Biology subjects

Wheeler, N.

Publications and source records attributed to Wheeler, N..

7 recordsLinked to original sources

bioRxiv2026

Safety First: Input Screening for Protein Design Tools

As biological AI models become more powerful, practical biosecurity approaches are needed to support beneficial applications while reducing misuse risks. Sequence-similarity-based screening approaches are no longer adequate to safeguard biological AI models because these models can design molecules with novel sequences and structures. Therefore, a screening approach that takes function into account is needed. To address this need, we propose a new screening method for AI-enabled protein binder design tools. Our framework screens protein binding targets, with a focus on the human proteome, as opposed to the binder molecule itself. We constructed a database of 14,541 potentially harmful proteoform targets from the human proteome (7.1% of all human protein proteoforms) classified by biosecurity risk level. To discern structural and functional features, we evaluated constructs with an embedding-based screening method using the ESM-C protein language model. ESM-C achieved high accuracy for detecting variants of known targets (F1 scores >97%), with performance similar to BLASTP. However, ESM-C proved to be more effective at capturing functional relationships, distinguishing benign mutations from damaging ones where BLASTP did not. To characterize how screening would affect bioscience research, we measured flagging rates across diverse protein datasets. Flagging rates were significant for mammalian proteins weighted by publication frequency (23% for human, 20% for mouse), and rates for organisms distantly related to humans were minimal (<1.1% for bacteria, fungi, plants, and viruses). Among commercially relevant targets, 63% of antibody patent targets were classified as dual-use, reflecting that therapeutically important proteins often perform critical biological functions. To identify and flag risky user requests from protein binder design tools without placing an undue burden on scientific research and innovation, it will be essential to deploy this screening approach in a way that addresses the overlap our analysis showed between targets of concern and therapeutic targets-possibly in concert with tiered trusted access frameworks. This new method provides a foundation for proportionate safeguards for biological AI models that reduce misuse risks while preserving their benefits for legitimate research and demonstrates a concrete proof of principle that can be generalized to other protein design tools and biological AI models.

synthetic biology↗

bioRxiv2025

Large-scale single-cell phylogenetic mapping of clonal evolution in the human aging esophagus

The human somatic genome evolves throughout our lifespan, producing mosaic individuals comprising clones harboring different mutations across tissues. While clonal expansions in the hematopoietic system have been extensively characterized and reported to be nearly ubiquitous, clonal mosaicism (CM) has more recently also been described across multiple solid tissues. However, outstanding questions remain about the parameters and processes of human somatic evolution in non-cancerous solid human tissues, including when clones arise, how they evolve over time, and what mechanisms lead to their expansion. Questions of timing and clonal dynamics can be addressed through phylogenetic reconstruction, which serves as a temporal microscope, while uncovering the mechanisms of expansion necessitates simultaneous phenotypic profiling. To address this gap, here we develop Single-cell Miniaturized Automated Reverse Transcription and Primary Template-directed Amplification (SMART-PTA) for joint single-cell whole-genome and whole-transcriptome sequencing for large scale and cost efficient interrogation of solid tissue CM. We established a workflow that generates hundreds of matched single-cell whole genome and transcriptome libraries within a week. We profiled phenotypically normal esophagus tissue from four aged donors and used somatic variants to build high-resolution single-cell lineages from >2,700 cells with accompanying transcriptomic information, reconstructing >70 years of somatic evolution. T cell expansions identified from T cell receptor (TCR) sequences validated the clonal structure of the single-nucleotide variant (SNV)-based phylogenies and phylogenetic cross-correlation analysis showed that epithelial cells had higher degrees of shared ancestry by spatial location compared to immune cells. Mapping mutation signatures to the phylogenetic tree revealed the emergence of tobacco/alcohol exposure-related signatures later in life, consistent with the donors exposure histories. We identified variants in driver genes that were previously reported in the phenotypically normal esophagus, detecting clonal expansions harboring mutations in genes including TP53 and FAT1. We mapped the evolution of clones with both monoallelic as well as biallelic TP53 loss, including a clone associated with high expression of cell cycling genes and higher chromosome instability. Leveraging the matched transcriptome data, we uncovered cell type biases in mutant clones, with a higher proportion of TP53 or FAT1-mutant cells in an earlier basal epithelial cell state compared to wild-type cells. We further observed copy-neutral loss of heterozygosity (CNLOH) events on chromosome 9q that spanned the NOTCH1 locus in up to [~]35% of epithelial cells. Mapping CNLOH events to the phylogenetic tree revealed a striking pattern in which CNLOH was separately acquired many times, reflecting convergent evolution. Cells with CNLOH events were biased towards the earlier basal epithelial state, suggestive of a selective advantage that leads to prevalent recurrence of chr9q CNLOH. Together, we demonstrate that SMART-PTA is an efficient, scalable approach for single-cell whole-genome and whole-transcriptome profiling to build phenotypically annotated single-cell phylogenies with enough throughput and power for application to normal tissue somatic evolution. Moreover, we reconstruct the evolutionary history of the esophageal epithelium at high scale and resolution, providing a window into the dynamics and processes that shape clonal expansions in phenotypically normal tissues throughout a lifespan.

genomics↗

bioRxiv2025

Inter-tool analysis of a NIST dataset for assessing baseline nucleic acid sequence screening

Nucleic acid synthesis is a dual-use technology that can benefit fields such as biology, medicine, and information storage. However, synthetic nucleic acids could also potentially be used negligently and ultimately cause harm, or be used with malicious intent to cause harm. Thus, this technology needs to be appropriately safeguarded. Sequence screening is one component of a biosecurity protocol for preventing such harm and consists of differentiating Sequences of Concern (SOCs) from benign sequences that are not associated with pathogenicity or toxicity. There exist many fit-for-purpose tools that have been developed for DNA synthesis sequence screening. However, questions remain regarding their performance with respect to consistency of screening. To aid in determining if screening tools are harmonized in regard to baseline sequence screening, NIST constructed a test dataset based on current screening recommendations. NIST then sent blinded datasets to sequence screening tool developers for testing. Overall, there was a general agreement between the tools and NIST assignments of the sequences and all tools had a baseline performance of greater than 95% sensitivity and 97% accuracy. Disagreement on specific sequences largely arose from single tools and could be traced to differences in defining a SOC and/or methodological differences in screening algorithms.

bioinformatics↗

bioRxiv2024

Antibiotic lethality dictates mycobacterial infection outcomes

Antibiotic development and treatment focus on bacterial growth inhibition, often with limited success. Here, we introduce Antimicrobial Single-Cell Testing (ASCT), an advanced imaging strategy to assess bacterial killing in real-time. By tracking 140 million bacteria and generating over 20,000 in vitro time-kill curves, we can predict Mycobacterium tuberculosis treatment outcomes in mice and humans and link strain-specific survival (drug tolerance) in Mycobacterium abscessus to clinical responses. Using ASCT, we reveal drug tolerance as a distinct genetically encoded bacterial trait conserved across drugs with similar targets and, via genome-wide associations, uncover molecular mechanisms that govern bacterial killing. This study establishes the technical framework and in vivo validation for large-scale bacterial killing assessments to advance our understanding of bacterial survival, antibiotic development and clinical decision-making.

microbiology↗

bioRxiv2022

Data Descriptor: Human whole exome genotype data for Alzheimer's Disease

Bigger sample size can help to identify new genetic variants contributing to an increased risk of developing Alzheimers disease. However, the heterogeneity of the whole-exome sequencing (WES) data generation methods presents a challenge to a joint analysis. Here we present a bioinformatics strategy for joint calling 20,504 WES samples collected across nine studies and sequenced using ten different capture kits in fourteen sequencing centers in the Alzheimers Disease Sequencing Project. gVCFs of samples were joint-called by the Genome Center for Alzheimers Disease into a single VCF, containing only positions within the union of capture kits. The VCF was then processed using specific strategies to account for the batch effects arising from the use of different capture kits from different studies. We identified 8.2 million autosomal variants. 96.82% of the variants are high-quality, and are located in 28,579 Ensembl transcripts. 41% of the variants are intronic and 15% are missense variants. 1.8% of the variants are with CADD>30. Our new strategy for processing these diversely generated WES samples has shown to generate high-quality data. The improved ability to combine data sequenced in different batches benefits the whole genomics research community. The WES data are accessible to the scientific community via https://dss.niagads.org/.

genetics↗

bioRxiv2019

Evolution of Salmonella enterica serotype Typhimurium driven by anthropogenic selection and niche adaptation

Salmonella enterica serotype Typhimurium (S. Typhimurium) is a leading cause of gastroenteritis and disseminated disease worldwide. Two S. Typhimurium strains (SL1344 and ATCC14028) are widely used to study host-pathogen interactions, yet genotypic variation results in strains with diverse host range, pathogenicity and risk to food safety. A robust fully parsimonious phylogenetic tree constructed from recombination purged variation in the whole genome sequence of 131 diverse strains of S. Typhimurium revealed population structure composed of two high order clades ( and {beta}) and multiple subclades on extended internal branches, that exhibited distinct signatures of host adaptation and anthropogenic selection. Clade contained a number of subclades composed of strains from well characterized epidemics in domesticated animals, while clade {beta} predominantly contained subclades associated with wild avian species, with the notable exception of a subclade containing the DT204/49 complex. The contrasting epidemiology of and {beta} strains was reflected in a distinct distribution of antimicrobial resistance (AMR) genes, accumulation of hypothetically disrupted coding sequences (HDCS), and signatures of functional diversification associated with invasiveness of host adapted serotypes. Gene flux was predominantly driven by acquisition, loss or recombination of prophage. The acquisition of large genetic islands (SGI-1 and 4) was limited to two recent pandemic clones (DT104 and monophasic S. Typhimurium ST34) in clade . Together, our data are consistent with the view that a broad host range common ancestor of S. Typhimurium diversified with clade lineages remained largely associated with multiple domesticated animal species, while clade {beta} spawned multiple lineages that underwent diversifying selection associated with adaptation to various niches, predominantly in wild avian species.

genomics↗

bioRxiv2019

Evaluation of parameters affecting performance and reliability of machine learning-based antibiotic susceptibility testing from whole genome sequencing data

Prediction of antibiotic resistance phenotypes from whole genome sequencing data by machine learning methods has been proposed as a promising platform for the development of sequence-based diagnostics. However, there has been no systematic evaluation of factors that may influence performance of such models, how they might apply to and vary across clinical populations, and what the implications might be in the clinical setting. Here, we performed a meta-analysis of seven large Neisseria gonorrhoeae datasets, as well as Klebsiella pneumoniae and Acinetobacter baumannii datasets, with whole genome sequence data and antibiotic susceptibility phenotypes using set covering machine classification, random forest classification, and random forest regression models to predict resistance phenotypes from genotype. We demonstrate how model performance varies by drug, dataset, resistance metric, and species, reflecting the complexities of generating clinically relevant conclusions from machine learning-derived models. Our findings underscore the importance of incorporating relevant biological and epidemiological knowledge into model design and assessment and suggest that doing so can inform tailored modeling for individual drugs, pathogens, and clinical populations. We further suggest that continued comprehensive sampling and incorporation of up-to-date whole genome sequence data, resistance phenotypes, and treatment outcome data into model training will be crucial to the clinical utility and sustainability of machine learning-based molecular diagnostics.\n\nAuthor SummaryMachine learning-based prediction of antibiotic resistance from bacterial genome sequences represents a promising tool to rapidly determine the antibiotic susceptibility profile of clinical isolates and reduce the morbidity and mortality resulting from inappropriate and ineffective treatment. However, while there has been much focus on demonstrating the diagnostic potential of these modeling approaches, there has been little assessment of potential caveats and prerequisites associated with implementing predictive models of drug resistance in the clinical setting. Our results highlight significant biological and technical challenges facing the application of machine learning-based prediction of antibiotic resistance as a diagnostic tool. By outlining specific factors affecting model performance, our findings provide a framework for future work on modeling drug resistance and underscore the necessity of continued comprehensive sampling and reporting of treatment outcome data for building reliable and sustainable diagnostics.

microbiology↗