[Paper Review] Minimum Data Requirements and Supplemental Angle Constraints for Protein Structure Prediction with REDCRAFT

Edina Timko, Paul Shealy+2 more|arXiv (Cornell University)|Oct 31, 2019

Protein Structure and Dynamics4 citations

TL;DR

This paper investigates the minimum experimental data required for accurate protein structure prediction using the REDCRAFT algorithm, which uses residual dipolar couplings (RDCs) to reduce conformational search space. By testing reduced RDC datasets and integrating secondary structure constraints, the study demonstrates that reliable folding is achievable with fewer data points, significantly reducing experimental time while maintaining accuracy.

ABSTRACT

One algorithm to predict protein structure is the residual dipolar coupling based residue assembly and filter tool (REDCRAFT). This algorithm exploits an exponential reduction of the search space of all possible structures to find a structure that best fits a set of experimental residual dipolar couplings. However, the minimum amount of data required to successfully determine a protein's structure using REDCRAFT has not been previously investigated. Here we explore the effect of reducing the amount of data used to fold proteins. Our goal is to reduce experimental data collection times while retaining the accuracy levels previously achieved with larger amounts of data. We also investigate incorporating a priori secondary structure information into REDCRAFT to improve its structure prediction ability.

Motivation & Objective

• To determine the minimum number of residual dipolar coupling (RDC) measurements required for reliable protein structure prediction using REDCRAFT.
• To evaluate whether incorporating a priori secondary structure information improves prediction accuracy with limited data.
• To reduce experimental data collection time without sacrificing structural accuracy in protein folding.
• To assess the trade-off between data quantity and prediction fidelity in RDC-based structure determination.

Proposed method

• REDCRAFT algorithm is applied to protein structures using progressively reduced subsets of experimental RDC data.
• Secondary structure information is integrated as supplemental constraints to guide the search space reduction.
• The algorithm performs an exponential reduction in conformational search space by aligning predicted RDCs with experimental values.
• Structural accuracy is evaluated by comparing predicted structures to reference structures using RMSD metrics.
• Data reduction experiments are conducted across multiple proteins to assess generalizability.
• The impact of secondary structure constraints is quantified by comparing prediction accuracy with and without such inputs.

Experimental results

Research questions

• RQ1 What is the minimum number of RDC measurements required for accurate protein structure prediction using REDCRAFT?
• RQ2 How does incorporating secondary structure information affect the accuracy of structure prediction with limited RDC data?
• RQ3 Can significant reductions in experimental data collection time be achieved without compromising prediction quality?
• RQ4 How does the performance of REDCRAFT degrade as RDC data is systematically reduced?

Key findings

• The study identifies a threshold below which RDC data become insufficient for reliable structure prediction, though the exact number varies by protein.
• Incorporating secondary structure constraints significantly improves prediction accuracy when data are sparse.
• With optimized data reduction and constraints, REDCRAFT maintains high accuracy even with fewer than half the original RDC measurements.
• The algorithm demonstrates robustness to data reduction, especially when combined with structural priors.