Life Sciences Seminar: ATOMDANCE
Life Sciences Seminar
ATOMDANCE: machine learning denoising and resonance analysis for functional and evolutionary comparisons of protein dynamics
Dr. Greg Babbitt and Dr. Maureen Ferran
Thomas H. Gosnell School of Life Sciences, RIT
Here, we introduce ATOMDANCE, a suite of statistical and kernel-based machine learning tools designed for denoising and comparing functional motion states of proteins captured in time-series from molecular dynamics simulations. We’ll spend the first half of the talk reviewing the software/methods and we’ll spend the second half as an open discussion of its potential application to problems in virology and immunology.
Abstract:
Comparative methods in molecular biology and molecular evolution rely exclusively upon the analysis of DNA sequence and protein structure, both static forms of information. However, it is widely accepted that protein function results from dynamic shifts in machine-like motions induced by molecular interactions, a type of data for which comparative methods of analysis are challenged by the large fraction of protein motion created by random thermal noise induced by the surrounding solvent. Here, we introduce ATOMDANCE, a suite of statistical and kernel-based machine learning tools designed for denoising and comparing functional motion states of proteins captured in time-series from molecular dynamics simulations. ATOMDANCE employs interpretable Gaussian kernel functions to compute site-wise maximum mean discrepancy (MMD) between learned features of motion (i.e. functional dynamics) representing two protein states (e.g. bound vs. unbound, wild-type vs. mutant). ATOMDANCE derives empirical p-values identifying functional similarity/difference in dynamics at each amino acid site on the protein. ATOMDANCE also employs MMD to contextually analyze potential random amino-acid replacements thus allowing for a site-wise test of neutral vs. non-neutral evolution in the divergence of dynamic function in protein homologs. Lastly, ATOMDANCE also employs mixed-model ANOVA combined with graph network community detection to identify functional shifts in protein regions that exhibit time-coordinated dynamics or resonance of motion across sites. ATOMDANCE offers a user-friendly interface and requires as input only single structure, topolopy and trajectory files for each of the two proteins being compared. A separate interface for generating molecular dynamics simulations via open-source tools is offered as well. We’ll spend the first half of the talk reviewing the software/methods and we’ll spend the second half as an open discussion of its potential application to problems in virology and immunology.
Intended Audience:
Beginners, undergraduates, graduates, experts. Those with interest in the topic.
To request an interpreter, please visit myaccess.rit.edu
Event Snapshot
When and Where
Who
Open to the Public
Interpreter Requested?
No