The field of protein engineering primarily aims at improving the functional properties of a protein such as thermo-stability, binding affinity, and/or catalytic activity and has become a vital step in engineering industrial enzymes. In this work, we have developed a machine learning (ML) based approach called MLProScape to build an accurate model of protein fitness landscape (ProScape) and then, use it to design synthetic protein designs with superior functional properties. Unlike directed evolution based approaches which require screening of millions of variants, our method requires screening of just a handful of variants (in the order of tens to hundreds) owing to the power of statistical inference. MLProScape consists of three major steps - i) numerically encode a protein sequence using amino acid based physio-chemical properties as features ii) build a machine learning based model using a subset of highly informative features iii) identify synthetic designs with improved characteristics. To demonstrate the success of our proposed workflow, we have attempted to engineer the catalytic activity of glycoside hydrolases (enzymes that can break down cellulosic and hemi-cellulosic polysaccharides) by using experimentally measured specific activity values corresponding to a diverse set of glycoside hydrolases for training the models. The resulting elastic net regression based models have a high predictive power (with correlation coefficient and R2 values as high as 0.896 and 0.714 respectively between predicted and experimentally measured specific activity values) surpassing previously published ML based enzyme engineering studies. Moreover, the use of position specific features helps us identify positions distal to active site that might play a key role in modulating the activity level. Lastly, this ML based workflow is capable of modeling complex design criteria such as optimizing the protein sequence for hydrolyzing multiple substrates simultaneously as well as to account for other desirable traits such as high stability and better in vivo expression levels.

MLProScape: Machine learning guided approach for engineering enzymes faster

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.