Sampling bottlenecks in de novo protein structure prediction

Area of Science:

• Computational Biology
• Biophysics
• Structural Bioinformatics

Background:

• De novo protein structure prediction is hindered by conformational sampling difficulties.
• Accurate prediction for large proteins remains challenging with current methods like Rosetta.
• The computational resources needed for successful prediction are largely unknown.

Purpose of the Study:

• To develop a method for estimating the computational power required for accurate protein structure prediction.
• To reformulate protein conformational search as a discrete combinatorial sampling problem.
• To identify key features limiting conformational sampling in proteins.

Main Methods:

• Reformulating conformational search as a combinatorial sampling problem in a discrete feature space.
• Analyzing the impact of
• linchpin
• features on sampling efficiency.
• Investigating the role of backbone torsion angles as critical sampling constraints.

Main Results:

• Conformational sampling is often limited by rare, critical "linchpin" features.
• Constraining these linchpin features dramatically enhances sampling of the native protein state.
• These features are often located in strained regions important for protein function.
• Linchpin features in some proteins correspond to experimentally observed late-folding regions.

Conclusions:

• A new approach quantifies computational needs for protein structure prediction.
• Identifying and constraining linchpin features is key to efficient conformational sampling.
• This work suggests a link between in silico folding and experimental protein folding pathways.