New paper: predicting rifampicin resistance via free energy simulation

This work was carried out by Xibei Zhang, who is doing her PhD with Peter Coveney at UCL. It builds on earlier work I did using alchemical free energy methods to calculate whether individual mutations in the protein target of an antibiotic reduce how well the drug can bind, thereby conferring resistance. The method was first shown to work on DHFR in S. aureus and this was followed up by another paper looking to see just how fast we could do this (as the method is extremely computationally intensive so even with access to high performance computing takes a while to return a result). DHFR is a fairly small protein so proved the method but wasn’t a good test. So myself and Alice then tried the RNA polymerase and DNA gyrase in M. tuberculosis with mixed results, mainly due to the longer convergence time required due to the sheer size of the protein complexes.

Xibei has picked up the baton and applied their in-house free energy method, TIES_PM, to the RNA polymerase in this paper increasing the number of mutations considered from six to 61, in part through using the Polaris supercomputer at Argonne National Labs in the USA; it was the 12th fastest supercomputer on the planet back in 2021.

Hence whilst these methods remain computational expensive, they are getting more tractable, even for large protein complexes like the RNA polymerase, as computer processing power continues to increase, especially as molecular dynamics codes can run on GPUs which continue to get bigger and faster due to the current focus on artificial intelligence. We have proposed a few research projects and, should one be funded, hope to return to this idea in the near future.