UC researchers use Artificial Intelligence to boost the development of new drugs for cancer

English version: Diana Taborda

A research team of the University of Coimbra (UC) has created an innovative computational model that may enable faster and less expensive development of new drugs to be used in cancer treatment, focused on the biological context of the disease. The results of the study were published in the journal "Briefings In Bioinformatics".

Bearing in mind that drug design is a highly complex, lengthy and expensive process, this work – a collaboration between the Faculty of Sciences and Technology (FCTUC) and the Faculty of Pharmacy (FFUC) – aimed to shorten the initial stages of drug design, using Artificial Intelligence (AI), by applying computational methods that can generate pharmacologically interesting compounds in a faster and more automated way.

In order to develop the new model, the team of the Department of Informatics Engineering of FCTUC used Machine Learning techniques, namely Deep Learning – a method that uses Artificial Neural Networks (ANN). These structures allow the design of intelligent models "by mimicking the learning capacity of biological models. They are thus able to identify patterns embedded in data sets and, from there, it is possible to obtain models that generate new molecular structures and that predict biological properties of interest", explains Tiago Oliveira Pereira, first author of the study, and PhD student supervised by Professors Maryam Abbasi and Joel P. Arrais (Principal Investigator).

According to the authors, "the method to implement targeted molecular generation that employs biological information, namely, disease-associated gene expression data, to conduct the process of identifying interesting hits. The model developed provides a novel and reliable method for generating new promising compounds focused on the biological context of the disease, without causing undesirable side effects.”

With the collaboration of the laboratory of FFUC professor Jorge Salvador, it was possible to apply the model in a case study for the generation of compounds capable of inhibiting the USP7 protein (Ubiquitin specific protease 7). This protein, stresses Tiago Oliveira Pereira, plays a key role "in the progression of various types of cancer and is currently seen as an important receptor for the development of drugs".

The results obtained in the experiments conducted are highly promising, with the model showing a high capacity to generate potential USP7 inhibitor molecules: "More than 90% of the molecules contained physical, chemical and biological properties essential for the interaction with the receptor to occur. In addition, we verified that some compounds generated by the model present similarities with anti-cancer drugs at the level of their active groups, which validates the implemented approach", says Tiago Oliveira Pereira.

The researcher adds that the next steps of the research will focus on the improvement of the implemented structure and the "setting of validation methods to filter the molecules obtained and, depending on the results, move on to the synthesis of the best compounds".

This study was co-funded by the Portuguese Foundation for Science and Technology (FCT), by the PIDDAC programme and by European funds through the project “D4-Deep Drug Discovery and Deployment”. The scientific article is available here: https://doi.org/10.1093/bib/bbac270.