Pred-AHCP: Robust\nFeature Selection-Enabled Sequence-Specific\nPrediction of Anti-Hepatitis C Peptides via Machine Learning

Every\nyear, an estimated 1.5 million people worldwide contract\nHepatitis C, a significant contributor to liver problems. Although\nmany studies have explored machine learning’s potential to\npredict antiviral peptides, very few have addressed the problem of\npredicting peptides against specific viruses such as Hepatitis C.\nIn this study, we demonstrate the application and fine-tuning of machine\nlearning (ML) algorithms to predict peptides that are effective against\nHepatitis C virus (HCV). We developed a fine-tuned and explainable\nML model that harnesses the amino acid sequence of a peptide to predict\nits anti-hepatitis C potential. Specifically, features were computed\nbased on sequence and physicochemical properties. The feature selection\nwas performed using a combined strategy of mutual information and\nvariance inflation factor. This facilitated the removal of redundant\nand multicollinear features, enhancing the model’s generalizability\nin predicting anti-hepatitis C peptides (AHCPs). The model using the\nrandom forest algorithm produced the best performance with an accuracy\nof about 92%. The feature analysis highlights that the distributions\nof hydrophobicity, polarizability, coil-forming residues, frequency\nof glycine residues and the existence of dipeptide motifs VL, LV,\nand CC emerged as the key predictors for identifying AHCPs targeting\ndifferent components of HCV. The developed model can be accessed through\nthe Pred-AHCP web server, provided at http://tinyurl.com/web-Pred-AHCP. This resource facilitates the prediction and re-engineering of\nAHCPs for designing peptide-based therapeutics while also proposing\nan exploration of similar strategies for designing peptide inhibitors\neffective against other viruses. The developed ML model can also be\nused for validating peptide sequences generated using generative artificial\nintelligence methods for further optimization.