DROIDSv3.0/maxDemonv1.0 - software for machine learning applied to comparative protein dynamics

The application of statistical methods to comparatively framed questions about protein dynamics can potentially enable investigations of biomolecular function beyond the current sequence and structure based methods in comparative genomics. However, addressing this problem requires proper statistical inferences obtained from large ensembles of individual molecular dynamic (MD) simulations that represent the comparative functional states of a given protein under investigation. Meaningful interpretation of such large and temporally rich forms of data poses serious challenges to users of MD. Here, we announce the release of DROIDS v3.0, an Amber18/Chimera-based software package for comparative protein dynamics, incorporating many new features including maxDemon v1.0, a multi-method machine learning application that trains on large ensembles of MD comparisons generated by DROIDS and deploys learned classifications of dynamic states to newly generated protein simulations. Up to seven different machine learners can be deployed on the dynamics of each amino acid, and local canonical correlations in learning patterns generated from self-similar MD runs are used to identify regions of functionally conserved protein dynamics. The subsequent impacts of genetic and drug class variants on conserved dynamics can also be analyzed by deploying the trained classifiers on new MD runs. Still and moving images of comparative dynamics allow users to see both when and where a protein dynamic simulation displays a specific functional state defined by the functional protein comparison. Our software allows quantitative visualization of complex changes in functional dynamics caused by temperature changes, chemical mutations, or binding interactions with nucleic acids or a variety of small molecules. Here, we demonstrate the utility of DROIDS 3.0 + maxDemon in four case studies of the impact of genetic and drug class variation on functionally conserved protein dynamics in ubiquitin, TATA binding protein and Hsp90.