Comprehensive Analysis of Computational Models for Prediction of Anticancer Peptides Using Machine Learning and Deep Learning

Abstract

Anti-cancer peptides (ACPs) represent promising candidates for cancer therapy because they can target cancer cells selectively while leaving healthy cells unaffected. ACPs offer a multifaceted approach to cancer treatment by combining targeted cytotoxicity, immune system activation, and the potential to overcome drug resistance. Their development is aided by computational tools that expedite the discovery of promising candidates. As a result, they have received significant attention and broadly studied by many researchers. Currently, numerous peptide-based drugs are undergoing evaluation in preclinical and clinical trials. Accurately identifying ACPs has become a major focus of research, leading to the construction of diverse methods for their detection in silico. These methods implemented different training/testing datasets, classifiers, feature engineering, and feature selection techniques. Thus, it is indispensable to highlight the strengths and weaknesses of current methods and provide insights to improve novel computational tools for identification of ACPs. To address this, we conducted a comprehensive investigation of 26 available existing methods for ACPs, examining their feature engineering methods, classification learning algorithms, performance validation parameters, and availability of web servers. Subsequently, we performed a thorough performance assessment to examine the robustness of these studies using different benchmark datasets. Based on our findings, we offer potential strategies for enhancing model performance and effectiveness.