Machine Learning approaches to improve prediction of target-drug interactions

BALATTI, GALO E; BARLETTA, PATRICIO G; PEREZ, ANDRES, D; GIUDICESSI, SILVANA L.; MARTINEZ CERON, MARIA C

Design of New Drugs Using Machine Learning

Wiley Scrivener Publishing

Lugar: Beverly; Año: 2022; p. 21 - 96

From the initial steps of early drug discovery, the traditional techniques, like docking, QSAR, or molecular dynamics, have been used for decades identifying targets, ranking molecule candidates and optimizing the lead compounds chemically to decrease toxicity and improve drug absorption, distribution, metabolism, and excretion (ADME) properties. Nowadays, computational tools are increasingly used not only in the drug discovery process but also in drug development. Information technologies like artificial intelligence (AI) and machine learning (ML) participate in practically every step in the pharma value chain, improving and accelerating the overall drug development and design. Besides the rise of new methodologies, the improvement of relative old computational techniques, like docking, QSAR, or cavity search with new ML-based algorithms like random forest, support vector machines, or neural networks are being developed and promises to reduce time and operational costs mainly in the drug design process. In this chapter, we review some of these approaches, briefly introducing the most used ML techniques in study drug-target interactions. Also, a special section about peptide-based drugs and its advantages over small organic molecules is discussed.