(27ac) Meta-Analysis of Proteolytic Cleavage Specificity Using Machine Learning

Proteases are enzymes that cleave the peptide bonds of proteins which play crucial roles in several physiological and pathological processes, including digestion of dietary proteins, apoptosis, cell differentiation, inflammation, and nociception. Proteases are highly diverse enzymes that can differ in their substrate specificity, catalytic mechanism, and regulation. Alterations in proteolytic systems are underlying factors contributing to multiple pathological conditions, including cancer, neurodegenerative disorders, and inflammatory and cardiovascular diseases. [1] Understanding protease biology is critical for the development of effective protease-targeted therapeutics, however, identifying protease cleavage sites is a challenging task. Recently, computational have been developed to predict proteolytic cleavage sites using machine learning and artificial intelligence models. [3, 4] We developed a machine learning-based algorithm that uses support vector machines (SVMs) to predict peptide substrates of cysteine proteases, specifically cathepsins K, S, L, and V. To train and test our models, we used a publicly available online database, MEROPS, and cross-validation. Moreover, we conducted a meta-analysis using SVM-based feature selection in order to compare the substrate specificity of cathepsins K, S, L, and V. This analysis has potential to contribute to the advancement of protease-targeted therapeutics by providing a guide for design novel cathepsin-specific substrates.

1. Lopez-Otin, C. and J. S. Bond (2008). "Proteases: multifunctional enzymes in life and disease." J Biol Chem283(45): 30433-30437.
2. Bleuez, C., et al. (2022). "Exploiting protease activation for therapy." Drug Discov Today 27(6): 1743-1754.
3. Bhasin, M. and G. P. Raghava (2005). "Pcleavage: an SVM based method for prediction of constitutive proteasome and immunoproteasome cleavage sites in antigenic sequences." Nucleic Acids Res 33(Web Server issue): W202-207.
4. Li, F., et al. (2020). "DeepCleave: a deep learning predictor for caspase and matrix metalloprotease substrates and cleavage sites." Bioinformatics 36(4): 1057-1065.