Protein design: A perspective from simple tractable models

TL;DR

This paper reviews recent computational methods for protein design based on statistical-mechanical models, focusing on designability, model evaluation, and improving sequence optimization techniques to better mimic experimental results.

Contribution

It introduces a simple condition for protein designability and critically evaluates existing models and optimization approaches, suggesting improvements for experimental relevance.

Findings

Evaluation of degeneracy in protein folding

Analysis of strengths and weaknesses of stochastic optimization methods

Discussion on features of proteins that can be designed in or out

Abstract

We review the recent progress in computational approaches to protein design which builds on advances in statistical-mechanical protein folding theory. In particular, we evaluate the degeneracy of the protein code (i.e. how many sequences fold into a given conformation) and outline a simple condition for ''designability`` in a protein model. From this point of view we discuss several popular protein models that were used for sequence design by several authors. We evaluate the strengths and weaknesses of popular approaches based on stochastic optimization in sequence space and discuss possible ways to improve them to bring them closer to experiment. We also discuss how sequence design affects folding and point out to some features of proteins that can be deigned ''in'' or designed ''out''}