Bettina M. Nestl

Machine learning (ML) is rapidly turning into a key technology for biocatalysis. By learning patterns in amino acid sequences, protein structures, and functional data, ML models can help navigate complex fitness landscapes, uncover new enzymes in databases, and even design biocatalysts de novo. Along with advances in DNA synthesis and sequencing, laboratory automation, and high-throughput screening, ML is increasing the speed and efficiency of enzyme development. In this Outlook, we highlight recent applications of ML in the fields of enzyme discovery, design, and engineering, with a focus on current challenges and emerging solutions. Furthermore, we discuss barriers that impede a broader and faster adoption of ML-based workflows in the biocatalysis community. We conclude by suggesting best practices for fostering effective collaborations in this interdisciplinary field.

Reductive amination catalysed by imine reductase (IRED) and reductive aminase (RedAm) enzymes has recently been established as a powerful method for the asymmetric synthesis of chiral amines. While this biocatalytic technology has rapidly progressed from proof of concept to initial industrial applications, its scope and limitations remain to be fully explored. In this work, we report a broad and systematic profiling of reductive amination performance in the sequence space of IREDs and RedAms. This investigation employed an iterative strategy for activity and stereoselectivity screening, guided by chemo- and bioinformatic modelling as well as machine learning. By evaluating the catalytic performance of 175 IREDs against structurally diverse panels of 36 carbonyl compounds and 24 amines, we show that the majority of these enzymes is capable of asymmetric reductive amination at equimolar concentrations of the two substrates (50 mM each). The most effective enzymes identified in this study display sequence characteristics of RedAms, are active on 29–42% of the analysed substrate combinations, and combine high specific activities for the most favourable substrate pair (1.7−27.7 U/mgIRED) with excellent stereoselectivity. Beyond assembling this high-performance enzyme panel, we demonstrate extrapolation from our collected screening data to new substrate combinations by deep learning and the scale-up of selected reactions to a preparative batch size (10 mmol substrate, 200 mL reaction volume), delivering gram amounts of reductive amination products in high yield (63–89%) and optical purity (98% to >99% ee).

Light‐dependent fatty acid photodecarboxylases (FAPs) hold significant potential for biotechnology, due to their capability to produce alka(e)nes directly from the corresponding (un)saturated natural fatty acids requiring light as the only reagent. This study expands the family of FAPs through cavity‐based enzyme discovery methods. Thirty enzyme candidates with potential photodecarboxylation activity were identified by matching the cavities of four related template structures against the Protein Data Bank's flavoproteins, a library of proteins identified via the Foldseek Search Server, and homology models of sequences resulting from BLAST. Subsequent docking experiments narrowed this library to ten promising enzymes, which were expressed and assessed in vitro, identifying four photodecarboxylases. Out of these enzymes, the GMC oxidoreductase from Coccomyxa sp. Obi (CoFAP) was characterized in detail, which revealed high activity in the decarboxylation reactions of palmitic acid and octanoic acid and a broad pH tolerance (pH 6.5–9.5).

Imine reductases (IREDs) provide promising opportunities for the synthesis of various chiral amines. Initially, asymmetric imine reduction was reported, followed by reductive aminations of aldehydes and ketones via imines. Herein we present the reductive amination of structurally diverse carbonyls and dicarbonyls with hydrazines (reductive hydrazination), catalyzed by the IRED from Myxococcus stipitatus. In analogy to IRED‐catalyzed reductive aminations, various carbonyls and dicarbonyls could react with simple hydrazines to produce substituted acyclic and cyclic N‐alkylhydrazines. By incorporating and scaling up a hydrogenase cofactor regeneration system, we demonstrated the scalability and atom‐efficiency of an H₂‐driven double reductive hydrazination, highlightling the potential of IREDs in biocatalysis.

Recent advances in bioeconomy allow a holistic view of existing and new process chains and enable novel production routines continuously advanced by academia and industry. All this progress benefits from a growing number of prediction tools that have found their way into the field. For example, automated genome annotations, tools for building model structures of proteins, and structural protein prediction methods such as AlphaFold2^TM or RoseTTAFold have gained popularity in recent years. Recently, it has become apparent that more and more AI‐based tools are being developed and used for biocatalysis and biotechnology. This is an excellent opportunity for academia and industry to accelerate advancements in the field further. Biotechnology, as a rapidly growing interdisciplinary field, stands to benefit greatly from these developments.