MSMBuilder: Statistical Models for Biomolecular Dynamics

Area of Science:

• Computational Biophysics
• Machine Learning
• Data Science

Background:

• High-dimensional time-series data analysis is crucial in biomolecular dynamics.
• Markov state models (MSMs) are increasingly favored for analyzing such data.
• Existing tools may lack comprehensive features or broad applicability.

Purpose of the Study:

• To introduce MSMBuilder, a software package for building statistical models of high-dimensional time-series data.
• To provide tools for analyzing atomistic simulations, focusing on protein folding and conformational changes.
• To offer a user-friendly interface and compatibility with the machine learning community.

Main Methods:

• Utilizes Markov state models (MSMs) for time-series analysis.
• Incorporates hidden Markov models and time-structure based independent component analysis.
• Features a command-line interface and a Python API, designed with scikit-learn compatibility.

Main Results:

• MSMBuilder facilitates the construction of MSMs for biomolecular dynamics.
• The package offers complementary algorithms for comprehensive time-series data understanding.
• It provides a flexible and accessible platform for researchers.

Conclusions:

• MSMBuilder is a versatile software package for statistical modeling of time-series data.
• It is particularly beneficial for analyzing biomolecular dynamics simulations.
• The package supports both molecular dynamics practitioners and broader time-series analysis applications.