Predictive profiling of SARS-CoV-2 variants by deep mutational learning

Taft, J. M.; Weber, C. R.; Gao, B.; Ehling, R. A.; Han, J.; Frei, L.; Metcalfe, S. W.; Yermanos, A.; Kelton, W.; Reddy, S. T.

doi:10.1101/2021.12.07.471580

Home/Archive/bioRxiv

bioRxiv · 10.1101/2021.12.07.471580

Predictive profiling of SARS-CoV-2 variants by deep mutational learning

Taft, J. M.·Weber, C. R.·Gao, B.·Ehling, R. A.·Han, J.·Frei, L.·Metcalfe, S. W.·Yermanos, A.·Kelton, W.·Reddy, S. T.

View original source ↗Export metadata ↓

Abstract

The continual evolution of the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and the emergence of variants that show resistance to vaccines and neutralizing antibodies (1-4) threaten to prolong the coronavirus disease 2019 (COVID-19) pandemic (5). Selection and emergence of SARS-CoV-2 variants are driven in part by mutations within the viral spike protein and in particular the ACE2 receptor-binding domain (RBD), a primary target site for neutralizing antibodies. Here, we develop deep mutational learning (DML), a machine learning-guided protein engineering technology, which is used to interrogate a massive sequence space of combinatorial mutations, representing billions of RBD variants, by accurately predicting their impact on ACE2 binding and antibody escape. A highly diverse landscape of possible SARS-CoV-2 variants is identified that could emerge from a multitude of evolutionary trajectories. DML may be used for predictive profiling on current and prospective variants, including highly mutated variants such as omicron (B.1.1.529), thus supporting decision making for public heath as well as guiding the development of therapeutic antibody treatments and vaccines for COVID-19.

Explore related subjects

synthetic biology Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Taft, J. M., Weber, C. R., Gao, B., Ehling, R. A., Han, J., Frei, L., Metcalfe, S. W., Yermanos, A., Kelton, W., Reddy, S. T.. 2021-12-09. Predictive profiling of SARS-CoV-2 variants by deep mutational learning. https://doi.org/10.1101/2021.12.07.471580

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Creation of genomic chimeras via megabase scale genome expansion

An organism's physiology is controlled by its genomic content. Building genomic chimeras would enable melding of the diverse functions and properties of life. However, prior arts in genome synthesis are limited to reconstituting pre-existing functions within a singular genome rather than combining diverse genomic functions across multiple distinct genomes. Existing methods are also prohibitively expensive, labor intensive, and time consuming, and thus fundamentally not scalable for creating large synthetic chimeric genomes. To address all these limitations, we invented Additive Conjugative-CAST Engineering (ACE) combining conjugation with CRISPR-associated transposition (CAST) to deliver and integrate up to half a genome per step from a donor cell into a precisely defined position in the recipient's genome. We have demonstrated ACE's capability to integrate at least a 2-megabase donor genomic segment in a single step and at least 3 megabase in two consecutive steps. Importantly, we confirmed ACE's generality through the creation of multiple genomic chimeras across species, genus, and at least order barriers. With ACE, we further demonstrate such chimeric organisms, or what we call genome expanded organisms (GEOs), can be stably forged from at least three starting bacterial strains, and can be stably maintained to express all their acquired genomic parts in a single chimera. ACE offers a new paradigm of creating artificial lifeforms to combine desired properties and potentially create novel functions beyond the constraints of nature. ACE also offers a unique perspective to probe unexplored plasticity in genome architecture and expression patterns of novel, massively expanded genomes.

synthetic biology↗

bioRxiv2026

Functional analysis of small-molecule and protein binding domains using monomeric streptavidin-based two-component biosensors

Protein biosensors are important analytical and diagnostic tools that convert molecular recognition into measurable activity. Most common two-component biosensors require two non-competing recognition domains, yet binding affinity and specificity alone do not reliably predict their compatibility with the rest of the biosensor. We report a modular platform that exploits monomeric streptavidin (mSA2) as a surrogate recognition domain, enabling rapid functional evaluation of candidate binders in the context of biosensor scaffold without requiring a second ligand-specific binding protein. We established a chromatography-free protocol for recovering functional mSA2-based biosensor components from Escherichia coli. Using biotinylated small molecules methotrexate and vitamin D3, as well as heart failure protein biomarkers galectin-3 and S100A7 we demonstrate ligand-dependent reporter biosensor activation. The developed platform provides a scalable approach for rapid binder evaluation and biosensor prototyping, streamlining biosensor development using binding domains generated by directed evolution or computational protein design.

synthetic biology↗

bioRxiv2026

Developing synthetic auxin-based trans-kingdom signaling

Synthetic trans-kingdom signaling systems with tunable transduction properties provide a unique avenue to explore how communication between organisms coordinate behaviors without the confounding complexity of natural systems. However, existing trans-kingdom signaling systems suffer from either toxic signal molecules, a need for rare precursors for signal generation, or transduction machinery which is either species-specific or cannot be easily tuned. This limits their capacity to be broadly applied across kingdoms and their utility as a model system to study trans-kingdom interactions. In this work we elucidate the design principles for a trans-kingdom signaling system based on the plant phytohormone, Indole 3-Acetic Acid (auxin), that overcomes these challenges. We demonstrate how auxin biosynthesis can be engineered into a model sender organism, the cyanobacteria Synechococcus elongatus PCC 7942, and how this signal can be tuned with low leak and high dynamic range. We also show these signals can be connected to gene expression using tunable synthetic auxin-responsive signaling systems across multiple kingdoms. We demonstrate that S. elongatus generated auxin signals can tune gene expression in co-cultured Saccharomyces cerevisiae via auxin-induced degradation. Additionally, we show that when the S. elongatus are associated with the roots of the model plant, Arabidopsis thaliana, the auxin they generate can control gene expression via auxin-responsive synthetic transcription factors expressed from the plants genome. Taken together, our results demonstrate the design principles for a tunable auxin-based trans-kingdom signaling system that can be broadly applied across species.

synthetic biology↗