Biology subjects

DeWitt, W. S.

Publications and source records attributed to DeWitt, W. S..

14 recordsLinked to original sources

bioRxiv2026

Additive baselines furnish no evidence for epistasis learning by MULTI-evolve

Recent work from Tran et al. (Science, 2026) introduced MULTI-evolve, a framework for protein engineering that combines single-mutant nomination via a protein language model (PLM) or a deep mutational scan (DMS), experimental single- and double-mutant characterization, and neural networks to engineer hyperactive multimutant proteins. The authors attribute the frameworks performance to "epistasis-aware modeling" and claim that their neural networks "learn the epistatic landscape" and "identify synergistic interactions" from limited double-mutant training data. Additive models, by definition, cannot represent epistasis, making them a natural null baseline for such claims. Here we show that MULTI-evolves multimutant predictions are almost perfectly correlated with an additive models across all three engineering applications (APEX, dCasRx, and HuABC2), such that the engineering of multimutants reduces to combining beneficial mutations with the largest additive effects--a standard protein engineering strategy for over four decades. We also find that MULTI-evolves neural networks do not outperform an additive model on held-out test set predictions, and do not even represent epistasis in their training data. Finally, we revisit a DMS benchmark finding presented as evidence of epistasis learning and show that the same pattern is expected even under a null additive model, due to an elementary statistical phenomenon; when we fit an additive model to the benchmark data, it reproduces the reported pattern. More broadly, our findings underscore the need to benchmark models for machine learning-guided directed evolution against additive null baselines before attributing performance to learned epistasis.

bioinformatics↗

bioRxiv2025

Inference of germinal center evolutionary dynamics via simulation-based deep learning

B cells and the antibodies they produce are vital to health and survival, motivating research on the details of the mutational and evolutionary processes in the germinal centers (GCs) from which mature B cells arise. It is known that B cells with higher affinity for their cognate antigen (Ag) will, on average, tend to have more offspring. However, the exact form of this relationship between affinity and fecundity, which we call the "affinity-fitness response function", is not known. Here we use deep learning and simulation-based inference to learn this function from a unique experiment that replays a particular combination of GC conditions many times. All code is freely available at https://github.com/matsengrp/gcdyn, while datasets and inference results can be found at https://doi.org/10.5281/zenodo.15022130.

evolutionary biology↗

bioRxiv2025

The mutation rate of SARS-CoV-2 is highly variable between sites and is influenced by sequence context, genomic region, and RNA structure

RNA viruses like SARS-CoV-2 have a high mutation rate, which contributes to their rapid evolution. The rate of mutations depends on the mutation type (e.g., A[->]C, A[->]G, etc.) and can vary between sites in the viral genome. Understanding this variation can shed light on the mutational processes at play, and is crucial for quantitative modeling of viral evolution. Using the millions of available SARS-CoV-2 full-genome sequences, we estimate rates of synonymous mutations for all 12 possible nucleotide mutation types and examine how much these rates vary between sites. We find a surprisingly high level of variability and several striking patterns: the rates of four mutation types suddenly increase at one of two gene boundaries; the rates of most mutation types strongly depend on a sites local sequence context, with up to 56-fold differences between contexts; consistent with a previous study, the rates of some mutation types are lower at sites engaged in RNA secondary structure. A simple log-linear model of these features explains [~]15-60% of the fold-variation of mutation rates between sites, depending on mutation type; more complex models only modestly improve predictive power out of sample. We estimate the fitness effect of each mutation based on the number of times it actually occurs versus the number of times it is expected to occur based on the model. We identify several small regions of the genome where synonymous or noncoding mutations occur much less often than expected, indicative of strong purifying selection on the RNA sequence that is independent of protein sequence. Overall, this work expands our basic understanding of SARS-CoV-2s evolution by characterizing the viruss mutation process at the level of individual sites and uncovering several striking mutational patterns that arise from unknown mechanisms.

bioinformatics↗

bioRxiv2024

Learning antibody sequence constraints from allelic inclusion

Antibodies and B-cell receptors (BCRs) are produced by B cells, and are built of a heavy chain and a light chain. Although each B cell could express two different heavy chains and four different light chains, usually only a unique pair of heavy chain and light chain is expressed--a phenomenon known as allelic exclusion. However, a small fraction of naive-B cells violate allelic exclusion by expressing two productive light chains, one of which has impaired function; this has been called allelic inclusion. We demonstrate that these B cells can be used to learn constraints on antibody sequence. Using large-scale single-cell sequencing data from humans, we find examples of light chain allelic inclusion in thousands of naive-B cells, which is an order of magnitude larger than existing datasets. We train machine learning models to identify the abnormal sequences in these cells. The resulting models correlate with antibody properties that they were not trained on, including polyreactivity, surface expression, and mutation usage in affinity maturation. These correlations are larger than what is achieved by existing antibody modeling approaches, indicating that allelic inclusion data contains useful new information. We also investigate the impact of similar selection forces on the heavy chain in mouse, and observe that pairing with the surrogate light chain significantly restricts heavy chain diversity.

immunology↗

bioRxiv2023

Deep mutational scanning of CYP2C19 reveals a substrate specificity-abundance tradeoff

Cytochrome P450s (CYPs) are a family of enzymes responsible for metabolizing nearly 80% of small molecule drugs. Variants in CYPs can substantially alter drug metabolism, which may result in improper dosing and severe adverse drug reactions. CYPs have low sequence conservation, making it difficult to anticipate whether variant effects measured in one CYP may extend to others based on sequence alone. Even closely related CYPs, like CYP2C9 and its closest homolog CYP2C19, have distinct phenotypic properties despite sharing 92% amino acid sequence identity. Thus, we used Variant Abundance by Massively Parallel sequencing (VAMP-seq) to measure the steady-state protein abundance, a proxy for protein stability, of 7,660 missense variants in CYP2C19 expressed in cultured human cells. Our results confirmed positions and structural features critical for CYP function and revealed how variants at positions conserved across all eukaryotic CYPs influence abundance. We jointly analyzed 4,670 variants whose abundance was measured in both CYP2C19 and CYP2C9, finding that the homologs have different variant abundances in substrate recognition sites within the hydrophobic core, and that substitutions in some regions reduced abundance in CYP2C19 but not CYP2C9. We also measured the abundance of all single and some multiple WT amino acid exchanges between CYP2C19 and CYP2C9. While most exchanges had no effect, substitutions in substrate recognition site 4 (SRS4) reduced abundance in CYP2C19. When nearby amino acids were exchanged in double and triple mutants, we found distinct interactions between the sites in CYP2C19 and CYP2C9, revealing a region that is partially responsible for the difference in thermodynamic stability between the two homologs. Since these positions are also important for determining substrate specificity, there may be an evolutionary tradeoff between stability and altered enzymatic function. Finally, we used our data to analyze 368 previously unannotated human variants, finding that 43% had decreased abundance. Thus, by comparing variant effects between two closely related and important human genes, we have uncovered regions underlying their functional differences and paved the way for a more complete understanding of one of the most versatile families of enzymes.

genetics↗

bioRxiv2023

Jointly modeling deep mutational scans identifies shifted mutational effects among SARS-CoV-2 spike homologs

Deep mutational scanning (DMS) is a high-throughput experimental technique that measures the effects of thousands of mutations to a protein. These experiments can be performed on multiple homologs of a protein or on the same protein selected under multiple conditions. It is often of biological interest to identify mutations with shifted effects across homologs or conditions. However, it is challenging to determine if observed shifts arise from biological signal or experimental noise. Here, we describe a method for jointly inferring mutational effects across multiple DMS experiments while also identifying mutations that have shifted in their effects among experiments. A key aspect of our method is to regularize the inferred shifts, so that they are nonzero only when strongly supported by the data. We apply this method to DMS experiments that measure how mutations to spike proteins from SARS-CoV-2 variants (Delta, Omicron BA.1, and Omicron BA.2) affect cell entry. Most mutational effects are conserved between these spike homologs, but a fraction have markedly shifted. We experimentally validate a subset of the mutations inferred to have shifted effects, and confirm differences of >1,000-fold in the impact of the same mutation on spike-mediated viral infection across spikes from different SARS-CoV-2 variants. Overall, our work establishes a general approach for comparing sets of DMS experiments to identify biologically important shifts in mutational effects. Significance StatementAmino-acid mutations to a protein have effects that can shift as the protein evolves or is put under new selective pressure. The effects of amino-acid mutations to a specific protein under a defined selective pressure can be measured by deep mutational scanning experiments. Here, we devise an approach to quantify shifts in mutational effects between experiments performed on different homologs (i.e. variants) of the same protein, or on the same protein selected under different conditions. We use this approach to compare experiments performed on three homologs of SARS-CoV-2 spike, identifying mutations that have shifted in their effect on spike-mediated viral infection by >1,000 fold across SARS-CoV-2 variants.

molecular biology↗

bioRxiv2022

A biophysical model of viral escape from polyclonal antibodies

A challenge in studying viral immune escape is determining how mutations combine to escape polyclonal antibodies, which can potentially target multiple distinct viral epitopes. Here we introduce a biophysical model of this process that partitions the total polyclonal antibody activity by epitope, and then quantifies how each viral mutation affects the antibody activity against each epitope. We develop software that can use deep mutational scanning data to infer these properties for polyclonal antibody mixtures. We validate this software using a computationally simulated deep mutational scanning experiment, and demonstrate that it enables the prediction of escape by arbitrary combinations of mutations. The software described in this paper is available at https://jbloomlab.github.io/polyclonal.

evolutionary biology↗

bioRxiv2022

Increased mutation rate and interlocus gene conversion within human segmental duplications.

Single-nucleotide variants (SNVs) within segmental duplications (SDs) have not been systematically assessed because of the difficulty in mapping short-read sequence data to virtually identical repetitive sequences. Using 102 phased human haplotypes, we constructed 1:1 unambiguous alignments spanning high-identity SDs and compared the pattern of SNVs between unique and SD regions. We find that human SNVs are elevated 60% in SDs compared to unique regions. We estimate that at least 23% of this increase is due to interlocus gene conversion (IGC) with >7 Mbp of SD sequence converted on average per human haplotype. We develop a genome-wide map of IGC donors and acceptors, including 498 acceptor and 454 donor hotspots affecting the exons of ~800 protein-coding genes. The latter includes 171 genes that have "relocated" on average 1.61 Mbp in a subset of human haplotypes. Using a coalescent framework, we show that SD regions are evolutionarily older when compared to unique sequences with most of this signal originating from putative IGC loci. SNVs within SDs, however, also exhibit a distinct mutational spectrum where there is a 27.1% increase in transversions that convert cytosine to guanine or the reverse across all triplet contexts. In addition, we observe a 7.6% reduction in the frequency of CpG associated mutations when compared to unique DNA. We hypothesize that these distinct mutational properties help to maintain an overall higher GC content of SD DNA when compared to unique DNA, and we show that these GC-favoring mutational events are likely driven by GC-biased conversion between paralogous sequences.

genomics↗

bioRxiv2020

Robustness of phylogenetic inference to model misspecification caused by pairwise epistasis

Likelihood-based phylogenetic inference posits a probabilistic model of character state change along branches of a phylogenetic tree. These models typically assume statistical independence of sites in the sequence alignment. This is a restrictive assumption that facilitates computational tractability, but ignores how epistasis, the effect of genetic background on mutational effects, influences the evolution of functional sequences. We consider the effect of using a misspecified site-independent model on the accuracy of Bayesian phylogenetic inference in the setting of pairwise-site epistasis. Previous work has shown that as alignment length increases, tree reconstruction accuracy also increases. Here, we present a simulation study demonstrating that accuracy increases with alignment size even if the additional sites are epistatically coupled. We introduce an alignment-based test statistic that is a diagnostic for pair-wise epistasis and can be used in posterior predictive checks.

evolutionary biology↗

bioRxiv2020

mutyper: assigning and summarizing mutation types for analyzing germline mutation spectra

SummaryCharacterization of germline mutation spectrum variation from population genomics data has shed light on the biological complexity of the mutation process, and its evolution within and between species. This analysis augments available population SNP data with estimates of local ancestral genomic context to assign mutation types and aggregate summary statistics thereof, and is increasingly common. There is a need for standardized computational tools to extract mutation spectrum information from sequencing data. Here I describe mutyper, a command-line utility and Python package that uses an ancestral genome estimate to assign mutation types to SNP data, compute mutation spectra for individuals, and compute sample frequency spectra resolved by mutation type for population genetic inference. Availability and implementationmutyper can be installed using the pip package manager and is compatible with Python 3.6+. Documentation is provided at https://harrispopgen.github.io/mutyper; source code is available at https://github.com/harrispopgen/mutyper.

bioinformatics↗

bioRxiv2020

Joint nonparametric coalescent inference of mutation spectrum history and demography

Booming and busting populations modulate the accumulation of genetic diversity, encoding histories of living populations in present-day variation. Many methods exist to decode these histories, and all must make strong model assumptions. It is typical to assume that mutations accumulate uniformly across the genome at a constant rate that does not vary between closely related populations. However, recent work shows that mutational processes in human and great ape populations vary across genomic regions and evolve over time. This perturbs the mutation spectrum: the relative mutation rates in different local nucleotide contexts. Here, we develop theoretical tools in the framework of Kingmans coalescent to accommodate mutation spectrum dynamics. We describe mushi: a method to perform fast, nonparametric joint inference of demographic and mutation spectrum histories from allele frequency data. We use mushi to reconstruct trajectories of effective population size and mutation spectrum divergence between human populations, identify mutation signatures and their dynamics in different human populations, and produce more accurate time calibration for a previously-reported mutational pulse in the ancestors of Europeans. We show that mutation spectrum histories can be productively incorporated in a well-studied theoretical setting, and rigorously inferred from genomic variation data like other features of evolutionary history.

evolutionary biology↗

bioRxiv2019

Estimation of cell lineage trees by maximum-likelihood phylogenetics

CRISPR technology has enabled large-scale cell lineage tracing for complex multicellular organisms by mutating synthetic genomic barcodes during organismal development. However, these sophisticated biological tools currently use ad-hoc and outmoded computational methods to reconstruct the cell lineage tree from the mutated barcodes. Because these methods are agnostic to the biological mechanism, they are unable to take full advantage of the datas structure. We propose a statistical model for the mutation process and develop a procedure to estimate the tree topology, branch lengths, and mutation parameters by iteratively applying penalized maximum likelihood estimation. In contrast to existing techniques, our method estimates time along each branch, rather than number of mutation events, thus providing a detailed account of tissue-type differentiation. Via simulations, we demonstrate that our method is substantially more accurate than existing approaches. Our reconstructed trees also better recapitulate known aspects of zebrafish development and reproduce similar results across fish replicates.

bioinformatics↗

bioRxiv2018

Human T cell receptor occurrence patterns encode immune history, genetic background, and receptor specificity

The T cell receptor (TCR) repertoire encodes immune exposure history through the dynamic formation of immunological memory. Statistical analysis of repertoire sequencing data has the potential to decode disease associations from large cohorts with measured phenotypes. However, the repertoire perturbation induced by a given immunological challenge is conditioned on genetic background via major histocompatibility complex (MHC) polymorphism. We explore associations between MHC alleles, immune exposures, and shared TCRs in a large human cohort. Using a previously published repertoire sequencing dataset augmented with high-resolution MHC genotyping, our analysis reveals rich structure: striking imprints of common pathogens, clusters of co-occurring TCRs that may represent markers of shared immune exposures, and substantial variations in TCR-MHC association strength across MHC loci. Guided by atomic contacts in solved TCR:peptide-MHC structures, we identify sequence covariation between TCR and MHC. These insights and our analysis framework lay the groundwork for further explorations into TCR diversity.

immunology↗

bioRxiv2017

Unbiased Definition Of A Shared T-Cell Receptor Motif Enables Population-Based Studies Of Tuberculosis

Peptide-specific T cells that are restricted by highly polymorphic major histocompatibility complex (MHC) proteins express diverse T-cell receptors (TCRs) that are rarely shared among unrelated individuals. T-cells can also recognize bacterial lipid antigens that bind the relatively non-polymorphic CD1 family of proteins. However, genetic variation in human CD1 genes and TCR diversity expressed by CD1-restricted T-cells have not been quantitatively determined. Here, we show that CD1B is nearly nucleotide-identical across all five continental ancestry groups, providing evidence for purifying selection during human evolution. We used CD1B tetramers loaded with a mycobacterial glycolipid antigen to isolate T-cells from four genetically unrelated South African adults and cataloged thousands of TCRs from in-vitro expanded T-cells using immunosequencing. We identified highly conserved motifs that were co-expressed as a functional heterodimer and significantly enriched among tetramer-positive T-cells sorted directly from peripheral blood. Finally, we show that frequencies of these TCR motifs are increased in the blood of patients with active tuberculosis compared to uninfected controls, a finding that is confirmed by ex-vivo frequencies of tetramer-positive T-cells determined by flow cytometry. These data provide a framework for unbiased definition of TCRs targeting lipid antigens, which can be tested for clinical associations independently of host genetic background.\n\nBrief SummaryWe used human genetics and immunosequencing to define a shared T-cell receptor motif that is specific for a mycobacterial lipid antigen and associated with tuberculosis independently of host genetic background.

immunology↗