Accuracy benchmark for nucleic acid virtual site parameters

NAISS 2023/5-417


NAISS Medium Compute

Principal Investigator:

Erik Marklund


Uppsala universitet

Start Date:


End Date:


Primary Classification:

10402: Physical Chemistry

Secondary Classification:

10603: Biophysics

Tertiary Classification:

10203: Bioinformatics (Computational Biology) (applications to be 10610)




Virtual interaction sites (vsites) enable classical MD simulations of biomolecules to take 2.5 times longer time steps, meaning 2.5 times more useful data per core hour on supercomputing centers. Vsites have been available for proteins for over a decade, but to date no such parameters exist for nucleic acids, which means that simulations of DNA/RNA systems, or mixed protein-DNA/RNA systems are quite limited from a performance perspective. To enable widespread deployment of our vsite parameters we need to test them thoroughly and make sure vsite-containing simulations reproduce the properties of non-vsite counterparts. After initial rounds of testing (SNIC 2022/22-854) we have made additional tests on Dardel that will underpin the next round of tests involving free energy calculations. Here, we intend to run the latter tests to see if vsites shifts the balance between different states or the energy barriers between them