Evgenii Lobzaev

Engineering proteins is a challenging task requiring the exploration of a vast design space. Traditionally, this is achieved using Directed Evolution (DE), which is a laborious process. Generative deep learning, instead, can learn biological features of functional proteins from sequence and structural datasets and return novel variants. However, most models do not generate thermodynamically stable proteins, thus leading to many non-functional variants. Here we propose a model called PRotein Engineering by Variational frEe eNergy approximaTion (PREVENT), which generates stable and functional variants by learning the sequence and thermodynamic landscape of a protein. We evaluate PREVENT by designing 40 variants of the conditionally essential E. coli phosphotransferase N-acetyl-L-glutamate kinase (EcNAGK). We find 85% of the variants to be functional, with 55% of them showing similar growth rate compared to the wildtype enzyme, despite harbouring up to 9 mutations. Our results support a new approach that can significantly accelerate protein engineering.

Engineering proteins with desired functions and biochemical properties is pivotal for biotechnology and drug discovery. While computational methods based on evolutionary information are reducing the experimental burden by designing targeted libraries of functional variants, they still have a low success rate when the desired protein has few or very remote homologous sequences. Here we propose an autoregressive model, called Temporal Dirichlet Variational Autoencoder (TDVAE), which exploits the mathematical properties of the Dirichlet distribution and temporal convolution to efficiently learn high-order information from a functionally related, possibly remotely similar, set of sequences. TDVAE is highly accurate in predicting the effects of amino acid mutations, while being significantly 90% smaller than the other state-of-the-art models. We then use TDVAE to design variants of the human alpha galactosidase enzymes as potential treatment for Fabry disease. Our model builds a library of diverse variants which retain sequence, biochemical and structural properties of the wildtype protein, suggesting they could be suitable for enzyme replacement therapy. Taken together, our results show the importance of accurate sequence modelling and the potential of autoregressive models as protein engineering and analysis tools.

Type 2 diabetes mellitus (T2D) presents a major health and economic burden that could be alleviated with improved early prediction and intervention. While standard risk factors have shown good predictive performance, we show that the use of blood-based DNA methylation information leads to a significant improvement in the prediction of 10-year T2D incidence risk. Previous studies have been largely constrained by linear assumptions, the use of cytosine–guanine pairs one-at-a-time and binary outcomes. We present a flexible approach (via an R package, MethylPipeR) based on a range of linear and tree-ensemble models that incorporate time-to-event data for prediction. Using the Generation Scotland cohort (training set n_cases = 374, n_controls = 9,461; test set n_cases = 252, n_controls = 4,526) our best-performing model (area under the receiver operating characteristic curve (AUC) = 0.872, area under the precision-recall curve (PRAUC) = 0.302) showed notable improvement in 10-year onset prediction beyond standard risk factors (AUC = 0.839, precision–recall AUC = 0.227). Replication was observed in the German-based KORA study (n = 1,451, n_cases = 142, P = 1.6 × 10⁻⁵).

Summary: We describe construction of the 660 kilobase synthetic yeast chromosome XI ( synXI ) and reveal how synthetic redesign of non-coding DNA elements impact the cell. To aid construction from synthesized 5 to 10 kilobase DNA fragments, we implemented CRISPR-based methods for synthetic crossovers in vivo and used these methods in an extensive process of bug discovery, redesign and chromosome repair, including for the precise removal of 200 kilobases of unexpected repeated sequence. In synXI , the underlying causes of several fitness defects were identified as modifications to non-coding DNA, including defects related to centromere function and mitochondrial activity that were subsequently corrected. As part of synthetic yeast chromosome design, loxPsym sequences for Cre-mediated recombination are inserted between most genes. Using the GAP1 locus from chromosome XI , we show here that targeted insertion of these sites can be used to create extrachromosomal circular DNA on demand, allowing direct study of the effects and propagation of these important molecules. Construction and characterization of synXI has uncovered effects of non-coding and extrachromosomal circular DNA, contributing to better understanding of these elements and informing future synthetic genome design.

Enzymatic deficiencies cause the accumulation of toxic levels of substrates in a cell and are associated with life-threatening pathologies. Restoring physiological enzymes levels by injecting a recombinant version of the defective enzyme could provide a viable therapeutic option. However, these enzyme replacement therapies have had limited success, as the recombinant enzymes are less catalytically active, cause immune response and are difficult to manufacture. Moreover, the vast sequence design space makes finding enzymes with desired therapeutic properties extremely challenging. Here, we present a new enzyme engineering framework, which builds on recent advances in deep learning, variational calculus and natural language processing, to design variants of human enzymes with biochemical features comparable to the wild type protein as a way to rapidly build targeted libraries for downstream screening. We applied our method to design variants of human Sphyngosine-1-phosphate lyase (HsS1PL) as potential therapeutic treatments for nephrotic syndrome type 14 (NPHS14), and characterized their biochemical properties through extensive sequence and molecular dynamics analyses.