New preprint: rapid prediction of AMR by free energy methods Philip Fowler, 15th January 202015th January 2020 0 shares The story behind this preprint goes back to the workshop on free energy methods run by BioExcel in Göttingen in May 2019. I gave a talk, based in part on the work I’d previously published showing how alchemical free energy methods are able to predict which mutations in S. aureus DHFR confer resistance to trimethoprim. Several of us were talking in the coffee break, I think about how the lambda simulations in the DHFR study are very short (250 ps), and I remember thinking “I wonder how short we could make them and still be able to qualitatively predict whether a mutation confers resistance or not?”. Turns out that answer is very, very short! But, you need to do a large number of repeats. The combined effect, however, is that, in principle, one can make a prediction (for DHFR at least) using about a tenth of the computational resource we originally used. Not only that, but since all the individual simulations are now only 50 ps long, one could theoretically run them all in parallel in less than ten minutes using GPUs. Overall, however, you would spend far longer creating the simulation input files (grompp in GROMACS), copying them onto the high performance computer, submitting them to the queue and retrieving and analysing the files! That is a problem for another day… Share this:Twitter Related antimicrobial resistance clinical microbiology computing GPUs molecular dynamics publication research
New publication: fast human read decontamination for SARS-CoV-2 16th May 202216th May 2022 ReadItAndKeep is a new human-read decontamination algorithm that works by mapping the reads in a… Share this:Twitter Read More
clinical microbiology New preprint: processing SARS-CoV-2 genetics in the cloud 31st January 202431st January 2024 In this preprint, we describe how in July 2022 for two weeks seven sites in… Share this:Twitter Read More
antimicrobial resistance AMyGDA now available from GitHub 27th January 202027th January 2020 AMyGDA is a python module that analyses photographs of 96-well plates and, by examining each… Share this:Twitter Read More