Random Field Model Reveals Structure of the Protein Recombinational Landscape

Abstract

We are interested in how intragenic recombination contributes to the evolution of proteins and how this mechanism complements and enhances the diversity generated by random mutation. Experiments have revealed that proteins are highly tolerant to recombination with homologous sequences (mutation by recombination is conservative); more surprisingly, they have also shown that homologous sequence fragments make largely additive contributions to biophysical properties such as stability. Here, we develop a random field model to describe the statistical features of the subset of protein space accessible by recombination, which we refer to as the recombinational landscape. This model shows quantitative agreement with experimental results compiled from eight libraries of proteins that were generated by recombining gene fragments from homologous proteins. The model reveals a recombinational landscape that is highly enriched in functional sequences, with properties dominated by a large-scale additive structure. It also quantifies the relative contributions of parent sequence identity, crossover locations, and protein fold to the tolerance of proteins to recombination. Intragenic recombination explores a unique subset of sequence space that promotes rapid molecular diversification and functional adaptation.

Computational modelling of enzymes: predicting and understanding the selectivity of an epoxide hydrolase

“…In contrast, generative methods attempt to produce a full model representing the distribution of features for each class and compares how the distributions differ between them [54]. Discriminative methods are often chosen to predict the fitness of a protein from sequence data [41,[55][56][57][58][59][60][61][62][63][64]. Linear regression methods have typically been used [41,55,56,[58][59][60]62], however kernel methods [65] such as support vector machines (SVM) and Gaussian processes have also been applied to improve the thermostability, catalytic activity and ligand binding affinity of enzymes [61,63] and to classify the structural viability of proteins [57].…”

“…Studies that have used kernel methods have represented pairwise interactions between amino acid residues with residue-residue contact maps derived from resolved protein structures. Accounting for residue interactions helps capture potential epistatic effects and increases the predictive performance of a model [40,59,61,63]. Epistatic effects are prevalent between residues within the active site region and play a significant role in influencing enzyme selectivity [39].…”

“…Non-additive fitness contributions will complicate the exploration of the "fitness landscape" by making its topology more rugged [125,126] (Figure 1.2). It has not been established whether modelling sequence-activity relationships under the assumption of additivity [56,80,55,59] is sufficient to accurately predict beneficial mutations and their contributions to enzyme selectivity in the presence of strong epistatic effects, or whether non-additive methods [55,108,57,61] are required. In this study linear models and counterparts representing both pairwise and higher-order (three or more) residue interactions are constructed and evaluated on enzymes whose enantioselectivities have been experimentally characterised.…”

“…Simple models using linear regression have also been successfully applied to predict protein functional status, thermostability and biological activity [56,59,60,58,128,129]. Linear models based on characterised sequences generated with SCHEMA-guided recombination [56,59,60,58] have demonstrated good predictive ability with minimal sampling of the protein fitness landscape accessible by recombination, also referred to as the protein recombinational landscape [59]. The SCHEMA algorithm [77,130,8] (see 2.…”

“…Proteins sampled from these optimised libraries are more likely to be folded and functional. Minimising the number of residue-residue contacts broken during recombination will tend to partition epistatic interactions into structural sub-units, promoting an additive fitness contribution from each fragment [56,59,92]. In contrast, when using a focused mutagenesis strategy such as saturation mutagenesis [131] this partitioning will not occur, thus increasing the likelihood of non-additive fitness contributions.…”