(366e) Developing Data-Driven Models for Computational Drug Discovery to Enhance Therapeutic Specificity and Enable Precision Medicine

Computational drug discovery, Prediction and design of therapeutic peptides, Computational approaches in Precision medicine, Bioinformatics, Cheminformatics

Computational prediction, optimization, and design of therapeutics play a crucial role in reducing the cost, time, energy, and labor associated with experimental work and manufacturing. Recently, novel membrane-active peptides (MAPs) have attracted significant attention as alternatives to antibiotics and small molecule-based drugs, particularly due to the issue of drug resistance. In this study, we conducted an in-depth machine-learning prediction and analysis on a diverse array of therapeutic peptides within the broad class of MAPs, focusing on their capacity to either traverse cellular membranes (cell-penetrating peptides, CPPs) or disrupt these membranes (membranolytic peptides, MDPs). To accomplish this, we compiled datasets from the literature that encompass cell-penetrating, anticancer, antifungal, antiparasitic, antibacterial, antiviral, and mammalian-targeting peptides. Furthermore, as hemolytic activity and understanding and predicting the solubility of these peptides have been a primary concern for scientists, we also predicted the hemolytic activity and solubility of the therapeutic peptides. Leveraging known periodicities in peptide properties that correlate with their structure and function, we employed Fourier transforms to generate predictive descriptors by measuring the amplitude of amino acid property oscillations. We applied an in-house feature selection procedure, based on non-linear support vector machines (SVMs), to derive structure-function fingerprints for each class of MAPs. A comparison of our approach, based on the Fourier transform, with the state-of-the-art shows that our approach leads to models with significantly fewer features with at least comparable performance. Finally, we used the derived structure-function fingerprints to cluster the classes of MAPs, which provides insight into the design of MAPs with improved specificity. Additionally, we employed our promising SVM classifier and the same feature selection framework to predict individual biomarker responses to different cancer treatment components. This innovative study holds the potential to expedite and revolutionize the design and development of novel membrane-active peptides with high specificity, as well as precision oncology, offering promising avenues for drug discovery and clinical trials.

Keywords: Computational drug discovery, Support vector machine models, Feature selection.