Modeling large-scale protein motions, such as those involved in folding and
binding interactions, is crucial to better understanding not only how
proteins move and interact with other molecules but also how proteins
misfold, thus causing many devastating diseases. Robotic motion planning
algorithms, such as Rapidly Exploring Random Trees (RRTs),
have been successful in simulating protein folding pathways.
Here, we propose a new multi-directional Rapidly Exploring Random
Graph (mRRG) specifically tailored for proteins.
Unlike traditional RRGs which only expand a parent conformation in a single direction,
our strategy expands the parent conformation in multiple directions
to generate new samples.
Resulting samples are connected to the parent conformation and its nearest neighbors.
By leveraging multiple directions, mRRG can model the protein
motion landscape with reduced computational time compared to several
other robotics-based methods
for small to moderate-sized proteins.
Our results on several proteins agree with experimental hydrogen out-exchange, pulse-labeling,
and Phi-value analysis.
We also show that mRRG covers the conformation space better
as compared to the other computation methods.