High throughout prediction of toxin-antitoxin protein interfaces with AlphaFold2
Abstract
The Atkinson lab has recently been awarded a KAW project grant to work on high-throughput prediction along with biochemical and structural characterisation of novel toxin-antitoxin systems of bacteria. Toxin-antitoxin systems are two-gene operons that encode interacting proteins. These proteins are usually small often with completely novel folds. We plan to use AlphaFold2 to predict the domain structures of these tens of thousands of proteins, along with their protein-protein interaction interfaces. This project is a collaboration with Arne Elofsson, who has already successfully used the Berzelius cluster for prediction of human protein complexes.
We expect the results to be of the highest international impact as they will help us answer an important question of how toxin-antitoxin systems evolve to switch domain partners so readily. Our previous work on toxin-antitoxin systems has been published in high-impact journals (PNAS 2020 and 2021 (in press) and Mol Cell 2021). Our toxin-antitoxin predictions are generated using our tool FlaGs, which was published in Bioinformatics in 2020.
The longer term impact of these results will be in the field of phage therapy to treat antibiotic resistant infections. We know that toxin-antitoxin systems protect bacteria agains phages, so we want to design phages that carry antitoxins or anti-defence systems that can overcome these bacterial defences.
