Scientists at MIT Devise New Ways for Faster Drug Discovery through Machine Learning

·2-min read

Scientists at the Massachusetts Institute of Technology (MIT) have developed a new technique that quickly calculates the binding affinities between drug candidates and their targets, which would most likely quicken the pace of drug discovery and protein engineering.

Published in the Journal of Physical Chemistry Letters on March 16, 2021, the research shows how the new technique called DeepBAR, yields precise calculations in a fraction of the time, compared to previous state-of-the-art methods. Bin Zhang — who is an associate member of the Broad Institute of MIT and Harvard, a Pfizer-Laubach Career Development Professor in Chemistry at MIT, and the co-author of the research paper — said in a statement to MIT News that their method is orders of magnitude is faster than before, which means that they can have drug discovery that is both efficient and reliable.

The probability of attraction between a drug molecule and a target protein is gauged by a quantity called the binding free energy. The smaller the number, the stickier the bond. Zhang explains that a lower binding free energy means that the drug can compete against other molecules in a better way, which in turn means that it can more effectively disrupt the protein's normal function. Pointing out the hurdles faced in making effective medication drugs, Zhang said that calculating the binding free energy of a drug candidate provides an indicator of a drug's potential effectiveness but it is a difficult quantity to find out.

DeepBAR computes the binding free energy exactly, but it needs just a fraction of the calculations demanded by previous methods. This new technique blends traditional chemistry calculations with recent advances in machine learning.

BAR in DeepBAR stands for the "Bennett acceptance ratio," which is a decades-old algorithm used in precise calculations of binding free energy. This new technique has done away with the in-between states by deploying the Bennett acceptance ratio in machine-learning frameworks called deep generative models. DeepBAR calculated binding free energy nearly 50 times faster than previous methods in tests using small protein-like molecules, mentions the MIT News report. Zhang said that this kind of efficiency means that scientists can start to think about using this to do drug screening, particularly in the context of a pandemic situation like coronavirus.