Research and development in drug discovery will need to find significant efficiency gains if the industry is to continue generating novel drugs. There is great expectation for machine learning (ML) to provide this boost in R&D productivity, but to harness the full potential of ML, the generation of new, high-quality datasets will be necessary. Here, the authors present a platform that combines high-throughput display and selection data generation with ML. More specifically, deep learning is used to inform the directed evolution of novel biotherapeutics using DNA library synthesis, ultra-high throughput selections, and next generation sequencing. By combining the learnings of multiple in silico models, their platform enables multi-parameter optimisation across multiple important protein characteristics. They also present a model for benchmarking these ML-driven drug discovery platforms according to the accuracy of their underlying in silico models, in conjunction with the throughput of their empirical experimentation.

Plant evolution has produced enzymes that may not be optimal for maximizing yield and quality in today\'s agricultural environments and plant biotechnology applications. By improving enzyme performance, it should be possible to alleviate constraints on yield and quality currently imposed by kinetic properties or enzyme instability. Enzymes can be optimized more quickly than naturally possible by applying directed evolution, which entails mutating a target gene in vitro and screening or selecting the mutated gene products for the desired characteristics. Continuous directed evolution is a more efficient and scalable version that accomplishes the mutagenesis and selection steps simultaneously in vivo via error-prone replication of the target gene and coupling of the host cell\'s growth rate to the target gene\'s function. However, published continuous systems require custom plasmid assembly, and convenient multipurpose platforms are not available. We discuss two systems suitable for continuous directed evolution of enzymes, OrthoRep in Saccharomyces cerevisiae and EvolvR in Escherichia coli, and our pilot efforts to adapt each system for high-throughput plant enzyme engineering. To test our modified systems, we used the thiamin synthesis enzyme THI4, previously identified as a prime candidate for improvement. Our adapted OrthoRep system shows promise for efficient plant enzyme engineering.

Protein-protein interactions (PPIs) are involved in nearly all cellular processes. PPIs are particularly crucial for mediating selectivity along signaling pathways. Thus, measuring the competitive interplay between PPIs in a cell is important for both understanding fundamental cellular regulation and developing therapeutics targeting those whose dysregulation is associated with disease. A variety of split protein reporter-based tools are available to measure if two proteins interact within a cell and thereby characterize the general determinants of their interactions. PPIs, however, occur within complex networks facilitated by dynamic biophysical nuances that determine activity and selectivity. Evolved, proximity-dependent split T7 RNA polymerase (RNAP) biosensors have recently been used to perform deep mutational scanning of PPI interfaces, and to create synthetic gene circuits. In this chapter, we present the application of proximity-dependent split RNAP biosensors as a method to measure multidimensional PPIs in live cells. Orthogonal split RNAP \"tags\" encode each interaction in a unique RNA signal, thereby enabling the study of multiple competitive PPIs in live cells. Each unique RNA signal can be quantified via established RNA analysis methods. Herein, we provide advice and protocols to aid other researchers in using the split RNAP biosensor, focusing primarily on how to detect multiple PPIs in mammalian cells, including their dynamic interplay in the presence of small molecule inhibitors.

A collection of the editors of Journal of Molecular Evolution have gotten together to pose a set of key challenges and future directions for the field of molecular evolution. Topics include challenges and new directions in prebiotic chemistry and the RNA world, reconstruction of early cellular genomes and proteins, macromolecular and functional evolution, evolutionary cell biology, genome evolution, molecular evolutionary ecology, viral phylodynamics, theoretical population genomics, somatic cell molecular evolution, and directed evolution. While our effort is not meant to be exhaustive, it reflects research questions and problems in the field of molecular evolution that are exciting to our editors.

The work aiming to unravel the correlation between protein sequence and function in the absence of structural information can be highly rewarding. We present a new way of considering descriptors from the amino acids index database for modeling and predicting the fitness value of a polypeptide chain. This approach includes the following steps: (i) Calculating Q elementary numerical sequences (Ele_SEQ) depending on the encoding of the amino acid residues, (ii) determining an extended numerical sequence (Ext_SEQ) by concatenating the Q elementary numerical sequences, wherein at least one elementary numerical sequence is a protein spectrum obtained by applying fast Fourier transformation (FFT), and (iii) predicting a value of fitness for polypeptide variants (train and/or validation set). These new descriptors were tested on four sets of proteins of different lengths (GLP-2, TNF alpha, cytochrome P450, and epoxide hydrolase) and activities (cAMP activation, binding affinity, thermostability and enantioselectivity). We show that the use of multiple physicochemical descriptors coupled with the implementation of the FFT, taking into account the interactions between residues of amino amides within the protein sequence, could lead to very significant improvement in the quality of models and predictions. The choice of the descriptor or of the combination of descriptors and/or FFT is dependent on the couple protein/fitness. This approach can provide potential users with value added to existing mutant libraries where screening efforts have so far been unsuccessful in finding improved polypeptide mutants for useful applications.

Positions identified in directed evolution campaigns or by (semi)rational design can be recombined iteratively or simultaneously. Iterative recombination has yielded many success stories and is beneficially used if screening capabilities are limited (four iterative SSMs generate 20×4=80 different enzyme variants). Simultaneous site saturation mutagenesis offers significantly higher diversity (204 =160 000 variants) and enables greater improvements to be found, especially if the selected positions are in close proximity to each other (cooperative effects). Here we report a first comprehensive comparison of iterative and simultaneous saturation of four residues in Candida parapsilosis alcohol dehydrogenase 5 (cpADH5) with methyl 3-hydroxyhexanoate as substrate. Screening of 7200 clones from 33 site saturation mutagenesis libraries (exploring 17 recombination paths) yielded the cpADH5 W286A variant, with a 82-fold improved initial activity toward methyl 3-hydroxyhexanoate. Screening 3500 clones from a single OmniChange library with four positions (C57, W116, L119, and W286; 1.8 % of the generated sequence space) yielded the cpADH5 C57V/W286S variant, with a 108-fold improvement in initial activity toward methyl 3-hydroxyhexanoate. A 1.8 % coverage of the sequence space of the simultaneous multisite saturation library was, in comparison to the investigated 17 recombination paths, sufficient to identify a cpADH5 variant with improved activity.

Directed evolution is an incredibly powerful strategy for engineering enzyme function. Applying this approach to glycosidases offers enormous potential for the development of highly specialized tools in chemical glycobiology. Performing enzyme directed evolution requires the generation, by random mutagenesis, of mutant libraries from which large numbers of variant enzymes must be screened in high-throughput assays. A structure-guided \"semirational\" method for library creation allows researchers to target specific amino acid positions for mutagenesis, concentrating mutations where they might be most effective in order to produce mutant libraries of a manageable size, minimizing screening effort while maximizing the chances of finding improved mutants. Well-designed assays, which may use specially prepared substrates, enable efficient screening of these mutant libraries. This chapter will detail general methods in the structure-guided directed evolution of glycosidases, which have previously been employed in engineering a blood group antigen-cleaving enzyme.

Synonymous mutations in protein coding genes significantly impact translation efficiency. We synthesized a pair of genes encoding green fluorescent protein that were separated by 160 synonymous mutations to investigate key factors that affect translation efficiency. One sequence was optimized for Escherichia coli (GFP(Eco)) and the other for Bacillus subtilis (GFP(Bsu)). When the genes were expressed in E. coli, GFP(Eco) fluoresced 12-fold stronger than GFP(Bsu), confirming the suboptimal nature of the GFP(Bsu) gene. We then employed directed evolution to improve the expression of GFP(Bsu). Random mutagenesis and DNA shuffling was used to generate mutant libraries, which were screened for fluorescence. A variant showing 6-fold fluorescence enhancement was identified, which contained a single mutation (G10A) in a rare codon for Gly-4. However, the substitution generated another type of rare codon, AGA, for Arg, suggesting that the improvement was caused by a factor other than the rare codon. We next applied saturation mutagenesis to Gly-4. The darkest variant contained a GGG codon (GFP(Bsu)-G) for Gly-4. Taking the location of the mutation into account, we hypothesized that destabilization of the mRNA secondary structure around the initiation codon improved the expression. We then randomized the nucleotide triplet in 5\'-untranslated region (5\'UTR) of GFP(Bsu), which is complementary to the Gly-4 codon. A variant showing 6-fold fluorescence enhancement was identified, which exhibited a destabilized secondary structure. When this 5\'UTR sequence was combined with GFP(Bsu)-G, 22-fold fluorescent improvement was achieved. Collectively, the stability of the mRNA secondary structure around the initiation codon predominantly affected the translation efficiency.