Unbiased, scalable sampling of protein loop conformations from probabilistic priors

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Dynamics

Background:

• Protein loops are crucial for function but challenging to model computationally.
• Existing methods like discrete search, optimization, and molecular dynamics have limitations in speed and accuracy for loop conformation generation.
• Incorporating diverse structural preferences into loop modeling remains a significant hurdle.

Purpose of the Study:

• To introduce a novel Markov chain Monte Carlo algorithm, Sub-Loop Inverse Kinematics Monte Carlo (SLIKMC).
• To enable the generation of protein loop conformation ensembles based on heterogeneous structural preferences.
• To address the computational challenges in modeling protein loop dynamics and conformations.

Main Methods:

• Developed SLIKMC, a Markov chain Monte Carlo algorithm for generating closed loop conformations.
• Utilized a probabilistic graphical model (PGM) to specify structural preferences within a unified framework.
• Implemented a novel sub-loop sampling method to ensure statistically unbiased sampling under loop-closure constraints.

Main Results:

• SLIKMC significantly outperforms standard Monte Carlo and discrete search methods in speed for large loops (>10 residues).
• The use of a sparse PGM enhances scalability by exploiting the locality of atomic and residue interactions.
• The sub-loop sampling technique generates statistically unbiased probability densities restricted by loop-closure constraints.

Conclusions:

• SLIKMC successfully generates protein conformation ensembles consistent with specified structural preferences.
• The method achieves rapid sampling of protein conformations (100+ residues) on standard hardware.
• SLIKMC shows potential for applications in loop ensemble generation and completing missing protein structures, as demonstrated in ion-binding proteins.