Protein engineering is often applied to tailor substrate specificity, enantioselectivity, or stability of enzymes according to the needs of a process. In rational engineering approaches, molecular docking and molecular dynamics simulations are often used to compare transition states of wild-type and enzyme variants. Besides affecting the transition state energies by mutations, the entry of the substrate and its positioning in the active site (Michaelis complex) is also often studied, and mutagenesis of residues forming the substrate entry tunnel can have a profound impact on activity and selectivity. In this study, we combine the strengths of such a tunnel approach with MD followed by semiempirical QM calculations that allow the identification of beneficial positions and an in silico screening of possible variants. We exemplify this strategy in the expansion of the substrate scope of Chromobacterium violaceum amine transaminase toward sterically demanding substrates. Two double mutants (F88L/C418­(G/L)) proposed by the modeling showed >200-fold improved activities in the conversion of 1-phenylbutylamine and enabled the asymmetric synthesis of this amine from the corresponding ketone, which was not possible with the wild-type. The correlation of interaction energies and geometrical parameters (distance of the substrate’s carbonyl carbon to the cofactor’s amino group) as obtained in the simulations suggests that this strategy can be used for in silico prediction of variants facilitating an efficient entry and placement of a desired substrate as a first requirement for catalysis. However, when choosing amino acid positions for substitution and modeling, additional knowledge of the enzymatic reaction mechanism is required, as residues that are involved in the catalytic machinery or that guarantee the structural integrity of the enzyme will not be recognized by the developed algorithm and should be excluded manually.

Figure 1 Principal component analysis on LD-pruned data set of 12 013 SNP loci across 66 individuals (one outlier from the Appaloosa breed removed). PCs 1 and 2 explain 2.08 and 2.05% of the variation respectively. QH, Quarter Horse. ECA1, 62.5 kb from TRPM1, also had a high F ST in pairwise comparisons including the Appaloosa. Therefore, although the divergence between breeds is low, these results are evidence that genomic signatures of selection for major, breed-defining phenotypes in the Appaloosa can be detected.