Perturbation-Theory Machine Learning for Multi-Target Drug Discovery in Modern Anticancer Research

Cancers constitute a group of biological complex diseases, which are associated with great prevalence and mortality. These medical conditions are very difficult to tackle due to their multi-factorial nature, which includes their ability to evade the immune system and become resistant to current anticancer agents. There is a pressing need to search for novel anticancer agents with multi-target modes of action and/or multi-cell inhibition versatility, which can translate into more efficacious and safer chemotherapeutic treatments. Computational methods are of paramount importance to accelerate multi-target drug discovery in cancer research but most of them have several disadvantages such as the use of limited structural information through homogeneous datasets of chemicals, the prediction of activity against a single target, and/or lack of interpretability. This mini-review discusses the emergence, development, and application of perturbation-theory machine learning (PTML) as a cutting-edge approach capable of overcoming the aforementioned limitations in the context of multi-target small molecule anticancer discovery. Here, we analyze the most promising investigations on PTML modeling spanning over a decade to enable the discovery of versatile anticancer agents. We highlight the potential of the PTML approach for the modeling of multi-target anticancer activity while envisaging future applications of PTML modeling.