Transcript No Slide Title

Using Motion Planning to
Study Protein Folding Pathways
Susan Lin, Guang Song and Nancy M. Amato
Department of Computer Science
Texas A&M University
http://www.cs.tamu.edu/faculty/amato/
Protein Folding
Protein folding is a “grand challenge” problem in biology the deciphering of the second half of the genetic code, of
pressing practical significance
Problem 1: given a protein’s amino acid sequence, predict its
3D structure, which is related to its function
Problem 2: “… use the protein’s known 3D structure to predict
the kinetics and mechanism of folding” [Munoz & Eaton, PNAS’99]
–Finding protein folding pathways - OUR FOCUS - will assist in
understanding folding and function, and eventually may lead to
prediction.
PRMs for Protein Folding
Node Generation [Singh,Latombe,Brutleg 99]
• randomly generate conformations (determine
all atoms’ coordinates)
• compute potential energy E of conformation
and retain node with probability P(E):
Querying the Roadmap
• Add start (extended conformation)
and goal (native fold) to the roadmap
•Extract smallest weight path
(energetically most feasible)
Roadmap Connection
• find k closest nodes to each roadmap node
• calculate weight of straightline path
between node pairs - weight reflects the
probability of moving between nodes (the
smaller the weight the lower the energy)
Validating Folding Pathways
Protein GB1 (56 amino acids)
—
1 alpha helix & 4 beta-strands
Hydrogen Exchange Results
first helix, and beta-4 & beta-3
Our Paths
60%: helix, beta 3-4, beta 1-2, beta 1-4
40%: helix, beta 1-2, beta 3-4, beta 1-4
Protein A:
Potential Energy vs. RMSD for roadmap nodes
hypothetical roadmap
for Protein A
start:
amino acid string
funnel
goal:
native fold
‘funnel’ for RMSD< 10 A,
suggests packing of
secondary structure
(similar potentials)