Abstract

Functional proteins for a variety of useful applications, as binders and catalysts, are required, but currently not known. Functional metagenomics and directed evolution promise access to such new proteins, but the chances of finding them are low. Therefore high-throughput technologies are crucial to beat the odds: screening in picoliter water-in–oil emulsion droplets produced at kHz rates in microfluidic devices allow screening of >10e7 clones and permit successful selections. While much faster than traditional screening approaches, the vastness of sequence space (and the scarcity of ‘solutions’ in it) require strategies for the identification and interconversion of enzymes. Decoding the selected clones by short or long-read sequencing methods leads to large sequence-function datasets that may allow extrapolation from experimental directed evolution to further improved mutants beyond the observed hits. We will introduce experimental approaches for drawing up ‘fitness landscapes’ that illustrate trajectories in sequence space. Starting with large datasets harvested from droplet microfluidics, this information is used for AI/ML-assisted enzyme engineering, allowing meaningful predictions that accelerate biocatalyst engineering and leveraging the synergy of experimental and in silico approaches.