PSSMHCpan

PSSMHCpan predicts peptide binding affinity to HLA class I alleles to enable identification of neoantigens for cancer immunotherapy.

Key Features:

• Position Specific Scoring Matrix (PSSM)-Based Approach: PSSMHCpan uses a Position Specific Scoring Matrix (PSSM) methodology to predict peptide–HLA class I binding affinities.
• Broad Coverage: Predicts binding for a wide range of HLA class I alleles, including HLA-A*0202, HLA-A*0203, HLA-A*6802, HLA-B*5101, HLA-B*5301, HLA-B*5401, and HLA-B*5701.
• Performance Evaluation: Trained and evaluated with 10-fold cross-validation on a training database of 87 HLA alleles, achieving average AUC of 0.94 and ACC of 0.85.
• Comparison with Existing Tools: In independent evaluations on the Peptide Database of Cancer Immunity, achieved sensitivity 0.90 and outperformed NetMHC-4.0 (sensitivity 0.74), NetMHCpan-3.0, PickPocket, Nebula, and SMM.
• Computational Efficiency: Measured to be more than 197 times faster than existing methods when predicting neoantigens from a large breast tumor peptide dataset.
• Large-scale Neoantigen Identification: Applied in a pipeline that identified 117,017 neoantigens across 467 cancer samples from the TCGA database.

Scientific Applications:

• Neoantigen discovery: Identification of tumor-specific neoantigens to inform cancer vaccine design and immunotherapy target selection.
• High-throughput cancer cohort screening: Large-scale prediction of peptide–HLA binding across TCGA cancer samples for cohort-wide neoantigen profiling.

Methodology:

Implements a PSSM-based scoring method; trained and validated by 10-fold cross-validation on a database of 87 HLA alleles and independently benchmarked on the Peptide Database of Cancer Immunity against NetMHC-4.0, NetMHCpan-3.0, PickPocket, Nebula, and SMM.