A team of researchers at IRB Barcelona’s Structural Bioinformatics and Network Biology have developed a tool that predicts the biological activity of chemical compounds, which is the key to evaluating their therapeutic potential. The researchers used artificial neural networks to obtain experimental data for one million compounds and developed many tools to evaluate each type of molecule. The Structural Bioinformatics and Network Biology team, led by Dr. Patrick Aloy, a researcher at ICREA, used deep machine-learning computational models to complete the collection of biological activity information about 1 million molecules and introduced a tool to predict any molecule’s biological activity, even when experimental data is unavailable.
This new method is based on the Chemical Checker, which is by far the largest bioactivity profile database for fake medicines developed by the same laboratory and released in 2020. The database collects information from 25 bioactive areas for each molecule. These areas are related to the chemical structure of the molecule, the target that it interacts with, and the alterations it causes at the clinical or cellular level. However, for most of the compounds, this detailed information about the mechanism of action is incomplete. This means that for a given compound, there might be one or two biologically active areas of information available, but not all 25. With this new discovery in development, researchers would compare all available experimental information with deep machine learning techniques to complete all activity profiles for all compounds, from chemistry to clinical level.
The new tool also allows us to predict the biological activity space of new molecules, which is essential for the drug discovery process because we can select the most suitable candidates and discard those that do not work for some reason or other reasons