency values to suggest optimal de-risking mutations. Users can select recommendations or type mutations, generating a live "Before vs. After" comparison of predicted MHC-II and OASign scores directly within the modal before saving.

2026-07-17

• Sapiens Humanness Score Integration & IDC Calibration: Added the deep learning Sapiens humanness score ("Sapiens") as the first column in the Humanness Score category on the results dashboard. The score is computed using the Sapiens transformer model to calculate the mean residue probability of the variable regions of both heavy and light chains. Sapiens thresholds were calibrated against clinical trial anti-drug antibody (ADA) rates from the Immunogenicity Database Collaborative (IDC) V1 dataset to define optimal risk cutoffs: Good (≥ 0.85), Low/Warning (≥ 0.75), Medium (≥ 0.65), and Severe (< 0.65). Added scoring rules and customizable weights for Sapiens VL, VH, and Fv to the KBC configuration settings, UI dashboard rendering, and Excel/CSV exports, fully preserved across incremental analysis runs.

2026-07-16

• Affiliation and Role for Normal Accounts: Added "Affiliation / Organization" and "Primary Use Case / Role" fields to standard/normal accounts. These fields are now requested and required during account activation and initial account setup, displayed and editable in the user profile settings, and optionally editable by administrators when creating new users manually via the User Management panel.

• Convert Evaluation to Normal Account: Added a new bulk action "Convert from Evaluation" to the admin user management dashboard. This feature allows administrators to easily convert any evaluation account to a standard/normal account. The conversion process changes the account type, updates the user's expiration date to one week from today, permanently deletes the trial/evaluation projects and associated runs/files, logs the action in the historical signup log, and emails the user notifying them that their account has been converted. Also updated the login warning alert for expiring accounts to display "Warning" as its title (instead of "Error") and stripped the redundant "Warning: " prefix from the message body.

• Mol* 3D Viewer in PFA Dashboard: Integrated a PDBe Molstar window below the sequence alignment section in the PFA (Positional Frequency Analysis) dashboard. Added a row-resizable drag bar to allow custom window heights. Clicking a residue column in the alignment grid highlights corresponding 3D residues on the structure, with a top toolbar style selector to toggle between Spacefill (default) and Ball-and-Stick representations, matching the look, border, and settings of the Germline dashboard.

2026-07-15

• Dynamic KBC Scoring Rules: Expanded the KBC Scoring Weights configuration to support complete rule customization. The static "Column" column has been replaced with a dropdown menu containing all supportable biological and physical attributes (e.g., CDR lengths, PTMs, OASign scores, etc.). Added a red trash "Delete" action button to each rule row, and an "Add Rule Row" button to the top of the table. In custom presets, rules are saved and loaded as a fully customized, self-contained list rather than a simple set of overrides, enabling arbitrary additions and deletions. Also added drag-and-drop row handles to allow user-friendly sorting and custom reordering of the rule evaluation sequence. Restructured the table to dynamically filter out empty separator rows and group rules into clean, styled category sections (e.g., "CDR Lengths", "PTMs & Cysteines"), restricting drag-and-drop actions to stay within their respective category boundaries for a cleaner layout. Introduced a "Create New Rule" draft panel directly at the top above the table, letting users build and verify rule constraints before clicking "Publish" to save it immediately to the active scoring set. Added default "User entered" labeling in the rule's notes field when created to clearly track customized overrides, and configured row deletion to automatically save changes instantly.

• KBC Preset Name in Excel Exports: Updated Excel exports to dynamically show the active project KBC scoring preset name directly in the weights worksheet tab title (e.g. Weights - {preset_name}). Built automatic cleaning and truncation to ensure the sheet name conforms to Excel's 31-character naming limitations while maintaining link integrity with calculated score formulas.

• Editable KBC Scoring Rules: Expanded the KBC Scoring Weights configuration to allow editing of rule conditions. Instead of a single static rule string, the UI now splits each rule into three interactive fields: a comparison operator selector and two value inputs (with the second value input dynamically enabled for range/BETWEEN constraints). Selecting the count operator (#) automatically clears and disables both value fields. These structured inputs are serialized back into standard rule strings for seamless compatibility with the backend scoring engine.

• Mol* 3D Viewer in Humanness Dashboard: Integrated a PDBe Molstar window below the VL and VH Sequence Analysis grids in the Humanness dashboard. Added a row-resizable drag bar to allow custom window heights. Selected residues in the Sequence Analysis tables highlight corresponding 3D residues on the structure in spacefill representation with static labels, showing no residues by default. Included a clear selection button in the headers of both the VL and VH tables to quickly deselect all active selections and clear 3D residue display. Restricted the sequence analysis grids and tables to display only Fv (variable domain) residues, while maintaining cartoon backbone coloring for the constant domains on the 3D structure.

• Mol* 3D Viewer in Germline Dashboard and Heading Updates: Integrated a PDBe Molstar window below the sequence alignment section in the Germline dashboard. Added a row-resizable drag bar to allow custom window heights. Selected residue columns in the variable domain grid highlight corresponding 3D residues on the structure, with a sidebar style selector to toggle between Spacefill (default) and Ball-and-Stick representations. Also updated the Molstar viewer pane heading in both the Germline and Humanization dashboards to display "3D Structure (Mature Linear Numbering)" to match the Covariance view.

• Project-Level KBC Scoring Presets: Integrated project-level scoring preset selection directly in the project results page. Users can now assign a specific custom named preset to a project using a dropdown menu in the toolbar, allowing different weights to be applied to different projects. Projects default to 'KBC Default' (preset ID 0), ignoring the user's active preference in Scoring Preferences. If a user changes the scoring preset, the setting sticks and dynamically recalculates scores. If a preset is deleted, any project using it automatically reverts back to 'KBC Default'.

• Admin-Editable Global Default KBC Weights: Enabled administrators to modify the baseline global default KBC scoring weights directly from the Admin Settings dashboard. When saved, these configurations serve as the default baseline scoring weights for all users across the platform, bypassing the hardcoded factory configuration while still allowing users to create their own named presets.

• Dashboard and Project Results Value and Header Search: Enhanced both the dashboard and project results search bars to scan all visible cell values, notes, custom metadata, and column headers/keys in the results. Users can search for specific values in a data column (e.g. IGKJ1*01, rhesus, or numeric values like 2.0315) as well as search for custom metadata headers (e.g. titer) from the main dashboard search bar to locate projects, and inside the individual project results table to filter entries.

2026-07-14

• Named Scoring Presets in KBC Preferences: Added support for named scoring presets in the user KBC Scoring Preferences dashboard. Users can now save their customized scoring weights under unique names, switch between different presets from a dropdown menu to dynamically update their project KBC scores, rename custom presets, and delete them. A permanent read-only set called 'KBC Default' remains as the baseline preset and cannot be edited.

• Mol* 3D Viewer in Humanization Dashboard: Integrated a PDBe Molstar window below the sequence alignment sections in the Humanization dashboard. Added a row-resizable drag bar to allow custom window heights. Colored the ribbon structures by region (matching the Liabilities view), including specific sub-domain region colors for the constant regions (CL, CH1, hinge, CH2, CH3). Added a style selector at the bottom of the sidebar config panel to display selected residues as either ball-and-stick or spacefill models (defaulting to spacefill on load). Grid column click selection highlights corresponding 3D residues on the structure in CPK element coloring regardless of the active exclusion method, and is fully synced with immediate grid rendering updates. Supported fullscreen viewports which correctly overlay all sidebar and header content when expanded. Added user documentation detailing the viewer features in Humanization.md.

• FASTA Export Header Formatting: Cleaned up the FASTA sequence export file generation to omit redundant descriptions in headers (e.g., writing >SSJ32-SH11_LC instead of >SSJ32-SH11_LC SSJ32-SH11 Light Chain). Also updated the FASTA sequence parser to be fully robust to imported headers that contain trailing spaces/descriptions when the record ID itself matches a standard chain naming convention.

• Germline Excel Export Header Spacing: Fixed an issue where the species badge, germline gene name, and similarity percentage in the header row ran together (e.g. HIGKV1-17*03(65.2%)). Added proper spacing to improve readability in both the web interface and the downloaded Excel report, formatting the species abbreviation in brackets in the Excel export (e.g., [H] IGKV1-17*03 (65.2%)).

• Dashboard and Project Search (Find Function): Added a unified search bar on the main dashboard and inside individual project results views. Users can search for projects and entries by name, notes, labels, custom metadata, or sequence parts (heavy or light chain) using simple substrings or standard PROSITE/NCBI nomenclature motifs (e.g. G-F-N-I-K-[D]-[T]-Y). Supports optional domain scope modifiers (.Fv, .VH, and .VL) appended to the end of search queries to automatically extract and restrict the sequence search to Fv, heavy chain variable, or light chain variable regions. In the dashboard, matching antibody entries and their hit details are highlighted inline under the project row. In the project results view, the query dynamically filters the displayed entries in the results table.

• Tree Zoom and Scaling Controls: Added a tree scaling widget (with zoom out, reset, and zoom in controls) to the Clading Settings panel, enabling users to adjust the SVG cladogram size from 40% to 300% for optimal viewing of complex structures.

• Instant Layout Switching: Pre-computes and loads both linear and circular SVG representations of the sequence tree in the backend, allowing users to toggle between layouts instantly in the browser without initiating server-side re-runs.

• CVV Severity Threshold and Coloring Adjustment: Updated the CVV severe threshold to 3.0 and aligned the color mapping with the discrete step scores (1.0 = Low/Yellow, 2.0 = Medium/Orange, 3.0 = Severe/Red). Updated the default CVV cutoff on the Liabilities page to 3 so that the grid defaults to showing only severe violators, configured the cutoff input step to 1, and formatted the score values, legend thresholds, and overall CVV Full and CVV FR scores as integers (e.g. [3] instead of [3.00], Low >= 1, Med >= 2, Sev >= 3 instead of Low >= 1.0, Med >= 2.0, Sev >= 3.0, and CVV>= 3 in the viewer header title). Aligned the Covariance tool's CDR filtering behavior with the Liabilities page, so that toggling CDR violations only filters out active violations if the offender residue itself resides in a CDR region, preserving framework offenders with CDR partner constraints. Aligned the framework (FR) offender resolution logic in the backend calculations to ensure consistent scoring between the results grid and the Covariance tool. This ensures consistency across the dashboard, Liabilities page, Covariance tool, and exports.

2026-07-13

• Export Project from Results View: Added the "Export Project (.zip)" option directly to the Export dropdown menu in the project results page, allowing users to export the entire project structure (metadata, sequences, and PDB files) without returning to the main dashboard.

• Admin Signup Log Deletion: Decoupled registration log deletion from user account management. Deleting an entry in the "Registration & Invitation Log (Historical)" table now strictly removes the log record without deleting or modifying any associated user account, regardless of the account status.

• Circular Cladogram Layout: Added a "Circular" layout option in the Clading Analysis tool. This projects the sequence tree radially to provide a highly compact, visually premium visualization for large numbers of sequences, eliminating extensive vertical scrolling. It includes a concentric, drag-to-resize red dashed circle that allows interactive real-time clade distance threshold adjustment and clustering. SVG export natively supports downloading the circular tree layout. The tool dynamically defaults the initial layout view choice to "Circular" when analyzing 50 or more sequences, and "Standard (Linear)" for smaller sets.

2026-07-12

• Multiple Sequence Format Support: Expanded file import support to allow uploading sequence files in FASTQ, GenBank, EMBL, SwissProt, and Clustal formats in addition to standard FASTA. The system automatically parses these formats during import, saving them as clean sequences to the database. Detailed explanations of this automatic normalization have been added to the user documentation.

• Multiple Structure Format Support: Added native support for uploading, parsing, and rendering macromolecular structures in both PDB (.pdb) and mmCIF (.cif / .mmcif) formats. The system converts incoming .cif/.mmcif models to standard PDB format upon upload, storing them as clean PDB data in the database. This ensures compatibility with downstream utilities and the Mol* 3D rendering pipeline. Updated user documentation to reflect this automatic conversion.

2026-07-11

• Germline Database Explorer: Added a new Germline Data view in the Dashboard's File dropdown menu. This view aggregates reference germline sequences by gene type (IGKV, IGKJ, IGLV, IGLJ, IGHV, IGHJ) using tabbed panels. Within each panel, each species is displayed in its own separate grid laid out vertically (Human first), listing only its active subtypes as columns to eliminate sparse matrices. Users can click any non-zero count to inspect the exact sequences in a scrollable detail modal and download them as clean FASTA files.

2026-07-10

• Pairwise Conservation Score (PCS) Integration: Renamed the "Kannan" covariance model to "Pairwise Conservation Score (PCS)" in the Covariance tool sidebar and results dashboard. The documentation now directly references the method's official publication: Gunasekaran, Kannan, Arnold T. Hagler, and Lila M. Gierasch. “Sequence and Structural Analysis of Cellular Retinoic Acid-Binding Proteins Reveals a Network of Conserved Hydrophobic Interactions.” Proteins 54, no. 2 (2004): 179–94.

• Step-Weighted Covariance Scoring: Replaced the linear sum covariance penalty score with a step-weighted score metric (individual offending residues contribute a maximum of 3 points based on their number of partner violations). Recalibrated Good/Warning/Severe thresholds across all three models (Residue OMES, ABHAND OMES, and ABHAND PCS) based on empirical distributions from the human OAS database.

• Covariance Grid & Summary Table Score Visibility: Added a dedicated Score column in the active violations summary table directly before the Cons. Sum column showing the color-coded step-weighted score value (0, 1, 2, or 3) for each offending residue, with full sorting support. Also updated the alignment grid's row headers to display each offender's individual step score contribution dynamically in real-time.

2026-07-09

• AI Chat Assistant Integration: Added a native, Gemini-powered AI Assistant widget to the user dashboard. Users can toggle the assistant by clicking the inline "AI Assistant" link in the footer next to "Report an Issue," completely eliminating screen occlusion on data grids.

• Conversation Session Persistence: Enabled sessionStorage persistence so the assistant remembers the active conversation context and restores previous chat logs when you change pages or reload.

• Interrupted Chat Recovery: Implemented auto-resume logic. If you refresh or navigate away mid-thought, the widget automatically detects the interruption, opens the window, and resumes fetching your answer.

• Custom Warnings & Downloads: Replaced the native browser alert when clearing chats with a custom centered warnings card that invites you to download a backup .txt file before ending the conversation.

• Math Translation: Enhanced markdown processing to translate LaTeX symbols (like \le) into native unicode (≤).

2026-07-08

• Conditional Formatting for Custom Metadata Columns: Added a column setting for conditional formatting to user-uploaded or entered custom metadata columns. Users can open the Column Format Settings modal to specify the actual values or thresholds for each of the four severity levels: Good (Green), Low (Yellow), Med (Orange), and High (Red) along with explicit comparison operators (e.g. >=, >, <=, <, =, !=) for complete control over rules even when only a single severity level is defined. The system evaluates cell values against these defined operator-value thresholds to apply background colors in the results dashboard and Excel exports. Fixed Excel export to apply these as native conditional formatting rules rather than static cell background fills, ensuring continuous range matching for intermediate values. Corrected decimal place formatting in Excel cells to match the results grid precision settings. Enabled dynamic, real-time background color and format updates in the browser results grid immediately when cell values are edited inline.

2026-07-07

• Fix Standard PTM Liabilities UI Visibility: Fixed a critical bug in the Liabilities workspace where standard PTM liabilities (Deamidation, Oxidation, etc.) would fail to show up in the sidebar or results table for newly run or downloaded projects. The project export API now correctly forwards the system default motifs to the frontend so they can be processed and displayed successfully.

• Dynamic Chain Resolution for PyMOL & SVL Exports: Updated structural visualization script generation for PyMOL and SVL (MOE) to dynamically map variable domain liabilities by sequence similarity. This allows liabilities, CDR regions, and framework colors to be correctly mapped and highlighted on custom-labeled chains in the loaded structure. On structures with four or more chains (e.g. IgG/Fab), highlighting is limited to the first matched light chain and heavy chain (one L, one H) while keeping all unmatched/uncolored chains visible to preserve overall structural context.

• Engineering Feature Formatting & Cell Widths: Updated liabilities and features representation. For all engineering features, matched residues and residue coordinates are now displayed as comma-separated lists (e.g. 237, 238, 332 and A, A, G) rather than sequence ranges. Table columns (M. Linear, numbering schemes, etc.) now scale their widths dynamically to fit these content lists. Selecting a multi-residue row now highlights each individual residue in both the interactive 3D structure viewer and the PyMOL export script.

• Eu Numbering Warning in Alignment: Added a warning message in the Alignment tool when the user selects only the "Eu" numbering scheme. Because Eu numbering is designed exclusively for antibody constant domains, the tool now displays a clear, user-friendly prompt instructing the user to select a different variable domain numbering scheme rather than rendering empty sequence grids.

• Zero-Combination Explanatory Warning Modal: Added diagnostic checking to the Engineering workspace. If selected mutations and grouping rules (e.g., Single-paired mutations with disabled parental inclusion, empty mutant residue lists, or limits configurations) make it impossible to generate any variant designs, or if the generated variants are skipped as duplicates/parental-only, the tool now displays a detailed explanation modal detailing exactly which constraints or existing records are conflicting.

• Preserve Row Selections During Pairing/Unpairing: Updated the Engineering workspace grid behaviour to retain active row selections when pairing or unpairing design rows. This eliminates the need to manually re-select the mutated entries for downstream combinatorial generation after grouping them.

• Automatic Parental Fallback for Single-Paired Groups: Updated combinatorial generation to implicitly include parental residues in the choice sets for any position belonging to a "Single" pairing group. This prevents logical contradictions where turning off "Include Parent" on multiple Single-paired items would result in 0 possible combinations (by forcing all to mutate simultaneously, violating the "at most one mutation" rule). This parental option is preserved for Single-paired positions even when generating "Non-Parent" combinations, preventing combinatorial deadlocks when mixed with other non-paired or Together-paired mutations.

• Reuse Existing Pairing IDs when Re-Pairing: Refined the pairing creation behaviour in the Engineering grid. If all selected rows are already grouped under the exact same pair ID, setting the pairing option (e.g. changing between Single and Together) now modifies the pair type in-place and reuses that ID rather than generating a new pair number and inflating the pair count.

2026-07-06

• Alignment Selection Retention: Fixed column selection behavior in the Alignment view. Selecting a residue column and switching the primary numbering scheme (e.g., from IMGT to AHo) now translates the selected position to map to the same residue column, ensuring the selection persists on the intended biological residues.

• Kannan Covariance Method & Custom Region Schemes: Added the on-the-fly ABHAND (Kannan) Covariance Model to the Covariance tool. Selecting this model evaluates query and sandbox sequences against pre-aligned, representative family reference datasets (e.g. IGHV3 or IGKV1 mapped to ABHAND classes) using a configurable conservation percentage threshold (defaulting to 60%). The interface dynamically adapts its table header labels ("OMES Sum" → "Cons. Sum", "Max OMES" → "Max Cons."), input fields, sliders, and severity score badges (using violation count cutoffs: Good ≤ 2, Warning 3-5, Severe > 5) in real-time. Added a Region Scheme select dropdown (supporting North, Kabat, IMGT, Aho, and Martin) directly below Covariance Model in the sidebar, allowing dynamic recalculation and alignment of framework/CDR boundaries on the fly. Updated Covariance Model option labels to Residue (OMES), ABHAND (OMES), and ABHAND (Kannan), and added corresponding documentation to Covariance.md.

• Salt Bridge Pairing in First Pass Optimize: Updated the First Pass Optimize (FPO) tool to automatically group and pair recommended mutations at heavy chain IMGT positions 106 and 116 when both are recommended to fix a disrupted CDR3 salt bridge. This ensures they are treated as paired mutations (Together) in downstream combinatorial variant design.

2026-07-05

• Decoupled Paired Mutation Setting: Introduced a new Pairing feature in the Engineering workspace. Users can select multiple mutation designs in the grid and group them as a pair set to either Single (only one mutation is allowed at a time in any generated variant) or Together (the mutations must always be made together or not at all in any generated variant). This enables precise control over combinatorial variant generation for linked residues (e.g., salt bridges, unusual cysteines, or fragmentation sites).

• Unified Severe Liability Thresholds: Centralized severe liability thresholds for machine learning models (AbLang, AbLang2, IgBert, and CVV) into a single configuration. This ensures that the 'S' severe indicator in Positional Frequency Analysis (PFA) and the severe selections, colors, and descriptions on the Liabilities dashboard are always perfectly synchronized and easily updated from a single location.

• FPO Disrupted CDR3 Salt Bridge Repair Rules: Added optimization rules to First Pass Optimize (FPO) to automatically fix disrupted CDR3 salt bridge stability liabilities, recommending Arginine (R) at IMGT position 106 if the parent is not R/K, and Aspartate (D) at IMGT position 116 if the parent is not D.

• Restored PFA LLM Grid Numerical Values: Restored the rendering of numerical values in the Positional Frequency Analysis (PFA) grid for both LLM rows (AbLang, AbLang2, IgBert) and standard PFA rows. Configured the numbers to display to one decimal place (toFixed(1)) and rotated them 90 degrees vertically (writing-mode: vertical-rl) using an optimized centered font sizing to fit cleanly within the compact 24px grid cells without running together.

• Aligned PFA Grid Collapse/Expand Buttons: Repositioned the PFA block grid collapse/expand toggle buttons (plus/minus icons) to the far right side of the label columns. Utilizing a flexbox row container ensures that all toggle buttons align perfectly vertically regardless of the varying lengths of block names.

2026-07-04

• Liabilities Region Scheme Alignment: Fixed a region mismatch in the Liabilities Analysis page where stability/ML model liabilities (like AbLang, CVV, IgBERT) displayed region names that did not match the user's selected region scheme (e.g., showing IMGT 66 and 67 as Framework 3 instead of CDR2 when using the North region scheme). These region names are now dynamically resolved from the sequence's numbering data to ensure perfect alignment with the Alignment view.

• Dynamic Region Scheme for CVV Calculations: Integrated dynamic region scheme selection into CVV calculations, replacing the hardcoded region setting. New project analyses will compute and output CVV liability regions matching the project's selected region scheme (e.g. North).

• Severe Liability Visual Indicator in PFA: Updated the Positional Frequency Analysis (PFA) sequence grid to place an 'S' indicator inside AbLang, AbLang2, CVV, and IgBert liability cells when their values fall into the severe range (AbLang ≥ 5, AbLang2 ≥ 2, IgBert ≥ 2.5, CVV ≥ 2). The 'S' indicator is now checked and rendered based on severe stability thresholds directly, ensuring it displays even if a standard liability prioritizes a different cell background color. Matched the 'S' font color dynamically to ABHAND coloring standards (white font on red/purple/magenta backgrounds and black font on yellow/green backgrounds) to ensure consistent readability. Configured CVV processing to evaluate the full region model (incorporating CDRs), matching the LLM approach. Ordered the mouseover tooltips to display liabilities as AbLang, AbLang2, IgBert, and then CVV (following standard liabilities), reporting the value directly (e.g. AbLang (7.3)). Added the covariance-based humanness (CVV) liability color and indicator description to the "Liability Color Lookup" legend, and sorted the legend alphabetically.

• Stability Liabilities Cutoff Defaults and CVV Standardisation: Updated the default stability cutoffs on the Liabilities page to reflect the medium thresholds (AbLang: 3.0, AbLang2: 1.5, IgBert: 2.0, CVV: 0.5) and updated their respective reset handlers. Standardized CVV cutoff checking and legends to use >= (greater than or equal to) comparisons (e.g. Low > 0.0, Med >= 0.5, Sev >= 2.0), matching the comparison definition of the other models.

• Stability Liabilities Index Alignment Fix: Corrected index offset alignment (0-based vs 1-based) in templates/liabilities.html for stability models (AbLang/IgBert/CVV) when querying regions, assigning grid coordinates, and creating 3D Molstar selections. Aligned the row key's linear index value in the main table grid (addToPivot) to use 1-based al.LinearPosition + 1 to match the index used by Molstar, resolving a bug where selected stability residues were highlighted in blue at the wrong offset position (e.g. index 65 instead of 66) rather than in red. This fixes position shifts in the 3D Viewer overlay and ensures selected rows map to the correct highlighted residues.

2026-07-03

• First Pass Optimize Grouping and Exclusions: Updated the "First Pass Optimize" engineering tool to group recommended mutations by their specific liability category (e.g., "Unusual Cys", "Stability", "Deamidation", "Isomerization", "Fragmentation", or "Disrupted CDR3 Salt Bridge") rather than using a single generic "First Pass Optimize" group name. For mutations in the "Stability" group, the notes field now appends the specific severe stability methods identified for that site (e.g., "First Pass Optimize recommendation. Stability (IgBert, CVV)"). Additionally, expanded the parental residue retention rules (includeParent: true) to preserve the parental residue at both standard Honegger exclusion areas and classic Vernier zone positions.

• Analysis Report Export Enhancements: Improved the Word document analysis report by dynamically using the active user's company or organization name in the header, applying species-specific background colors to the Project Overview tables, and simplifying the alignment grid labels. Also, colored the covariance-based humanness (CVV) liability cells, included them and custom user-entered motifs in the color standards legend, updated references with ANARCI and IgBert citations, and removed the insert placeholder.

• BiTE Quick Build Template Update: Updated the BiTE (1-chain) Quick Build preset in the Assembler to construct chains using the VH/VL-VH/VL orientation for both domains with a heavy chain leader sequence at the N-terminus, matching user preference for consistent Fv orientation.

• Automatic PLAbDab Search on Load: Enabled the PLAbDab search interface to automatically execute search queries using the default settings (100 maximum hits and all regions selected) immediately when the page is opened, rather than waiting for the user to click the search button.

• Germline Tool Layout Rearrangement: Simplified the Germline tool sequence grid by removing the redundant "Linear" numbering row (keeping the "M. Linear" row). Moved the "Regions" row below the numbering rows in the Leader, Variable, and Constant sections to improve visual clarity and alignment. Updated the "Regions" row label to dynamically include the active region scheme name.

• PFA Row Label & Cell Sizing Alignment: Standardized the Positional Frequency Analysis (PFA) tool sequence grids to match the Germline tool's visual structure. Renamed the "Numbering (Linear)" row label to "M. Linear", simplified the system numbering row label (e.g. from "Numbering (IMGT)" to "IMGT"), left-aligned and styled all row labels, and repositioned the "Clear Section Selection" ban button next to the section title in the block header. Also reduced the PFA cell sizes from 35px to 24px and optimized grid font size (10px) to match the Germline tool's compact aesthetics. To prevent overflow in the smaller cell boxes, numerical values for both LLM probabilities (AbLang, AbLang2, IgBert) and LLM liability diff overlays are hidden from cell text and shown on mouseover tooltip instead. Applied overflow: hidden on all table cells and rounded PFA repertoire frequencies to nearest integer values (e.g. 99.3 to 99) to prevent horizontal text overlapping and merging.

• Humanization Layout & Row Label Alignment: Standardized the Humanization tool sequence grids to match the Germline tool's visual structure. Reduced cell sizes from 35px to 24px and name column width to 280px to accommodate action buttons and badges without text clipping while matching the compact aesthetics of the Germline page. Included species badge overlays in the row labels (e.g. human, mouse, cat) and styled identity percentages to be consistent with the Germline view. Configured row identity percentages to dynamically update to show either Full or Modifiable sequence identity based on the "Sort By" selection. Aligned the row header section labels ("M. Linear" instead of "Mature Linear", "Regions" instead of "Region", and "Query" instead of the antibody base name), ensured left-alignment with a white background on all numbering/region header columns, and placed the "Query" sequence row directly below "Regions" to match the Germline tool's vertical hierarchy. Restored explicit sticky positioning overrides (!important) on row header cells to prevent conflicting grid layout overrides.

2026-07-02

• Optimized Large Project Analysis: Implemented persistent memory caching for sequence-based LLM models (AbLang, AbLang2, IgBert) within the background analysis worker, completely eliminating the disk-load and weight-reloading overhead for sequential batches. Also increased the default analysis batch size from 10 to 50 to optimize database execution speed and reduce transaction overhead on high-capacity servers.

• Export PyMOL in Engineering: Added an "Export PyMOL" button to the 3D Viewer pane header in the Engineering tab. This generates and downloads a self-contained PyMOL Python script (.py) for the currently selected mutation design set. Executing the script inside PyMOL loads the cartoon ribbon colored by sequence region (CDR and framework domains) and displays the designed mutations as element-colored sidechain atoms, grouped in a dedicated Engineering_Mutations selection set.

• Clinical Comparison Surface Properties Fix: Corrected the Clinical Comparison tool to treat structure-based surface properties (SPH, SPP, SPN, and SPCD) as numeric values instead of categorical entries. They are now formatted to exactly two decimal places and compared against the newly updated clinical reference database (which has calculated surface properties) to apply standard color-coded percentiles.

• Bulk Project Export Order & Date Retention: Configured bulk export to index nested archives sequentially, ensuring the selection order is preserved. Updated bulk import to sort nested ZIP files alphabetically and restore the project's original creation date (created_at) from the manifest, keeping dashboard sorting accurate.

2026-07-01

• Bulk Project Re-run & Auto-refresh: Added a "Re-run Selected" action to the File menu on the dashboard. This allows users to select multiple projects from the grid, see a warning modal with the total count of entries to be analyzed, and queue them all for background re-analysis. Also, configured the dashboard to automatically refresh in real-time as soon as any active background analysis reaches completion.

• SaaS License Agreement & Site Disclosure Update: Updated the terms of the AbLead Software-as-a-Service License Agreement and Site Disclosure. The new terms introduce a token-based concurrent access model for account setup, registration, and user-activation flows, replacing the per-user subscription seat model.

• Rename Mutation Set: Added a "Rename Set" action to the Mutation Designs set management dropdown in the Engineering tab. This allows users to easily rename any custom mutation set via a prompt dialog, keeping all mutations and variants intact.

• First Pass Optimize: Added a "First Pass Optimize" menu item under the Edit dropdown in the Results view. This feature automatically scans for severe liabilities (AbLang >= 5.0, AbLang2 >= 2.0, IgBert >= 2.5, CVV > 2.0, PTMs High) for a selected antibody and generates a target mutation design set under Engineering, incorporating specific rules for Cys (to Ser), deamidation (to Gln), isomerization/fragmentation (to Glu), and model-suggested/PFA-derived optimal replacements. It retains the parental residue at standard Honegger exclusion positions, and only applies additional VHH exclusions (positions 42, 49, 50, 52) if the selected entry is a single-chain VHH.

• Preserve Selection in Engineering Mutation Designs: Fixed a bug where selecting mutations in the Engineering workspace and switching tabs or triggering a background data update/refresh would clear the checkboxes. The selected mutation tracker now matches using unique keys based on chain and linear position instead of object references, allowing selection state to persist through data reloads.

• Coordinate Index Alignment: Fixed index offset mapping bugs in both the Liabilities detailed view and First Pass Optimize backend pipeline:

• In Liabilities, reverted rawIdx to utilize the legacy check (isLegacy ? i + 1 : i), since the standard dictionary-format index keys in the results database are already 1-based.
• In First Pass Optimize, corrected pos_map keys to utilize item[0] (which is already 1-based in annotate_and_trim) and removed the incorrect + 1 offset from standard liabilities linear_pos extraction. This resolves discrepancies where boundary residues (like IMGT 34 or position 7) mapped differently between tools.

2026-06-30

• Restore System Motif Defaults & Grouping: Added a "Restore System Defaults" action to the Motif Detection settings to reset system-level developability motifs and features back to their original properties while preserving custom user-defined motifs. Visually partitioned system motifs and user-defined custom motifs into distinct visual sections to improve configuration clarity.

• Bulk Project Export and Import: Enabled exporting multiple selected projects into a single master ZIP archive from the dashboard, as well as importing a bulk ZIP containing multiple project ZIPs at once to restore them in one action.

• Surface Properties PyMOL Export: Added an Export dropdown menu in the Surface Properties dashboard featuring "Export Grid (Excel)" and "Export PyMOL (.py)". Selecting the PyMOL option exports a self-contained Python script (.py) to load the antibody structure in PyMOL, color active residues by the selected surface property (SPH, SPP, SPN, or SPCD), and render a transparent surface with the cartoon ribbon visible underneath.

• Spam Protection for Evaluation Requests: Integrated Cloudflare Turnstile spam protection on the "Request Evaluation Account" page. This adds verification to prevent automated bot signups while maintaining a friction-free experience for legitimate users.

• C-Terminal Deletion Numbering Support: Fixed constant region matching to allow successful alignment and numbering of sequences with C-terminal truncations (such as the K447del deletion).

• Knobs-into-Holes Deduplication: Configured the liabilities scanner to automatically suppress redundant standard KiH Knob and KiH Hole matches when their respective + Disulfide variants are detected.

• Sticky Liabilities Header Fix: Resolved a visual shifting bug in the Liabilities detailed grid where the "Liability" column header cell scrolled horizontally while the rest of the column's cells remained pinned.

• CrossMab Domain-Swapped Numbering: Added full support for numbering and aligning domain-swapped bispecific sequences (CrossMabs) where light chain constant domains are engineered on heavy chains. Swapped constant domains are fully numbered and matched across subsequent domains (Hinge, CH2, CH3) without early truncation. Grouped parallel constant domains by domain priority in the topological sort to prevent column-by-column interleaving (e.g. CH1 and KC mixed) in the alignment and clading visualization grids.

• IgG Hinge Numbering Alignment: Fixed an alignment coordinate shift bug in constant domain numbering for sequences containing gaps in their reference templates (such as IgG2 and IgG4 hinge regions). This ensures that S228P and other hinge modifications are correctly numbered and matched during feature scanning.

• Secure Password Reset: Added a secure "Forgot Password?" reset flow to the login interface. Users can request a timed verification link sent via email to configure a new passphrase, utilizing a unified status message modal to prevent username enumeration.

• Covariance Reference Path Migration: Relocated covariance reference databases from the shared PFA data folder (static/pfa_data/) to a dedicated covariance folder (static/cvv_data/), separating the Covariance tool's data assets from the Positional Frequency Analysis tool.

• Local PDBe Molstar Hosting: Downloaded and hosted the PDBe Molstar JavaScript (pdbe-molstar-plugin.js) and CSS (pdbe-molstar.css) files locally. Updated all 3D structure viewer templates to load assets locally, eliminating external jsDelivr CDN dependency issues.

• Display CDR Violations Toggle: Added a "Display CDR Violations" checkbox setting to the Covariance settings sidebar (default: off). This setting dynamically filters out CDR-associated violations from the alignment grids, summary tables, and Molstar 3D viewer selections.

• Covariance Model Selection Menu: Added a "Covariance Model" settings dropdown to toggle calculations between the standard Residue-level model and the ABHAND-group model, loading their corresponding reference JSON databases (_abhand.json) dynamically. Corrected the validation logic to compare raw residues directly in Residue mode and mapped group classes in ABHAND mode, resolving false positives and incorrect 0.00 scores.

• Calibrated Covariance Severity Ranges & Clinical Statistics: Recalibrated the empirical CVV severity thresholds against the OAS5K database (CVV Full cutoffs: Good ≤ 46.44, Warning ≤ 90.49, Severe > 90.49; CVV FR cutoffs: Good ≤ 8.36, Warning ≤ 22.41, Severe > 22.41). Re-analyzed the Thera-SAbDab human reference set and regenerated the Clinical Comparison statistics in clinical_stats_data.py to align the clinical database with the corrected covariance engine. Updated the KBC weights config (kbc_weights.json) and aligned the clinical comparison guides/attributes mapping.

2026-06-29

• Residue Deletion Support in Engineering: Added support for using the gap character '-' as a mutation in the antibody engineering tool to completely delete the residue at that position.

• Variant Grid Parental Display Character: Updated the Variant Designs table display and Excel export in the Engineering workspace to use '.' instead of '-' for residues that have no change/match the parental sequence. This frees up '-' to be used exclusively as the deletion gap character.

• LALAPG Constant Region Feature Matching: Fixed an incorrect IMGT position index definition in Liabilities.py for the P329G mutation in LALAPG (Feature) (changing position 101 to 99). Enabled deduplication to suppress the redundant LALA (Feature) match when the full LALAPG sequence modification is present.

• Forward Chain Type to Scanner: Updated the liability analysis loop in AbRankOrder.py to forward the active chain_type to the scanner to prevent cross-chain false positives.

• Restricted Cysteine Count Inline Editing: Reverted inline editing permission changes to ensure only # Unpaired Cysteine (severity: high) remains editable on double click, while # Unusual Cysteine and # Cysteine Missing are read-only.

• Salt Bridge Column Sorting: Corrected the canonical sorting order for the Disrupted CDR3 Salt Bridge columns to place the count column (# Disrupted CDR3 Salt Bridge) before the location column (Disrupted CDR3 Salt Bridge).

• Genetic Origin Germline Column Sorting: Sorted germline columns in the project view results table to follow a "Light then Heavy" order: VL Species, VL V-Gene, VL V-Gene %id, VL J-Gene, VL J-Gene %id, followed by VH Species, VH V-Gene, VH V-Gene %id, VH J-Gene, VH J-Gene %id.

• Fix Custom Motifs Column Restorations: Fixed a bug where custom user motifs (such as 'INS') would fail to restore their columns on the project view table due to an unhandled initialization error when checking active settings. Active motifs are now loaded directly from the database record.

• OASign Log Silencing: Silenced verbose binary database loading messages from the OASign library to prevent cluttering stage logs.

• Motif Matching Across Gaps: Implemented a gap-aware regex matcher in Liabilities.py. Custom and system motifs (such as 'INS') are now successfully matched across gaps in aligned sequences (such as I--NS), correctly translating residue indexes back to the gapped alignment coordinate system.

• Fix Deleted Custom Motifs Columns Persistence: Fixed a bug where columns for custom user motifs would remain visible on the dashboard after deletion if they were previously calculated in some entries. Columns for deleted custom motifs are now dynamically filtered out on load.

• Uncalculated Motif NC Placeholder: Missing cell values for active Cysteines, Potential PTMs, and User Motifs (e.g. when a motif was disabled during a past subset run) are now automatically filled with "NC" (Not Calculated) for the count column on the project view table, while the location column remains empty. This avoids displaying "NC" on calculated empty locations.

• Fix Duplicate Column Rendering: Fixed a bug where category-prefixed database keys (such as Potential PTMs: Deamidation (severity: high)) in row dictionaries were processed as new headers during dynamic sync, resulting in duplicate/prefixed columns. The workspace loader now ignores category-prefixed keys.

• Cysteine Count Inline Editing: Enabled double-click inline editing support for all Cysteine count columns (# Unpaired Cysteine, # Unusual Cysteine, and # Cysteine Missing) on the project view results table.

• Fix Dynamic Header Synchronization & Subset Runs: Fixed a bug where subset runs would overwrite/clobber global project headers, causing previously calculated columns to disappear. The backend now merges headers during subset runs and dynamically synchronizes dashboard headers with active settings on load, restoring columns instantly when toggled back on without needing a full re-run.

• Excel Export Custom Motif Formatting: Added support for custom motif and cell background colors inside the main Excel spreadsheet export. Custom motifs are now correctly grouped under the "User Motifs" header, and their count columns receive the standard conditional formatting colors while location cells receive the user-defined motif colors.

• Fix Custom Motif Color Persistence: Fixed a bug where adding or deleting a custom motif would cause the highlight colors of all custom motifs to be stripped and reset to red when compiling the JSON configuration saved to the database.

• Motif Detection Workspace Help Documentation: Created a comprehensive help document docs/MotifDetection.md detailing settings configurations, toggles, positional builds, and highlight color mappings, and registered the page under Project View in the documentation menu.

• Draggable Liabilities Settings Sidebar: Added a resizable grab-handle to the right edge of the settings sidebar panel in the Liabilities view, allowing users to dynamically expand or narrow the settings panel width (from 200px to 600px) to easily view long feature/PTM names.

• Resizable Liabilities Column: Added a resizable grab-handle to the "Liability" column header in the Liabilities view, allowing users to manually expand or shrink the column width. The other sticky columns adapt their layout dynamically as the column size changes.

• Fix Custom and System Motif Merging: Fixed a critical merge bug where saving user settings without positions would clobber system-defined coordinates (such as Electrostatic Steering positions) with an empty list. Positional and regular expression configurations for canonical motifs are now preserved from system definitions, allowing them to be correctly scanned.

• Debug Custom Motif Match Ranges: Fixed an off-by-one error in PTM sequence and IMGT position range reconstruction where exclusive regex matching boundaries were improperly treated as inclusive. This ensures patterns such as PAS{T} correctly match and report VL:PASI:18-21 (4 residues) instead of VL:PASIS:18-22 (5 residues).

• User Settings Menu & Dropdown Layout: Migrated the user settings pages into full-width layout screens (removing the page-level sidebar box completely) and moved the tab selection into a hover-activated dropdown menu on the user profile navigation bar. Users can now click directly to Profile, KBC Scoring Preferences, or Motif Detection from anywhere in the app via the header.

• Custom Motif Detection & Position Checks: Built a new Motif Detection settings screen allowing users to toggle canonical developability motifs (Deamidation, Glycosylation, etc.) and define custom motifs. Supported NCBI/PROSITE nomenclature patterns (e.g. N{P}[ST]{P}) and multi-position residue conditions across Fv and constant domains (e.g., CH3-CHS:88L, CH3-CHS:94S). Custom motifs support low/medium/high severities with conditional formatting, scoring weights, or designation as unscored "Features". Custom motifs are grouped under their own "User Motifs" category in the results grid with names matching system conventions (appending the severity suffix). Added custom conditional formatting count ranges (operators and values for Good, Low, Medium, and High) directly definable and editable within the motif editor. Removed the scoring weight fields from the Motif Detection screen to make KBC Scoring Preferences the single central authority for configuring weights (assigning new custom motifs a default weight of 0.0 on creation). Added a descriptive warning note in Motif Detection pointing to KBC Scoring Preferences for weight configuration. Added an explicit UI help note under the position label clarifying that positional conditions require the IMGT number alone (e.g. 88 or 94) as shown in the Alignment tool. Fixed a critical parsing type bug in the liability scanner (Liabilities.py) to correctly extract the residues list when AntPackHelper.annotate_and_trim() returns a dictionary, resolving scanning failures for custom positional motifs across variable and constant domains. Excluded "Feature" severity custom motifs from appearing as grid columns in the results dashboard, ensuring unscored features only show up as rows in the Liabilities tool. Added a "Constant" region filter option and a dedicated "Features" multiselect list box in the Liabilities tool sidebar to isolate unscored system and user features (with severity "Feature") from standard liabilities. Added a highlight color selector to the custom motif editor/table and dynamically integrated the selected custom color into the Liabilities detailed grid, Molstar highlights, Excel sheet exports, and PyMOL/SVL script exports (with dynamic text-font contrast calculation). Resolved a grid display bug in the Liabilities detailed table where matched custom features without single-letter amino acid sequences would display ? in the grid cells, now correctly rendering the feature base name (e.g., LS) in both web and Excel spreadsheet exports. Fixed a crash (IndexError: list index out of range) when exporting SVL/PyMOL scripts for custom features matching outside the variable domains (e.g. constant regions) by fallback formatting empty numbering cache entries, and added full support to render and color custom user-defined motifs in the generated PyMOL/SVL scripts. Fixed a rendering bug in the Motif Detection settings backend where the color property was not loaded in custom motifs during GET requests, causing color selectors to revert to default values. Added a set of system-level Features (LALA, LALAPG, LS, YTE, S228P, N297Q, N297A, KiH Knob/Hole, KiH Knob/Hole + Disulfide, Electrostatic Steering, and K447del) for Fc effector silencing, stabilization, heterodimerization, and half-life extension, tracking them dynamically as unscored constant region features. Added "Hinge" to the region dropdown menus in the motif settings pages. Implemented self-healing merge logic in Liabilities.py to dynamically combine missing system-level canonical motifs into custom user configurations when initializing the scanner, ensuring newly introduced system features are always scanned regardless of when the settings were last saved.

2026-06-27

• Excel Alignment & Numbering Export Coloring: Updated Excel exports (both per-residue numbering grids and the aggregate sequence alignment sheets) to color constant regions (CL, CH1, Hinge, CH2, CH3) using the domain-specific color palette from colors.py and dynamically applied white/black text fonts to ensure maximum readability on darker backgrounds.

• Alignment View Constant Region Coloring: Updated the Sequence Alignment grid visualization to color constant domain rows by their specific region colors (CL, CH1, Hinge, CH2, CH3) defined in colors.py instead of displaying them in a single generic light blue/gray color.

• Liabilities Structure Viewer Constant Region Coloring: Color the constant (non-selected/non-Fv) regions in the liabilities 3D structure viewer by their specific domain colors (CL, CH1, Hinge, CH2, CH3, Linker) defined in colors.py rather than displaying them in white.

• Constant Domain PyMOL & SVL Export Styling: Added detailed region and cartoon ribbon coloring for all constant domains (CL, CH1, CH2, CH3, Hinge) and Linkers in both PyMOL and MOE SVL exports according to the central color palette. SVL scripts now construct a dedicated constant domains section to select, collect, and style constant regions directly using their mature sequence boundaries.

• User Account Initialization Fix: Fixed a bug where manually created users or users imported via CSV with an assigned initial password bypassed the initial settings configuration and legal terms acceptance. These users are now correctly marked as uninitialized and are prompted to select their timezone and accept the SaaS license agreement/legal terms on their first login.

• PyMOL Script Export: Added support for exporting antibody liabilities as dynamic PyMOL Python scripts (.py files) packaged inside a Zip archive, mirroring the existing MOE SVL export. The scripts automatically query the loaded structure's sequences inside PyMOL and map regional selections (cartoon ribbons for CDRs and frameworks) and spacefilled liability residues dynamically. This ensures 100% correct coloring and labeling regardless of how the PDB file was numbered (e.g. sequential linear, IMGT, Kabat, or custom). The scene is configured with premium palettes, highlighted Cysteine sidechains, and hidden hydrogen atoms.

• Export and Import Custom Metadata: Added full support for custom metadata columns and custom metadata values in project exports. When exporting a project to a .zip archive or downloading it as a .json file, all custom metadata columns and their values are now included. Importing a previously exported project .zip archive now fully restores the custom column definitions and their values on all antibody records.

2026-06-26

• Category Row Sticky Layout Fix: Resolved a layout misalignment in the main results grid category row under Chrome/Blink browsers by styling the table header row container (thead) to be sticky vertically rather than individual th cells. This fixes the horizontal layout collapse where cells with colspan > 1 (such as Genetic Origin) were collapsed into a single column width. Replacing the dynamic loop.first condition with an unconditional check for empty or Molecule Name category entries also guarantees proper alignment between category headers and sub-headers.

• Direct Custom Column Addition: Added the option to create a custom metadata column directly within the "Upload Metadata" dialog (via a new tabbed interface) without needing to upload a CSV or Excel file. Users can specify a name and choose the column's datatype (Text, Integer, or Float).

• Text Format Preset: Added "Text (e.g., %s)" format preset option to the column format settings modal, allowing users to configure and force custom columns to display values as text.

• Column Datatype Visibility: Added current column datatype (Text, Integer, or Float) visibility to the custom metadata column format settings modal.

• Automatic Datatype and Value Coercion: Configured the column format settings modal to automatically update the underlying column datatype in the database schema when changing formats (e.g. changing format to %s sets the type to text), and automatically coerce all existing database records to match the new datatype. Added automatic self-healing to repair mismatched legacy columns on project page load.

• Rearrange Custom Metadata Columns: Enabled draggable custom metadata column headers. Users can drag and drop custom columns to rearrange them within the "Other" group directly in the results grid, with a vertical insertion line indicator showing exactly where the column will be dropped.

• Column Sorting State Preservation: Upgraded the grid sorting persistence to track column headers by their text names instead of indices. This prevents the grid from resorting or losing its sorted state when dynamic columns are added (such as via CSV/Excel metadata upload) or deleted.

• Notes Column Mapping Support: Configured the metadata CSV/Excel importer to map any column named "Notes" (case-insensitive) directly to the native Notes field on the database records rather than creating a duplicate custom metadata column.

• Success Modal Visibility Fix: Resolved a layout nesting issue in the Upload Metadata modal where a missing closing </div> tag caused success alerts and overwrite warnings to render inside the upload modal container. This caused them to remain hidden unless the upload modal was reopened.

• Metadata Overwrite Confirmation Warning: Implemented an explicit warning modal that alerts the user if an uploaded CSV or Excel metadata file contains column names that already exist in the project. The user must explicitly confirm the overwrite before any database records are modified.

• Project View Documentation Update: Updated the Project View documentation (docs/ProjectView.md) to include instructions on uploading custom metadata, inline cell editing, configuring format notation presets (Integer, Decimal Float, Scientific Notation, Custom), column renaming, and column deletion.

• Custom Metadata Column Renaming: Added support for renaming custom metadata column titles directly from the settings popup dialog. Renaming a column updates the project schema definition and automatically migrates the metadata keys on all associated antibody records inside the database.

• Delete Metadata Confirmation Fix: Resolved a modal freezing bug where the delete confirmation modal's Cancel and Delete buttons failed to respond to clicks due to ID collision with the global confirmation modal. Unique results-specific IDs were assigned, and an in-progress spinner was added to indicate deletion status.

• Custom Metadata Scientific Float Formatting: Added standardized formatting for float custom metadata. Scientific notations like 32.4e-8 are mathematically normalized and formatted as 3.24e-7 inside the main interactive grid, Excel exports, and CSV exports.

• Grid Layout and Scrolling Stickiness Correction: Fixed header and column positioning in the main project results grid. Removed inline positioning overrides that were overriding sticky table styles and causing headers to mismatch. Changed the table wrapper container height definition from viewport-based height (100vh) to container-responsive flexbox sizing (flex: 1, min-height: 0), restoring full scrollability and proper layout boundaries across browsers.

• Custom Metadata CSV/Excel Upload: Introduced the ability to upload custom metadata (e.g., assay values, affinity, custom columns) to a project via CSV or Excel file. The system automatically matches rows based on entry name and appends the new fields directly to the main project grid right after the Notes column. These fields support inline double-click editing, robust float-type inference for scientific notation formats (e.g., 1e-09), dynamic column deletion from the grid interface, range filtering, custom sorting, and are fully preserved in Excel and CSV exports.

• Automated IMGT Germline Downloader: Introduced an automated command-line tool to programmatically query and download functional gapped amino acid sequences from the IMGT/GENE-DB database for any specified species. The script handles search criteria, extracts gene selections, requests aligned FASTA sequences with IMGT gaps, and saves them to the species germline directory.

• Chicken Germline Integration: Added support for Chicken (Gallus_gallus / Ck abbreviation) germlines. Processed the new FASTA source files and compiled them into the unified local germline JSON database. Updated the Germline and Humanization workspace layouts, legends, UI dropdown selectors, and product documentation to fully support chicken antibody references.

2026-06-25

• Species Color Hex Capitalization: Uppercased species-specific color hex codes in the central configuration to align with standard styling formats.

• Centralized Species Color Coding: Consolidated duplicate species configurations and introduced centralized, dynamic color coding in the Germline tool sidebar and results grids, including automatic fallback colors for future species additions.

• Feline Germline Integration: Added support for Cat (Felis_catus / Ct abbreviation) germlines. Processed the new FASTA source files into standard IMGT-aligned positions and compiled them into the unified local germline JSON database. Updated the Germline and Humanization workspace layouts, legends, UI dropdown selectors, and product documentation to fully support cat antibody references.

2026-06-23

• Disulfide Stabilization in Assembler: Added a new "Specialty Mutations" option in the Assembler tool containing the option "H44-L100 (IMGT H49-L120)" to introduce an interdomain disulfide bond to stabilize scFv/Fv interfaces. If selected, the system automatically mutates Kabat LC 100 (IMGT L120) and Kabat HC 44 (IMGT H49) to Cysteines.

• Rename Request Test Account: Renamed the request_test_account route, function, and templates to request_evaluation_account throughout the project to maintain terminology consistency with the evaluation workspace features.

• Custom Parts Deletion Bug Fix: Resolved an issue in the Assembler tool where newly created "User Custom Parts" could not be deleted from the sidebar before reloading the Assembler iframe. Also fixed a SortableJS cloning bug where dragging a custom part to a chain row caused the sidebar item to lose its delete click event handler, by migrating custom parts deletion to a robust event delegation listener on the main sidebar container.

• Conditional Formatting Documentation: Added a comprehensive table to the Project View documentation mapping all conditional formatting ranges and thresholds for stability LLMs (AbLang, AbLang2, IgBert), humanness (OASign), surface properties (SurfProp), pI, cysteines/PTMs, disrupted salt bridges, and CDR lengths.

2026-06-22

• Request Evaluation Account & Activation Queuing: Renamed the "Generate Test Account" option to "Request Evaluation Account". Implemented a FIFO activation queue for evaluation account requests when the concurrent active account limit is reached. Queued users do not receive activation links until active slots become available. If a user fails to activate their account within 24 hours of receiving the activation email, their request is dropped and the next queued user is automatically activated.

• Generate Test Account Public Sign-up: Added a "Generate Test Account" option to the public landing page. This page lets visitors register for a temporary, 1-week test account pre-loaded with an Example project (15 diverse mAbs) and a Humanization project (1 mouse Fv). Test accounts have complete access to all calculations, visualizations, and engineering tools but are restricted from importing new sequence/ZIP files or adding new sequences. Account verification links are sent automatically to users' verified email addresses. The registration form embeds the full scrollable Legal Disclosure terms (without requiring a checkbox agreement at sign-up, which is handled during self-activation).

• Test Account Safety Controls & Project Reset: Grayed out and disabled restricted menu items (Import, Add Fvs, and Deletions) in the dashboard and project workspaces for test accounts instead of hiding them. Completely blocked backend deletion APIs (for projects, runs, and antibodies) for test accounts to protect initial demo data. Added a "Refresh to Initial Account Projects" button on the dashboard allowing test account users to delete all custom work and instantly restore their demo workspace to its factory-seeded state.

• Account Invitation & Self-Activation Onboarding: Implemented a secure, invitation-based user creation flow. New users click a secure link, set their own passphrases, accept legal disclosures, and establish their preferences, removing the need for administrators to distribute temporary passwords manually.

2026-06-19

• Company-Level Concurrent Login Limits: Introduced a Company model to group user accounts by domain or direct configuration and set purchased concurrent login seat limits (tokens). The system automatically tracks active logins based on activity, preventing companies from exceeding their purchased tokens. Extended the User Management dashboard with inline company assignment selectors, bulk company assignment actions, and a comprehensive card to create, modify, and delete companies.

• Company Deletion Warning Modal: Integrated an explicit confirmation modal for deleting organizations from the User Management dashboard to prevent accidental deletions and unintended user unassignments.

• Bulk Project Sharing from Dashboard: Added a "Share" action inside the "File" dropdown menu on the main dashboard below "Import Library". This allows users to share multiple selected projects simultaneously. The share modal dynamically adapts to show a summary for multiple selected projects and applies the access permissions to all of them in parallel.

• Security Hardening and Access Control (IDOR Remediation): Conducted a comprehensive security audit of all API endpoints and page routes. Implemented robust authorization controls to prevent Insecure Direct Object Reference (IDOR) attacks across results views, progress status indicators, analysis log downloads, project and antibody workspaces, clading/clinical exports, residue observations, and engineering sets.
• Canine and Pig Germline Architecture Migration: Migrated the platform's closest germline and species classification logic from ANARCI's native databases to our local JSON database. ANARCI is now used solely for structural IMGT numbering alignment, and matches are evaluated locally on a unified sequence identity basis across all species. Added support for Pig (Sus_scrofa) germlines to the compiler, frontend UI, legend indicators (Pg), and help documentation to fully cover all previously supported ANARCI species under the new local architecture. Dropped the 4.5% human/mouse tie-breaker override so that the absolute closest germline match by identity is always chosen, prioritizing human in the event of a tie in percentage identity.
• Passphrase Generation Refactor: Refactored the passphrase generation algorithm to utilize cryptographically secure list sampling from an expanded dictionary of 223 words and increased the default length to 6 words to achieve higher cryptographic entropy (~46.7 bits). Exposed this generator via a new backend API endpoint, allowing the user interface to fetch passphrases directly and eliminating duplicated logic.

2026-06-18

• Canine Germline Integration: Added support for Canine (Canis_lupus_familiaris) germlines in the Germline and Humanization tools. Updated the compiler to process dog FASTA sequences into standard IMGT alignments. Developed a local sequence identity matching fallback to automatically align and match dog sequences since the ANARCI engine does not natively support canine germlines in its HMM profiles. Updated the UI and help documentation to include the new Dog option and legend indicator (Dg). Additionally, updated the Humanization and Germline workspace sorting keys to prioritize and promote canonical-species germlines to the top of alignment tables if they are within the 4.5% tie-breaker override threshold.

• Liabilities Structure Selection Persistence: Fixed a bug in the Liabilities workspace where selecting a different 3D structure for visualization automatically wiped out any custom row selections and selected only the severe rows. Custom row selections are now fully preserved when changing structures.

• Safari and Firefox Sticky Column Layout Correction: Fixed a bug where the left-hand sequence and parameter name columns (such as parent sequence, germline, and liability headers) in the PFA (Positional Frequency Analysis), Covariance, and Humanization grids failed to remain sticky and scrolled horizontally in Safari and Firefox. Removed conflicting CSS flex displays from the table cells (td and th) and transitioned them to standard table-cell vertical and horizontal alignment properties.

• Covariance Violation Navigation Scrolling Correction: Fixed an issue in Safari and Firefox where clicking a covariance violation in the summary table caused the entire main browser page layout to shift/scroll left and cut off side controls. The grid viewport now scrolls directly and centers the violating residue without shifting the main page context.

• Engineering Chain Label Fix: Corrected the layout in the Engineering sequence grid view so that the chain header labels (e.g., Lambda Chain, Heavy Chain) remain static and do not slide/scroll horizontally with the sequence tables, while preserving the single horizontal scrollbar on the main frame viewport.

• Covariance DB Generation Memory Optimization: Optimized the covariance database generator (generate_covariance_db.py) to aggregate single and joint frequencies on-the-fly, reducing memory footprint to a small, constant size. This prevents Out-Of-Memory (OOM) Killed: 9 crashes when processing massive sequence files (such as 30+ million heavy/light chain datasets) with unlimited sample limits.

---

2026-06-17

• Library Evaluation and Validation Import: Introduced a new "Import Library" workspace tool to validate high-throughput sequence files (FASTQ or FASTA) containing nucleotide or protein sequences before creating or importing them into projects. The tool automatically detects nucleotide sequences, translates them across all 6 reading frames to identify the correct open reading frame mapping to a valid antibody domain, evaluates structural domain limits, and presents a dynamic interactive grid showing CDR loops, sequence lengths, and warning flags. Users can filter and select valid clones using Gmail-style deselect behaviors and import them. Supported file formats include .fasta, .fastq, .txt, and compressed .gz formats. Added a real-time progress bar, validation counter (X of Y clones), and a responsive Cancel button to stop validation midway.

• Default Active Filter in User Management: Set the default view in the User Management dashboard to "Online/Active" instead of "All Statuses" to optimize initial dashboard loading and display. Also updated the search filter reset button to return to this "Online/Active" default state.

• Help Documentation Reorganisation: Split duplicate and overlapping navigation menu links in the documentation to ensure that every help page maps to a unique, single-purpose Markdown file. Extracted edit, analysis, and export sub-sections into 11 new documentation files (such as FullReRunAnalysis.md, BulkHumanize.md, AddFvsPDBs.md, etc.), eliminating compilation warnings and improving the documentation's overall navigation. Changed the sidebar navigation scheme so that top-level menu sections (Analysis, Edit, Export, etc.) are collapsible and toggleable, rather than expanded by default.

---

2026-06-16

• Standalone Web App Mode: Enabled true standalone PWA capability for AbLead. Created a web application manifest and added macOS/iOS PWA configurations to hide the settings, address, and navigation bars when the website is installed as a native app on macOS (using Safari's "Add to Dock" or Chrome's install feature).

• Clinical Comparison VHH Handling: Added handling to the Clinical Comparison tool to clearly indicate that comparing clinical-stage antibody reference ranges is not valid for VHH (nanobody) entries. VHH column header cell backgrounds are now colored severe red with black text in both the interactive dashboard and Excel exports, and a corresponding format legend has been added to the right of the Percentiles legend.

• PLAbDab-nano Integration: Expanded the PLAbDab Search tool to support searching against the PLAbDab-nano dataset (VHH and VNAR single-domain camelid/shark nanobody sequences). The tool automatically detects whether the query antibody is a standard Fv (Heavy + Light) or VHH (Heavy-only) and queries the appropriate database, with the user interface dynamically adapting to show either Fv or VHH tabs and information.

• PLAbDab Results Column Layering: Corrected the layering order in the PLAbDab Search view by increasing the z-index of the sticky results/patent metadata column and lowering the selection header layer's z-index. This prevents the selection overlay boxes from rendering on top of and obscuring the patent numbers and results text at both the body and header levels.

---

2026-06-15

• PFA Grid Parent Residue Highlighting: Enhanced the PFA (Positional Frequency Analysis) openable grids in both the PFA and Covariance workspaces to highlight the parental sequence residue value at each position in light red, matching the visual weight of the light green most probable/consensus highlights.

• Multispecific Parts and Quick Build Templates Documentation: Added a detailed section to the Assembler documentation (Assembler.md) detailing the available Knobs-into-Holes (KiH) CH3, CrossMab crossover, electrostatic steering, SEED, and orthogonal charge-pair domains, alongside their corresponding Quick Build templates. The parts are documented using their user-facing interface names (e.g., CH3 (Knob - T366W, IgG1)) rather than technical code identifiers to match the software's UI. Included key academic literature references for Electrostatic Steering (Gunasekaran et al.), SEED (Davis et al.), Orthogonal Fab Interfaces (Lewis et al.), and Fc engineering mutations (LALA, LALAPG, LS, YTE).

• Dynamic Lambda Quick Build Assemblies: Fixed a bug where Quick Build templates in the Assembler automatically inserted Kappa leaders and Kappa constant regions for all Light Chains. The system now dynamically selects Lambda-specific leaders and Lambda constant domains (or crossover equivalent CL Lambda parts) when the corresponding variable region is detected as a Lambda chain.

---

2026-06-14

• Covariance DB Generation Progress Tracking: Added ongoing progress tracking to the covariance database generation script, indicating the file name being evaluated and its percentage progress through the compressed stream.

---

2026-06-13

• Interactive User Management Controls: Streamlined the User Management dashboard by replacing static columns and toggle buttons with real-time dropdown menus. The Status, Admin role, and Read Only status can now be set directly via inline dropdowns that save instantly using AJAX, with visual confirmations (green/red border flashes). Repositioned the Impersonate and Force Passphrase Reset buttons side-by-side to the left of the user's name, clean-up the layout, and removed unnecessary action buttons from the far-right column.

• Surface Properties Charge Dipole (SPCD) Calibration: Calibrated the SPCD calculation to match standard FvCSP/SFvCSP definitions at pH 5.5. Histidine's charge value has been adjusted to \(+0.9\) in the new pH 5.5 charge scale, and the salt bridge distance cutoff has been aligned consistently to \(\le 4.0\) Å in both the calculation engine and documentation.

• Per-Entry Calculation Outdate Rerun Alerts: Added an administrator setting in the System Settings panel to configure a global "Rerun Cutoff Timestamp" (automatically localized to the admin's preferred timezone). Individual entries now record their respective calculation timestamps, and any entry analyzed before the cutoff is styled in red font in the results grid with an explanatory tooltip to alert the user that it needs to be rerun. Rerunning an individual entry will update only its specific timestamp, turning it back to black while leaving the remaining outdated entries styled in red.

• Multispecific Construct Exclusions for Project Outdated Alerts: Enhanced project-level outdated calculation checks on both the dashboard and project details views to correctly exclude multispecific constructs by cross-referencing row names with the database antibody source files. This ensures that projects containing multispecific assembler results are not falsely marked as outdated (red font) on the dashboard when all other individual variable domain/VHH sequences are up to date.

• Results Grid Sort State Preservation: Preserves the active sorting column and direction in sessionStorage. When an analysis re-run completes and refreshes the project data, or when the results view is reloaded, the active sorting selection remains intact instead of resetting.

• Clinical Comparison CSV & Ranges Update: Regenerated reference statistical limits and ranges (clinical_stats_data.py) using the human-specific subset of the updated Thera-SAbDab Jain full-length analysis cohort dataset.

---

2026-06-12

• Clinical Comparison Formatting Fix: Wrapped gene names containing asterisks in inline code blocks within the clinical comparison help documentation to prevent formatting errors and incorrect rendering of italics.
• Clinical Comparison Expanded Reference Set (Thera-SAbDab): Migrated the Clinical Comparison reference stats database to use the newly compiled and aligned set of 588 human clinical antibodies from Thera-SAbDab, updating the display text to match. Removed the redundant "Human Only" checkbox from the workspace interface since the entire dataset is now human-specific.
• Covariance Workspace Scrollbars: Added custom styled horizontal scrollbars to both the Light and Heavy Chain alignment grids and the summary violations table, and switched the summary table layout behavior to automatic. This resolves issues where the residue numbering grid and the Partners column listings were truncated and unscrollable on smaller screens.
• Covariance Framework Score Prominence: Updated the Covariance Settings sidebar to split the display into two separate colored badges: a prominent one for the Framework (FR) Violation Score and a smaller compact one for the Full Violation Score directly beneath it. Each badge dynamically color-codes its background and borders according to its respective severity thresholds (FR cutoffs: Good ≤ 2.34, Warning ≤ 7.82, Severe > 7.82; Full cutoffs: Good ≤ 18.24, Warning ≤ 41.27, Severe > 41.27). Removed the horizontal separation line and tight-spaced the margin between the two boxes for a cleaner layout.
• Session Cleanup on Login and Authentication: Resolved a session impersonation leak where timing out or navigating directly to the login page from an impersonated session, and then logging back in as admin, would mistakenly display the administrator as impersonating "admin". The application now explicitly clears the Flask session state during all login and authentication routes (standard password login, 2FA validation, recovery validation, and WebAuthn/passkey validation) before initializing the new user session.
• Jain and Thera-SAbDab Sequence Comparison: Performed sequence-level comparison between Jain mAbs and Thera-SAbDab analysis datasets to explain discrepant global properties (pI and surface charges), demonstrating that Jain sequences incorporated constant domains (full IgG/Fab format) while Thera-SAbDab sequences contained only variable domains (Fv).
• Molecules Export Utility: Exported the sequences of the 62 Jain clinical antibodies that are absent from the Thera-SAbDab cohort to FASTA and CSV formats.
• Human-only Missing Molecules Export: Exported the sequences of the 35 Jain clinical antibodies with both chains classified as human that are absent from the Thera-SAbDab cohort to FASTA and CSV formats.
• Full-Length Fv-Constant Joined FASTA: Created a utility script to combine the Fv-trimmed human missing Jain entries and the human Thera-SAbDab entries, appending appropriate Kappa, Lambda, and Heavy constant domains (KC IGKC*01, LC IGLC1*01, and HC Full IgHG1*01 respectively) based on their V-gene classifications, and exported the joined dataset as Thera-SAbDab_Jain_Human_FullLength.fasta.
• Germline Alignment Database Rebuild: Rebuilt the germline alignment JSON database (humanization_germlines.json) by running ANARCI on all V germline sequences during compilation. This ensures all database V-region entries map perfectly to standard IMGT unique numbering positions, correcting a sequence alignment shift bug that caused Rhesus matching scores to be calculated incorrectly (e.g., scoring at 29% instead of 90%).
• CDR3 Exclusion Documentation: Updated help files Germline.md and ProjectView.md to explicitly document that ANARCI ignores the CDR3 junction region (positions 105-108) when computing V-region species and germline assignments for the main Project View, whereas the Germline workspace alignments evaluate positions 1 to 108 (including the CDR3 junction).
• Germline Tool Default Species & Priority Sorting: Updated the Germline tool's species selection to default to selecting all available species (e.g., Human, Mouse, Rhesus, Alpaca, Rat, Rabbit) rather than defaulting only to Human. Additionally, matches are now sorted to prioritize the active chain's canonical species when sequence identity scores are tied, ensuring rhesus germlines are prioritized for rhesus-classified chains. Updated user-facing help documentation in Germline.md to describe the expanded list of species and their indicators.
• Clinical Comparison Germline Suffix Matching: Fixed a visual bug where germlines containing species suffixes (such as IGKV1-39*01 (human)) were incorrectly highlighted in red on the Clinical Comparison grid. The categorical comparison logic now extracts and compares suffix-stripped base germline names for both client values and clinical reference values to ensure robust matching.
• LLM Calculation Fix for Structure-Only Runs: Fixed a bug where sequence-based language model predictions (AbLang, AbLang2, and IgBert) were not calculated for subsequent sequence runs. Background worker processes now track whether sequence models have been initialized separately from structural models, preventing structure-only runs (like PDB additions or model building) from blocking sequence model imports in subsequent jobs.
• Germline Human/Mouse Species Prioritization: Implemented an automated tie-breaking prioritization rule in the Germline Assigner. When evaluating multiple species, the assigner now favors human or mouse germlines if their identity scores fall within 4.5% of the absolute top non-human/mouse hit. This prevents human/humanized therapeutic antibodies with minor framework mutations from being incorrectly classified as rhesus, pig, or other species, while correctly preserving primate classification for actual primatized therapeutics.
• V-Gene C-Terminus Alignment Correction: Corrected V-gene alignment logic during germline database compilation to prevent ANARCI from stretching trailing V-region residues (such as the conserved Arginine at the end of the CAR motif) to position 117. Trailing residues after Cysteine 104 are now correctly positioned sequentially starting at 105, ensuring V-germline residues align perfectly in the Germline analysis grid.
• Germline Tool Default Species Selection: Reconfigured the default selected species on page load in the Germline tool from "all species" to "Human" and "Mouse" only. This resolves slow page load times by preventing the backend from aligning the selected antibody against thousands of other species' V/J germlines on initial load, while still allowing the user to select them in the settings panel.

---

2026-06-11

• PDB Files Selection Counter: Added a permanent selection counter to the PDB Files manager toolbar styled like the main results dashboard. The counter displays the total count of PDB files when no items are checked (e.g. 568 PDB Files), updates dynamically to show selection status (e.g. 3 of 568 Selected), and is initialized immediately upon page load.
• Surface Properties Predicted Structure Sequence Tolerance: Fixed a sequence mismatch error (PDB/FASTA Fv sequences mismatch after trimming.) that occurred during Surface Properties calculations for ML-predicted structural models (such as Zinlirvimab and Zovostotug). The sequence validation algorithm now tolerates N-terminal and C-terminal trimming introduced by structural prediction pipelines.
• ABodyBuilder2 Model Building Validation Bypass: Implemented runtime monkey-patching in AbRankOrder.py to relax ImmuneBuilder's strict residue numbering checks. Conserved IMGT boundary assertions (e.g. min(numbers) < 8 and max(numbers) > 120) are bypassed when they reject valid antibody sequences due to unusual framework alignments (e.g. Zinlirvimab light chain) or long CDR3 loops (e.g. Zovostotug heavy chain), allowing structural models to build successfully.
• Structural Model Building PDB Harvesting Fix: Corrected a bug where newly built 3D structure models were not saved to the database. Updated the final database harvesting loop to search within the structure-only antibodies_with_new_pdbs list (instead of relying solely on the sequence-only antibodies_to_analyze list), and corrected the setup_complete callback payload to properly return the list of built structures so they are analyzed immediately.
• Structural Model Building Subset Filter Fix: Fixed a bug where ABodyBuilder2 failed to build 3D models for selected project entries during structure-only runs. Enabled run_model_building() to recognize and respect structure_subset limits alongside sequence analysis_subset constraints, ensuring model building successfully builds structures for the subset of targeted entries.
• Analysis Progress Modal Cancel Button Fix: Fixed an issue where the "Cancel Job" button could leak and remain visible in the progress modal footer after a run completes successfully or fails, particularly during zero-item runs. The button is now explicitly hidden at the start of modal initialization.
• Germline Assignment Warning Silence: Suppressed verbose and distracting ANARCI Limiting hmmer search warnings from the console output during germline species assignment. Wrapped all internal ANARCI calls in GermlineAssignment.py with stdout redirection to ensure logs and import progress messages remain clean and focused.
• Assembler CrossMab Hinge Fix: Removed the incorrect automatic insertion of the EPKSC upper hinge sequence at the junction between the crossover CL domain and core hinge domain in CH1-CL CrossMab assemblies. This ensures heavy crossover chains are correctly assembled as VH-CL-Hinge, aligning with biologically validated clinical standards (such as Vanucizumab).
• Clinical Comparison CVV Metrics Integration: Integrated covariance violations (CVV Full and CVV FR) into the Clinical Comparison dashboard, enabling users to evaluate sequence-level covariance anomalies against clinical development therapeutics using reference percentiles and ranges derived from the Jain mAbs dataset.
• Confirmation Modal Customization & Cancel Styling: Fixed button label crossover and styling in the confirmation modal. Destructive or cancellation actions now display with grey secondary ('Keep Running') and red primary ('Cancel Job') styled buttons, and all button labels/colors are correctly reset dynamically to prevent text leaks from previous modals.
• Queued Run Cancellation: Enabled cancellation of queued background runs (waiting in the serial orchestrator queue) through the progress modal cancel button. Queued runs marked as canceled are now automatically skipped when pulled by the orchestrator.
• Log Polling Deduplication: Resolved a client-side JavaScript bug where overlapping asynchronous network requests from concurrent polling could cause progress logs to duplicate. Refactored the polling loop in templates/base.html from setInterval to a recursive setTimeout pattern, ensuring a new polling request is only scheduled after the previous request has fully resolved. Stored the timeout ID in a global reference and hoisted clearActivePoll() to guarantee clean cancellation.

---

2026-06-10

• Inactivity Timeout Online Status Sync: Fixed an issue where users whose sessions timed out remained marked as "Online" in user management. Updated the backend login_required middleware in routes/auth_utils.py to set the user's last_activity status to None in the database when enforcing an inactivity timeout, mirroring a manual logout.
• CL-CH1 CrossMab Preset Enhancement: Aligned the CL-CH1 CrossMab preset template with biologically validated architectures (such as Faricimab and Vanucizumab). Configured the light chain crossover to end in +EPKSC (crossmab-ch1-cl-lc), the heavy chain crossover to use the Hole Fc variant with a new truncated hinge starting with DKTHT (hinge-dktht-g1), and the standard heavy chain to use the Knob Fc variant. Added a new ASVA-optimized heavy crossover CL Kappa part (crossmab-ch1-cl-hc-kappa-asva) to minimize V-C interface steric strain, and embedded a notice banner warning that presets serve as starting points only.
• CrossMAb Hinge Assembly Fix: Corrected a sequence truncation bug in the constant crossover (CH1-CL) Assembler. Implemented automatic insertion of the 5-amino-acid upper hinge spacer (EPKSC) when a crossover constant domain ending in RGEC is connected to a core heavy hinge sequence starting with DKTHT (hinge-dktht-g1), preventing expression failure and steric clash.
• Bug Fix for CL-CH1 CrossMab Presets: Corrected the preset domain search query for Hinge regions in the Assembler. Fixed a query name mismatch (from "Hinge IGHG101" to "H IGHG101") and made the category lookup function robust against empty/missing arrays, restoring complete chain construction including variable domains (VL/VH) and correct constant domain layouts for CrossMab templates.
• Dynamic Fc Mutations and Preset Templates: Integrated dynamic Fc mutations (LALA, LALAPG, YTE, LS) in standard and multispecific build modes to apply modifications on-the-fly, avoiding combinatorial library clutter. Added a "Quick Build" presets dropdown (e.g. Fv CrossMab, CL-CH1 CrossMab, scFv-Fc, BiTE) to instantly layout chains and pre-populate heterodimerization, crossover, and linker domains with automatic signal leader sequence insertion in the Assembler.
• Expanded Multispecific Parts Configuration: Added new engineered constant region domains to static/config/multispecific_parts.json including Electrostatic Steering (charge pairs DK/KK), Strand-Exchange Engineered Domains (AG/GA SEED), and Orthogonal Fab interface mutations (CH1/CL Kappa) to simplify modular multispecific antibody design in the Assembler.
• Obsolete CrossMab Part Removal: Removed the obsolete truncated light crossover part CrossMab CH1-CL -EPKSC (crossmab-ch1-cl-lc-no-epksc) from the multispecific parts JSON configuration to prevent unoptimized crossover assemblies.
• Unlimited Max-Total in Covariance Script: Updated generate_covariance_db.py to support setting -m / --max-total to 0 or -1 to process files without enforcing a total sequence limit.
• Updated Help Documentation: Added dedicated sections for 'CH2/CH3 Mutations (Dynamic Fc Mutations)' and 'Quick Build' templates/presets to the Assembler.md help documentation, including foundational scientific literature citations for CrossMAb, Knobs-into-Holes, and BiTE formats.
• Batch Mode Chain Label Counts: Added dynamic counts to the Kappa Chain, Lambda Chain, and Heavy Chain labels in Batch Mode to display the number of domains mapped from the total number of selected cohort entries (e.g. "Kappa Chain (2 domains from 4 selected entries)").

---

2026-06-09

• Multi-Species Germline Species Assignment Fix: Resolved an issue where ANARCI failed to evaluate alternative species germlines (such as alpaca) when "all" species were selected, incorrectly defaulting to human germlines. Updated the germline assigner tool to explicitly pass the full allowed species list to the ANARCI library during "all" species searches, and expanded the supported species database configuration to cover all species supported by ANARCI including cow. Prioritized the species of the actual closest V-gene germline over the raw HMMer hit species, ensuring human entries are correctly called human rather than rhesus or alpaca.
• Surface Properties Region Annotations for VHH: Fixed region scheme mapping and index alignment in the Surface Properties route. Pulls system defaults in lowercase and runs AntPackHelper.annotate_and_trim() directly on the PDB-derived sequence (full_seq) with include_leader=False, ensuring 1:1 parity between annotated residues and coordinates. Maps linear to the AntPack raw_idx and pdb_num to the renumbered idx + 1 to match the sequentially renumbered PDB coordinates served to Molstar. This completely resolves coordinate index shifts, wrong CDR/Framework region assignments on the ribbon, and mutation indicator misalignment on VHH and Fv structures.
• Solvent Exposures Workspace Enhancements: Updated residue mouseover tooltips in the Solvent Exposures workspace to display the selected numbering system name next to the position value (e.g., “IMGT 75”), ensuring consistency with other analysis workspaces. Changed the workspace tab label from “Solvent: [entry]” to “Solv. Exp.: [entry]” to prevent tab label crowding.
• PDB Files Loading Indicator: Added a visual loading spinner to the PDB Files manager workspace to indicate progress during initialization and model parsing.
• Surface Properties 3D Viewer Header: Added a descriptive "3D Structure (Mature Linear Numbering)" title header above the Molstar viewport in the Surface Properties tool to match layout consistency with Covariance and other structural tools.
• Residue Clustering & Neighborhood Highlighting: Added interactive 3D residue clustering display in the Surface Properties workspace. Selecting a residue now highlights all of its physical spatial neighbors (within 7.5 Å) in the Molstar 3D viewer as sidechain ball-and-stick representations, with optional static 3D labels.
• Surface Properties Excel Export: Added a toolbar button to export the detailed surface properties (SPH, SPP, SPN, and SPCD scores) for both variable chains to a styled Excel spreadsheet, including full color mappings for regions, amino acid classes, and score ranges matching the UI design. Added "Mature Linear" and "3D Neighbors" columns directly inside the main summary detailed table, kept the export simplified to a single worksheet, and removed background color fills from all zero (0.0) cells across all metrics to prevent spreadsheet color clutter. Resolved a 413 Request Entity Too Large payload size crash by stripping the heavy coordinates string (pdb_data) from the client-side POST form before submission.
• Surface Properties Indicator Support: Enabled dynamic display of observations and mutation designs on residue cells inside the VL/VH grids (excluding the sequence list track), including full automatic refresh capabilities upon saving mutations/observations via the Engineering modal.
• Surface Properties Legend Ranges: Embedded numerical score bounds directly inside the color scale labels in the Legend section (0.0 to 4.0+ for SPH, 0.0 to 2.0+ for SPP/SPN, and -2.0 to +2.0 for SPCD). Fixed a CSS gradient tiling glitch that caused a thin vertical color line to appear on the left edge of scale bars by setting background repetition to inactive.
• Surface Properties Numbering Alignment: Resolved a scheme mismatch in the Surface Properties analysis tool by configuring the backend calculations to retrieve and respect the project-specific numbering and region schemes from the project's analysis results (falling back to system config defaults). Recalculates single chain (VL/VH only) surface properties in isolation by detaching the unselected chain before executing the solvent accessibility (SASA) algorithm, enabling the newly exposed interface residues on the isolated chain to be calculated and displayed. This ensures residue annotations, labels, and regions align perfectly with the main results dashboard.
• Surface Properties Workspace Enhancements: Restructured grid pane headers to position the deselect button containing the ban icon (fa-ban) directly to the left of the chain title to match other analysis tools. Filtered the sequence panel and the residue grids to only show residues in the variable domain (Fv), excluding constant and leader regions. Configured the sequence alignment panel to compress without wrapping by laying out in a horizontal scrollable row. Added vertical system numbering (rotated 90 degrees) above the sequence track columns, and updated mouseover tooltips to display the selected numbering system name alongside the residue position matching Humanness. Preserved both the VL and VH grids side-by-side at all times (preventing layout collapse when switching chain modes). Added a "Show Both Chains" checkbox (enabled by default) under the Chains Display settings panel, which controls whether the non-selected chain's ribbon cartoon remains visible in the 3D Molstar viewer. Fixed a bug where the non-selected chain's ribbon cartoon disappeared in the viewport when a single chain was selected by mapping sequential linear indices to actual PDB coordinates (pdb_num) for Mol* selections. Optimized structure reloading to skip re-initializing the 3D viewer when coordinates are unchanged, enabling instant, flicker-free chain display toggling while preserving camera coordinates.
• Surface Properties Analysis Tool: Added a comprehensive Surface Properties analysis tool under the Analysis menu, offering per-residue calculation and 3D visualization of surface hydrophobicity (SPH), positive charge patches (SPP), negative charge patches (SPN), and charge dipole distributions (SPCD). Added static 3D text labels to active and selected residues inside the viewport matching the format in the Liabilities workspace, and created detailed user help documentation (docs/SurfaceProperties.md).
• Surface Properties 3D Viewer & Selection Enhancements: Refactored Molstar 3D rendering to color structure backbones by region and sidechain/atoms by property value (preserving property colors in spacefill view, and coloring SPCD dipole charges blue or red based on \(+1\)/\(-1\) charge values) or blue selection highlight. Grouped backbone cartoon selections into contiguous region ranges for reliable and correct rendering in the Mol* viewer. Restructured the viewer coloring loop to strictly limit the display of active residue sidechains (ball-and-stick or spacefill atoms) to the selected chain(s) based on the chain selection mode, while displaying the cartoon representation colored by region for the entire Fv ribbon cartoon regardless of chain selection. Aligned table row selection borders with standard results/Covariance grids exactly (using continuous blue bottom borders and side inset shadows), added a "Select Severe" button with subtext cutoff details, added a "Deselect" clear button for each VL and VH grid, embedded the ABHAND amino acid color legend, and resolved Molstar z-index overlap when expanded. Fixed PDB structure reloading in the PDBe Molstar wrapper when switching chain modes (e.g., to VL only) by clearing the canvas container and re-initializing the viewer plugin. Adjusted nonSelectedColor to light grey ({ r: 220, g: 220, b: 220 }) to ensure constant regions and other non-active structures remain fully visible instead of blending into the white background.
• Unlimited Covariance Database Sampling Option: Configured generate_covariance_db.py to allow unlimited sampling per V-gene and disable early stopping when --sample-limit is set to 0 or -1, enabling users to process entire input datasets without constraint.
• Residue & ABHAND Dual Output & Compression: Updated generate_covariance_db.py to run both standard per-residue (20 amino acids) and ABHAND (6 physicochemical classes, with Glycine mapped to deletion) covariance calculations in a single pass, outputting compressed JSON files (removing whitespaces and newlines) named <output>.json and <output>_abhand.json respectively.

---

2026-06-08

• Integrated Residue-Level Covariance Violations (CVV): Embedded residue-level covariance highlights directly in the Liabilities dashboard and export tools, introducing unified filtering and coloring matching the Covariance Analysis workspace.
• Enhanced Covariance Workspace Layout: Added a collapsible settings sidebar for maximizing screen real estate, aligned grid row styling with the PFA track layout, and introduced automated mutation details loading and real-time indicator refreshes when editing sequences.
• LaTeX Math Rendering in Help Docs: Enabled native browser-side LaTeX rendering via MathJax across all system help documentation, providing high-fidelity mathematical equations and threshold formulas.
• Refined Table Alignment and Grid Scaling: Optimized cell spacing, consensus highlights, and text sizing between the PFA and Covariance alignment grids to ensure a unified visual design.

---

2026-06-07

• AJAX Inline User Expirations: Added support for inline expiration dates editing with real-time automatic user deactivations inside the administration dashboard.
• Empirical Validation Documentation: Added comprehensive developability drop-off threshold reports and correlation scatter plots based on clinical Jain dataset statistics.
• Closest Human Germline in Covariance: Integrated closest human germlines mapping to correctly evaluate covariance partner constraints for non-human sequence targets.

---

2026-06-06

• Region and Vernier Columns in Covariance: Appended region labels and classic Vernier zone markers to active covariance violations summaries.
• Grid and Violations Row Sync: Implemented interactive selection highlighting and scrolling synchronization between alignment grid columns and summary violation rows.
• Closest Germline Row in PFA and Covariance: Embedded closest human germline sequence tracks with fading match styling directly beneath query residues.

---

2026-06-05

• OMES Sum Severity Grading: Aligned covariance severity metrics to be evaluated by cumulative partner OMES scores instead of raw counts.
• Multi-Severity Toggling controls: Added additive select/deselect settings buttons for Low, Medium, and Severe covariance violations.
• Optimized Covariance Databases: Re-ran the database pipelines using a streaming gzipped sequence reader and early-termination controls to create uniform reference models.

---

2026-06-04

• Sandbox Mutation Panel Updates: Implemented sandbox edit sequences with dynamic "Repaired" vs. "New" violation statuses, custom alert modals, and precise OMES threshold overrides.
• AbLang/IgBert Grid Integrations: Corrected API data matching paths to restore language model grids in the Covariance workspace.

---

2026-06-03

• Three-Way Grid Column Sorting: Added Ascending / Descending / Default sort cycles across results dashboards, user management tables, and engineering grids.
• Refined Residue Indicators: Standardized high-contrast eggshell (mutations) and orange (observations) markers with black outlines across workspaces.
• PDB Files sorting and filtering: Implemented popover search filters and sequence sorting inside the PDB database manager.

---

2026-06-02

• Fv Only Mutation Grid Toggle: Added a filter toggle to evaluate only variable domain sequences, excluding leader and constant regions.
• Molstar 3D Viewer in pI Combinations: Integrated PDBe Molstar structure alignments to visually map target surface mutants and combination designs.
• Auto-Saving Fv Limits: Swapped manual admin buttons with immediate inline AJAX auto-saving for user build allowances.

---

2026-06-01

• Optimized Humanization & Repair Queueing: Upgraded backend queues to target only newly created construct records during bulk humanization or repairs, saving computing resources.
• Honegger & VHH Humanization Legend Options: Visual overlays and selection tags representing Honegger and VHH residues within humanization alignments.

---

2026-05-31

• Refined Exclusions interface: Redesigned humanization exclusions panel with mutually exclusive radio selectors and inline custom CDR definition rules.

---

2026-05-30

• Account Expiration Boundaries: Implemented administrator-controlled user account expiration dates, including warning notices, login popups, and automatic access lockouts.

---

2026-05-29

• Molstar 3D Structure Viewer in Engineering: Pinned interactive PDBe Molstar structure viewers dynamically color-coded by region into the Engineering and pI workspaces.
• High-Severity Liabilities Shortcuts: Added a "Select Severe" settings button to instantly highlight candidates with high-risk liabilities.

---

2026-05-28

• Alpaca Germline Matching: Added VICUGNA PACOS germline V and JH matching database supports.
• Engineering Grids Column Sorting: Upgraded Variant Designs tables with interactive column sorting and bold header formatting.

---

2026-05-27

• Multi-Row Liabilities Selection: Interactive row highlighting carrying synchronized PDBe Molstar spacefill overlays.
• Dynamic Constant Domain Repair: Dynamic VL/CL split-codon junction warnings and automated insertion/deletion repair controls.

---

2026-05-26

• PFA Germline Fallback: Smart fallback mapping to closest reference human germlines when primary determined candidates lack database frequency scores.
• Surgical Workspace Grid Refreshes: Replaced full-page reloads with targeted iframe updates when saving structural models and mutation designs.

---

2026-05-25

• Bulk Notes Edit: Inline menu options to batch append or overwrite notes for selected sequences.
• Recalibrated IgBert Scoring: Concatenated heavy/light sequences to utilize IgBert paired-attention inference with updated warning thresholds.

---

2026-05-24

• VHH 3D Modeling Support: Integrated NanoBodyBuilder2 to automatically build heavy-chain-only nanobody models.
• Surface Mutagenesis Workspace: Restored the surface residue editor displaying sidechain SASA levels alongside custom representation styling.

---

2026-05-23

• Model-Specific Cutoffs in Liabilities: Replaced unified cutoffs with independent sliders for AbLang, AbLang2, and IgBert thresholds.
• Impersonate User Option: Added administrative impersonation toggles within the User Management dashboard.

---

2026-05-22

• IgBert Paired Scoring: Added paired-chain stability profiling across workspaces, Clinical Comparison modules, and exports.
• CSP Build Button Compliance: Refactored inline JS button handlers to programmatic event listeners to satisfy Content Security Policies.

---

2026-05-21

• Analytics & Consent Management: Integrated Cookie Consent Banners and Google Analytics 4 tracking.
• Account Deletion Safeguards: Database cascade deletion protections to prevent administrator lockouts.

---

2026-05-20

• Passkeys (WebAuthn): Biometric and hardware security key sign-ins from user profiles.
• Bulk User Import: CSV administrative uploads with temporary hyphenated word passphrase assignments.

---

2026-05-19

• Multispecific FASTA Imports: Automatic grouping of multispecific antibodies.
• Fv Pair Extraction Utility: Segment variable domains from multispecific IgG candidates for developability scoring.

---

2026-05-18

• Multispecific Assembler Workspace: Drag-and-drop peptide linkers, Knobs-into-Holes (KiH) domains, and CrossMab crossover parts.

---

2026-05-13

• Germline Performance Caching: Cached database lookup indexes to reduce page loading delays.

---

2026-05-12

• Molstar in PDB Manager: Added structural overlays directly inside the PDB Files manager dashboard.
• Multi-Species Germline Targets: Integrated Human, Mouse, Rat, Rabbit, and Rhesus databases.

---

2026-05-11

• Allotype Detection: Automatic constant domain allotype annotation.
• Germline Engineering Mutator: Selection modal and indicator dots integration within the Germline view.

---

2026-05-10

• Automated Fv 3D Modeling: Batch structural modeling utilizing ABodyBuilder2.
• Germline Analysis view: Split V and J germline identity and mapping views.