Computational models not yet powerful enough for drug discovery
Scientists want to use computer models to help reduce the cost and time associated with drug discovery, to develop new antibiotics to tackle the growing crisis of antimicrobial resistance. But a new study shows that using the latest tools together is little better than guesswork right now.
This is a barrier to drug development – at least to the extent that computer models currently exist – according to a study by a new study Posted in Biology of molecular systems.
Researchers at the Massachusetts Institute of Technology (MIT) investigated whether existing computer programs could accurately predict interactions between antibacterial compounds and bacterial protein structures generated by Google’s new tool called AlphaFold, an artificial intelligence program which generates 3D protein structures from their amino acid sequence.
Alpha folding fascinates the scientific world.
But the MIT team found that predictions from existing models, called molecular docking simulations, performed little better than chance.
“Breakthroughs like AlphaFold expand the possibilities of silicone (i.e. computationally) drug discovery efforts, but these developments need to be coupled with further advances in other aspects of modeling that are part of drug discovery efforts,” says the lead author James Collins, professor of medical engineering and science at MIT’s Institute for Medical. Engineering and Science (IMES) and Department of Biological Engineering.
“Our study speaks to both the current capabilities and the current limitations of computational platforms for drug discovery.”
The hope is that scientists could use modeling to perform large-scale screening for new compounds that affect previously untargeted bacterial proteins. The end result being the development of new antibiotics that work in unprecedented ways.
Read more: Google’s Protein-Folding AI AlphaFold Has Almost All Broke Them
The team studied the interactions of 296 essential proteins of Escherichia coli with 218 antibacterial compounds, using molecular docking simulations that predict the bond strength of two molecules based on their shapes and physical properties.
Previously, these simulations have been used successfully to screen large numbers of compounds against a single protein target to identify compounds that bind best. But here the predictions became much less accurate when trying to screen many compounds against many potential protein targets.
In fact, the model produced similar false positive rates to true positive rates when simulating interactions between existing drugs and their targets.
“Using these standard molecular docking simulations, we got an auROC value of around 0.5, which basically means you’re doing no better than guessing at random,” says Collins.
But that was no fault of AlphaFold, as similar results occurred when they used the same modeling approach with experimentally determined protein structures in the lab.
“AlphaFold appears to do about as well as experimentally determined structures, but we need to do a better job with molecular docking models if we are to use AlphaFold effectively and widely in drug discovery,” adds Collins.
One explanation for this poor performance is that the protein structures introduced into the model are static, but in real biological systems proteins are flexible and often change configuration.
The researchers were able to improve the performance of the molecular docking simulations by running them through four additional machine learning models trained on data that describes how proteins and other molecules interact with each other.
Read more: Tuberculosis drug discovery paves the way
“Machine learning models not only learn the shapes, but also the chemical and physical properties of known interactions, then use that information to reassess docking predictions,” says co-lead author Felix Wong, applied physicist and fellow postdoc at Collins. ‘ MIT lab.
“We found that if you were to filter interactions using these additional patterns, you can get a higher ratio of true positives to false positives.”
“We are optimistic that with improvements in modeling approaches and expansion of computing power, these techniques will become increasingly important in drug discovery,” Collins concludes. “However, we still have a long way to go to reach the full potential of silicone drug discovery.”