Immunogenicity Prediction

The foundational layer of immunogenicity screening. These models predict the binding affinity between a peptide fragment (typically a 9-mer) and a specific Major Histocompatibility Complex (MHC) allele.

• Allele Coverage: Pan-specific models like NetMHCpan are trained on binding data across multiple alleles, allowing predictions for HLA variants not seen during training.
• Mechanism: The model scores each overlapping peptide in an antibody sequence for its likelihood to be presented on the cell surface.
• Output: A ranked list of potential T-cell epitopes, often expressed as an IC50 value or a percentile rank against a background of natural peptides.

Peptide-MHC binding is necessary but not sufficient for immunogenicity. The presented peptide must also be recognized by a circulating T-cell receptor (TCR).

• Immunogenicity vs. Presentation: This methodology filters pMHC complexes to identify those likely to be immunogenic, not just presented.
• Models: Tools like IEDB T-cell Class I pMHC Immunogenicity Predictor use amino acid physicochemical properties at specific peptide positions to predict TCR recognition probability.
• Key Insight: Positions P4-P6 of the peptide are critical for TCR contact; non-self-like residues at these positions strongly correlate with immunogenicity.

A computational mimic of thymic negative selection. This approach filters out peptide sequences that resemble self-peptides to which the immune system is already tolerant.

• Reference Proteome: Candidate epitopes are BLASTed against the human proteome.
• Tolerance Induction: Peptides with high sequence identity to human proteins are flagged as likely tolerized and removed from the risk list.
• Germline-Encoded TCRs: Some algorithms specifically check for motifs recognized by germline-encoded TCRs that have been deleted during T-cell development, ensuring the predicted epitope has a naive T-cell repertoire available to respond.

Integrates multiple orthogonal predictions into a single, holistic immunogenicity risk score for a full antibody sequence.

• Composite Metrics: Algorithms like EpiMatrix or ISPRI combine pMHC binding, TCR recognition, and humanness scores into a normalized risk index.
• Tregitope Modulation: Advanced scoring accounts for the presence of T regulatory cell epitopes (Tregitopes) that can actively suppress immune responses, potentially offsetting the risk from effector T-cell epitopes.
• Benchmarking: The aggregate score is compared against a distribution of scores from antibodies with known clinical immunogenicity rates to contextualize risk.

Immunogenicity is not solely sequence-driven. Aggregated antibody drug product can break B-cell tolerance and provoke an anti-drug antibody (ADA) response independent of T-cell help.

• Spatial Aggregation Propensity (SAP): Calculates the solvent-exposed hydrophobic patches on the antibody's 3D structure. High SAP scores correlate with aggregation and immunogenicity.
• Computational Models: Tools like Aggrescan3D or CamSol perform structural analysis to identify aggregation-prone regions (APRs) that should be engineered out.
• T-Independent Response: This methodology addresses the risk of T-cell independent immunogenicity, where repetitive antigen arrays on aggregates cross-link B-cell receptors directly.

The primary driver of long-term anti-drug antibody (ADA) responses is T-helper cell activation via MHC Class II presentation.

• Mechanism: Antigen-presenting cells (APCs) internalize the therapeutic antibody, process it into peptides, and load them onto MHC-II molecules.
• Prediction Tools: Specialized models like NetMHCIIpan predict promiscuous binding across HLA-DR, HLA-DQ, and HLA-DP alleles.
• Clinical Relevance: A high density of predicted MHC-II epitopes correlates strongly with ADA incidence in the clinic, making this a critical filter for candidate selection.

The process of locating specific peptide sequences within a therapeutic antibody that can be recognized by CD4+ T-helper cells. Not all MHC-binding peptides are T-cell epitopes; the peptide-MHC complex must also be recognized by a cognate T-cell receptor (TCR).

• In silico screening: Sliding window approach scans the entire antibody sequence for MHC-binding peptides
• Germline filtering: Removing peptides identical to human germline antibody sequences, as these should be tolerated
• T-cell activation assays: In vitro validation using donor PBMCs to confirm predicted epitopes
• Immunodominance: Some epitopes dominate the immune response while others are subdominant or cryptic

Key tools include IEDB T-cell epitope prediction tools and EpiMatrix, which aggregate MHC binding predictions into a composite immunogenicity score.

The computational and experimental process of reducing immunogenicity by replacing murine framework regions with human germline sequences while preserving the complementarity-determining regions (CDRs) responsible for antigen binding.

• CDR grafting: Direct transplantation of murine CDRs onto a human framework
• Back-mutation analysis: Identifying framework residues that influence CDR conformation and reintroducing murine residues at critical positions
• Vernier zone residues: Framework positions that adjust CDR loop positioning
• Humaneering: A proprietary technology that uses homologous human frameworks to minimize back-mutations

Humanization directly reduces the number of foreign T-cell epitopes by replacing non-self framework peptides with sequences recognized as self by the human immune system.

A multi-parameter computational evaluation that extends beyond immunogenicity to assess the manufacturability and clinical viability of antibody candidates. Immunogenicity risk is one component of a broader developability profile.

• Aggregation propensity: Predicted using spatial aggregation propensity (SAP) and aggregation-prone region (APR) algorithms
• Chemical stability: Identifying deamidation (NG, NS motifs), oxidation (M, W residues), and isomerization (DG, DS motifs) hotspots
• Solubility prediction: Using CamSol and Solubis to forecast concentration-dependent behavior
• Polyreactivity screening: Detecting non-specific binding that correlates with poor pharmacokinetics

Therapeutic Antibody Profiler (TAP) integrates these metrics into a developability index, enabling early-stage triaging of candidates with both high immunogenicity risk and poor biophysical properties.

The use of next-generation sequencing to profile the vast diversity of B-cell and T-cell receptors in an organism. This data provides the biological ground truth for training and validating immunogenicity prediction models.

• B-cell receptor (BCR) repertoire: Captures the diversity of circulating antibodies, including those from healthy donors and patients with anti-drug antibody responses
• TCR repertoire sequencing: Identifies T-cell clones that recognize therapeutic antibody-derived epitopes
• Single-cell paired sequencing: Captures cognate heavy and light chain pairs from individual B cells
• Public vs. private sequences: Distinguishing commonly shared antibody sequences from unique individual repertoires

Repertoire data enables the identification of convergent antibody responses—shared anti-drug antibody sequences across multiple patients—indicating immunodominant epitopes.

Not all T-cell epitopes are detrimental. Tregitopes are peptides that activate regulatory T-cells (Tregs), which suppress immune responses and promote tolerance. Computational identification of Tregitopes is an emerging strategy for engineering tolerance to therapeutic proteins.

• Tregitope sequences: Highly conserved peptides found in human IgG constant regions
• Mechanism: Tregitopes expand Treg populations that suppress effector T-cell responses to co-administered antigens
• Engineering applications: Incorporating Tregitope sequences into fusion proteins or gene therapy vectors to induce antigen-specific tolerance
• EpiVax Tregitope database: Curated collection of validated Tregitope sequences

This concept inverts the traditional immunogenicity paradigm—rather than simply removing T-cell epitopes, tolerogenic epitopes can be actively introduced to promote immune acceptance.