PSE: an approach to improve the search for pairs of stabilizing mutations for in silico protein engineering using structural signatures and dimensional reduction

Protein engineering is a technique adopted in the production of synthetic proteins (mutants). Synthetic proteins have a wide range of functionalities for the industry, from the manufacture of new medicines to the production of second-generation biofuel. In some cases, the construction of mutants can improve the stability, solubility, resistance to inhibition, and also keep the activity of proteins even in unusual values of pH and temperature. Although mutations effects must be verified through experimental method, the number of possible mutations for each residue make this approach costly and time-consuming. The use of bioinformatics techniques is a promising alternative since several mutations can be proposed and later evaluated in silico as a screening prior to experimentation. Recently, our group has proposed a software called Proteus (Protein Engineering Supporter). Proteus collected pairs of residues in contact (hydrogen bond) of the Protein Data Bank (PDB), clustered them to remove redundancy, and constructed a database of residues in contact and their neighbors. Proteus hypothesizes that if the main chain of a pair of residues in contact, and their preceding and following residues present low RMSD compared to a target protein, these pairs of residues could be transferred to the target protein, improving its stability. However, Proteus requires structural alignment to compare the main chain of these residues, which generate a high computational cost. For example, a protein of 400 residues could present up to 159,600 pairs of residues that could perform a contact. In this example, Proteus would align all these pairs of residues against the ProteusDB. This process could take weeks, even using high-performance servers. To try improving the search, we propose here a method called PSE (Proteus Search Engine). PSE uses structural signatures to represent each tridimensional structure of residues in contact. PSE also applies dimensional reduction using SVD and uses Euclidean distance to construct a ”search ray", that could reduce the number of contacts tested. We performed a case study using the structure of bacteriophage t4 lysozyme (PDB: 2LZM) and detected that Proteus could run in a few hours using only one core of processing. Our new search method will allow the execution of Proteus in large scale, proposing mutations to be tested that could improve the activity of proteins with several applications for the industry.