Conformational Ensembles from Experimental Data
and Computer Simulations
Poster Abstracts
93
58-POS
Board 18
Matching Pursuit Genetic Algorithm for Structure Characterization of Large Intrinsically
Disordered Proteins
Wei Liu
, Daiwen Yang.
National University of Singapore, Singapore, Singapore.
Structure characterization of intrinsically disordered proteins (IDPs) remains a key obstacle in
understanding their functional mechanism. Due to the highly dynamic feature of IDPs, structure
ensembles instead of static unique structures are often derived from experimental data.
Determination of a structure ensemble usually uses a combinatorial optimization strategy, which
selects an optimal ensemble to fit the data from a structure pool without prior experimental
information. The search space of the combinatorial optimization problem could be extremely
huge, as it’s an exponential function of the ensemble size and a power function of the pool size.
In such a case, conventional algorithms become less efficient to find a good solution within
appropriate computational time. Here we present a matching pursuit genetic algorithm (MPGA),
which uses matching pursuit (MP) for search space reduction and genetic algorithm (GA) for
optimization. A sub-pool is selected from the original pool based on diverse criteria, and a
structure ensemble is selected from the sub-pool by GA. Like MP, the sub-pool is sequentially
adjusted according to the differences between the experimental and back-calculated data, and
then utilized in next round GA. This process is iterated until the outcome converges. We
demonstrate that the MPGA method outperforms other state-of-art algorithms in structure
ensemble selection from a large pool (>1.3 million) of p130CasSD, an IDP with 306 amino
acids.