First, I have been very bad about posting news. It has been a productive half year in the Salsbury Group.
However, I will break the drought of news, by announcing our latest accepted manuscript in Frontiers in Physics, "MutSα’s Multi-Domain Allosteric Response to Three DNA Damage Types Revealed by Machine Learning."
This article focuses on how the DNA complex, MutSalpha, responds differently when bound to DNA crosslinked by cisplatin, i.e., 1-2 cross-linking, carboplatin, more 1-3 cross-linking, and with DNA with an FDU-substitution. We have published on these on the past, except FDU, but there were GPU-enabled simulations an order of magnitude longer than before, and we used machine learning techniques to examine hydrogen bonds and to do unbiased clustering to really examine the conformational effects of binding these different damaged DNAs with minimal user bias. The selection of interface hydrogen bonds was motivated by previous work. Overall, using these machine learning on these long-scale simulations really cleanly identified different conformational responses and the hydrogen bonds associated with them, even though the changes were often non-local.
I might write a more detailed summary, but I am quite happy with this article, and you can read it here.
PS provisional article, pre-proof, so some typos/minor corrections are being made.