Inferring protein 3D structure from deep mutation scans.

Area of Science:

• Structural Biology
• Computational Biology
• Genetics

Background:

• Natural evolutionary sequence covariation has been successfully used to predict protein and RNA structures.
• The potential of synthetically generated sequence variations, such as those from laboratory mutational scans, for structure determination remained unexplored.

Purpose of the Study:

• To investigate whether laboratory-induced, synthetic sequence variations can be utilized for three-dimensional (3D) structure determination of macromolecules.
• To establish a novel experimental method for 3D structure prediction leveraging high-throughput mutational data.

Main Methods:

• Analysis of five large-scale experimental mutational scans.
• Identification of residue pairs exhibiting the largest positive epistasis within these scans.
• Computation of ab initio protein folds using identified epistatic pairs as contact restraints.

Main Results:

• The strongest epistatic pairings from genetic screens of proteins and a ribozyme accurately reveal 3D contacts within and between macromolecules.
• Ab initio structure prediction for a GB1 domain (1.8 Å accuracy) and a WW domain (2.1 Å accuracy) was achieved using experimental epistatic pairs.
• Strategies were proposed to reduce the number of mutants required for accurate contact prediction.

Conclusions:

• Laboratory-generated sequence variation, specifically positive epistasis, is a viable and powerful tool for determining macromolecular 3D structures.
• This genomics-inspired approach offers an efficient alternative to traditional methods for predicting protein and RNA folds.
• The findings suggest a paradigm shift towards utilizing synthetic genetic data for structural biology.