Neopepsee: Accurate Genome-level Prediction of Neoantigens by Harnessing Sequence and Amino Acid Immunogenicity Information

In cancer genomics, next generation sequencing data are usually used to detect somatic driver mutations for identifying the cause of tumorigenesis. However, non-driver somatic mutations, or passenger mutations, can also play an important role in cancer cell survival and treatment by generating aberrant short peptide sequences as known as "neoantigens". Tumor-specific mutations form novel immunogenic peptides called neoantigens, which could be important to the recent promising outcomes of cancer immunotherapy, including immune checkpoint blockade. Many studies have tried to identify various sequence characteristics for prediction of immunogenicity; however, practical applications rely on a single predicted value (MHC-I binding affinity) with an arbitrary cut-off. Here, we developed Neopepsee, a method that applies a machine learning to predict personal neoantigen with next generation sequencing data. Neopepsee not only automates the entire computational procedure for immunogenicity prediction from raw data but also improves accuracy by harnessing 10 different features for classification, including conventional MHC-I and T-cell receptor binding affinity and amino acid characteristics (e.g., hydrophobicity, polarity and charge). Additionally, we found that protein sequence similarity to known pathogenic epitopes is a novel strong feature for classification. Tests with validated epitope datasets and independently proven neoantigens confirmed the remarkable improvement in accuracy. Application of Neopepsee to 224 public stomach adenocarcinoma data predicted neoantigens, whose burden is strongly correlated with patient prognosis. By providing a convenient platform with better accuracy, Neopepsee will be of many uses in cancer immunotherapy research, such as in developing predictive biomarkers and in designing personalized cancer vaccines.