Engineering Dehalogenase Enzymes using Variational Autoencoder-Generated Latent Spaces and Microfluidics

Abstract

Enzymes play a crucial role in sustainable industrial applications, with their optimization posing a formidable challenge due to the intricate interplay among residues. Computational methodologies predominantly rely on evolutionary insights, leveraging homologous sequences to pinpoint conserved and functionally critical regions. However, despite their notable advancements, deciphering the evolutionary variability and complex dependencies among residues presents substantial hurdles. Here, we present a new machine-learning method based on variational autoencoders in combination with a simple evolutionary sampling strategy to address those limitations. We customized our method to generate novel sequences of haloalkane dehalogenases, enzymes widely used in biodegradation, biocatalysis, and biosensing. Three consecutive design-build-test cycles improved the solubility of variants from 11% to 75%. Thorough experimental validation using the state-of-the-art microfluidic device MicroPEX resulted in 20 multiple-point variants. Nine of them, sharing as little as 67% sequence similarity with the template, showed a melting temperature increase of up to 9°C and an average improvement of 3°C. The most stable variant demonstrated a 3.5-fold increase in activity compared to the template, while five variants exhibited average dehalogenase activities. High-quality experimental data collected with 20 variants represent a valuable dataset for the critical validation of novel protein design approaches and scoring functions. Python scripts and data sets are available on GitHub (https://github.com/loschmidt/vae-dehalogenases), and interactive calculations will be possible via an easy-to-use website: https://loschmidt.chemi.muni.cz/fireprotasr/.