bioRxiv Science⌕ Search

Biology subjects

Lees, J.

Publications and source records attributed to Lees, J..

At least 19 recordsLinked to original sources
bioRxiv2026

Autonomic neurons are intrinsically hyperexcitable in a Friedreich ataxia iPSC model

Friedreich ataxia (FRDA) is an inherited multisystem disorder caused by a GAA repeat expansion in the FXN gene and consequent frataxin deficiency, causing progressive neurodegeneration and heart disease. Clinical studies have documented autonomic dysfunction in FRDA, including elevated resting heart rate, raised plasma catecholamines and a high burden of urinary, bowel and sexual symptoms. Consistent with this, autopsy studies show degeneration of the cardiac conduction system and its associated ganglia in FRDA. Whether these features are intrinsic to FRDA autonomic neurons has been untestable without a human cellular model. Here, a human model of FRDA autonomic neurons was established by differentiating induced pluripotent stem cells (iPSCs) from individuals with FRDA and from their CRISPR/Cas9-corrected isogenic controls. FRDA autonomic neurons expressed canonical autonomic lineage markers and retained reduced FXN mRNA and frataxin protein levels. Electrophysiological analysis revealed pronounced hyperexcitability, with an approximately five-fold increase in spike number, shortened interspike intervals, and longer bursts. FRDA autonomic neurons also exhibited increased cell death, elevated mitochondrial reactive oxygen species, and a shift from networked to tubular mitochondrial morphology. Bulk RNA sequencing of all three isogenic pairs revealed loss of extracellular matrix programs and activation of stress response pathways. The top differentially expressed gene was MEG3, a maternally imprinted long noncoding RNA, which was consistently upregulated across all three isogenic pairs. MEG3 sits within the imprinted 14q32 DLK1-DIO3 locus, whose maternally expressed noncoding transcripts were co-ordinately de-repressed in FRDA autonomic neurons. Competitive gene-set testing of 43 imprinted genes across 20 domains showed this to be specific to 14q32: no other imprinted domain was co-ordinately altered. Together, these findings establish hyperexcitability of autonomic neurons as a cell autonomous consequence of frataxin deficiency, accompanied by mitochondrial dysfunction. This provides a cellular basis for the increased sympathetic tone observed in clinical studies.

neuroscience↗
bioRxiv2026

Elucidating the Design Space of Generative Models for Single-Cell Perturbation Prediction

Next-token prediction has produced predictable scaling in language, but the recipe presumes a sequence of tokens with a meaningful order. Single-cell RNA-seq counts have no natural gene ordering, so applying the recipe directly to raw expression fails under an ill-suited left-to-right bias. We instead ask whether a learned latent can supply the structure the recipe needs. We introduce ExpressionVAE (eVAE), a discrete-latent perturbation model that compresses each cell into a short sequence of discrete codes through a finite-scalar-quantization (FSQ) bottleneck and trains a perturbation-conditioned discrete prior over those codes. On Replogle and Parse 1M, eVAE sets a new state of the art on every distributional metric and leads on most cell-eval perturbation metrics, with Frechet distance and MMD2 roughly 3 to 20x lower than the strongest continuous-latent baseline. Swapping the prior between autoregressive and masked discrete diffusion leaves performance near-identical, isolating the gain to the discrete latent itself rather than the prior family. A decoder-head ablation then exposes a single design axis, the richness of the predictive distribution at inference, that splits the standard metrics into two groups, variance-sensitive and mean-sensitive, which move in opposite directions along the axis. Finally, on a held-out CRISPRi reversion benchmark of 1,732 perturbations under inflammatory cytokine stress, the frozen eVAE encoder outperforms UMAP and differential expression and matches scGPT on perturbation ranking at a fraction of the data.1 We release our code.2

bioinformatics↗
bioRxiv2026

An integrated cardiac microtissue proteome map extends therapeutic remodelling by nanovesicles

Human cardiac microtissues are a promising model to study cardiac biology and disease, but their application is constrained by therapeutic remodelling strategies and limited knowledge of their functional protein expression profiles. Here, we define the use of human cardiac microtissue (hCMT) model generated by assembling iPSC-derived endothelial cells, cardiac fibroblasts, and cardiomyocytes to model ischemia-reperfusion injury (IRI) through a model of hypoxia and reoxygenation and nanovesicle-mediated functional remodelling. Engineered nanovesicles (NVs), generated directly from human stem cells, have been shown to influence cardiac tissue and cell repair, and provide a platform for scalable and reproducible cell free-mediated therapy. We show the functional regulation of the hCMT model and define that administration of NVs (from human induced pluripotent stem cell origin) during reoxygenation significantly increase cardiomyocyte survival and preserve contractility function (contractile duration, relaxation time, relaxation:contraction velocity). Quantitative proteomics was applied to decipher the cell proteome dynamics and molecular mechanisms of IRI in our in vitro model following NV treatment, linked with networks associated with cell survival, energy production, and stress response regulation. Conserved proteome dynamics in NVs from different iPSC source reveal conserved upregulation of cellular protein networks involved in tissue repair (HSP70, CYFIP1), cardiac function (XIRP1, SLMAP, MYH6, CTNNA1, NDUFS2, GPD2), response to stress (CANX, PDCD6,), pro-survival (MDH2, LRPPRC, NIPSNAP1) and pro-angiogenic (FARSA, ECE1, RRAS) relative to vehicle treatments in context of IRI. Finally, we show that NVs also mediate differential remodelling in hCMT in response to IRI based on their cell origin, including altered wound healing and tissue repair response. Our findings provide an advanced human stem cell-based platform to understand underlying mechanisms of IRI and assess cell-free therapeutic cardioprotective strategies. SummaryAdvanced human stem cell-based platform provides a cardiac microtissue model to understand nanovesicle-based function and proteome remodelling, with potential applications for disease modelling and therapeutic intervention.

cell biology↗
bioRxiv2026

Global population structure and phase variation of serotype 12F Streptococcus pneumoniae following the introduction of pneumococcal conjugate vaccine

BackgroundAfter the global deployment of pneumococcal conjugate vaccines (PCVs), serotype 12F has become the predominant serotype responsible for invasive pneumococcal disease (IPD) worldwide. As PCVs that include serotype 12F are gradually introduced, we aim to characterise the global population structure and genetic diversity of the 12F capsule locus using whole-genome sequencing. Capsule variants with vaccine evasion potential were further investigated by functional experiments. MethodsA global collection of pneumococcal serotype 12F genomes (n=806) from 37 countries across six continents were included in this study. To characterise the serotype 12F population, Global Pneumococcal Sequence Cluster (GPSC), in silico serotype, and antimicrobial resistance profile were inferred from whole-genome data for each isolate. The capsule biosynthesis (cps) locus was analysed for gene content variations that could alter polysaccharide capsule production or structure, thereby influencing recognition by vaccine-induced antibodies. These isolates were further investigated by assessing their capsule production using immunofluorescence assays and its susceptibility to vaccine-elicited antibody killing by opsonophagocytosis assays. FindingsThe global increase in serotype 12F was driven by both distinct pneumococcal lineages across different continents, and a globally-disseminated and multidrug-resistant lineage GPSC26. We identified six capsule variants in nine isolates that had disruptive mutations in cps genes including wze, wcil, wciJ and fnlA. Most (6/9) of the disruptive mutations were a result of strand-slippage mutations. A convergent strand-slippage mutation disrupting the glycosyltransferase gene wciJ was identified in four isolates from distinct lineages and countries. Despite the truncation, three of four isolates with available Quellung typing results still identified them as 12F, indicating the production of the capsule. We then created a genetically engineered lab strain with wciJ knockout and complemented with wciJ containing the strand-slipppage mutation. The knockout strain did not produce any capsule. In contrast, the lab strain with wciJ containing the strand-slippage mutation produced a mixed population of encapsulated and non-encapsulated pneumococci, even within the same chain of pneumococcal cells. This observation indicated encapsulated subpopulation possesses a functional WciJ and rapidly reversible strand-slippage mutation during replication. Opsonophagocytosis assays indicated that the clinical 12F strain with strand-slippage mutation in wciJ exhibited reduced susceptibility to vaccine-elicited serum killing, compared to a genetically closely related 12F clinical strain with an intact wciJ. However, substantial inter-individual antisera variation limits definitive interpretation. InterpretationOur work revealed the global rise of serotype 12F pneumococci has been driven by both regional-specific lineages, and a globally-disseminated and multidrug-resistant lineage GPSC26. We demonstrated that strand-slippage mutation is one of the major drivers of serotype 12F capsule variants and represents a novel mechanism enabling reversible on-off switching of capsule production. The ability to switch off capsule expression in a subpopulation may enable evasion of antibody-mediated killing but increase susceptibility to innate immune clearance. FundingBill & Melinda Gates Foundation, Wellcome Sanger Institute, and the US Centers for Disease Control and Prevention.

genomics↗
bioRxiv2026

Discrete Diffusion for Single-Cell Gene Expression Modeling

AO_SCPLOWBSTRACTC_SCPLOWCurrent generative modeling of single-cell transcriptomics relies on continuous latent representations, transforming inherently discrete and sparse gene counts into continuous space. We propose Discrete Cell Models (DCM), a diffusion-based framework that learns cellular representations directly in the discrete domain. Our framework supports both unconditional and conditional generation, allowing for precise modeling of complex biological scenarios such as cell-type-specific transcriptional responses to genetic perturbations. We demonstrate that DCM scales effectively and achieves strong performance against current state-of-the-art methods, including scVI, CPA, STATE, scGPT, and scLDM. On the Dentate Gyrus benchmark, DCM achieves a 5-fold improvement in MMD2RBF and a nearly 2-fold improvement in W2 distance, over the leading continuous diffusion baseline (scLDM). On the conditional Replogle perturbation benchmark, DCM sets a new state of the art on W2 distance while remaining competitive on MMD2RBF. Together, these results establish discrete diffusion as a promising direction for foundational models of cellular biology.

cell biology↗
bioRxiv2026

DRP1 inhibition confers cardioprotection against doxorubicin while preserving anticancer efficacy

BackgroundAnthracyclines such as doxorubicin are effective chemotherapeutics but are limited by cardiotoxicity driven in part by mitochondrial dysfunction. Dysregulated mitochondrial dynamics, particularly excessive dynamin-related protein-1 (Drp1)-mediated fission, contribute to doxorubicin-induced cardiac injury and support selective survival of cancer cells. ObjectivesTo determine whether DRP1i2, a novel small molecule Drp1 inhibitor targeting a conserved domain shared between human and mouse, can function as a cardio-oncology therapeutic by reducing doxorubicin-induced cardiotoxicity while maintaining or enhancing anti-cancer efficacy. MethodsCardioprotective effects of DRP1i2 were evaluated in a murine model of chronic doxorubicin cardiotoxicity and in human induced pluripotent stem cell-derived cardiac microtissues exposed to acute doxorubicin injury. Anticancer activity was assessed across multiple cancer cell lines using 2D monolayers and 3D microtissues. ResultsIn vivo, DRP1i2 preserved left ventricular ejection fraction, reduced interstitial fibrosis and cardiomyocyte atrophy, and attenuated doxorubicin-induced myocardial proteomic remodelling. In human cardiac microtissues, DRP1i2 improved viability and restored contractile function despite persistent mitochondrial oxidative stress. DRP1i2 showed modest anticancer activity in MG63 osteosarcoma cells in both 2D and 3D systems and did not diminish doxorubicin efficacy in other cancer models (MDA-MB-231 breast, OVCAR3 ovarian, and A549 lung adenocarcinoma). Combined treatment further enhanced cytotoxicity selectively in MG63 cells. ConclusionsDRP1i2 exerts complementary cardioprotective and anticancer actions through modulation of shared mitochondrial pathways, identifying Drp1 as a druggable target in cardio-oncology. These findings support DRP1i2 as a first-in-class Drp1 inhibitor and highlight mitochondrial dynamics as a promising therapeutic axis to preserve anthracycline efficacy while reducing cardiotoxicity. Clinical PerspectivesExcessive Drp1-mediated mitochondrial fission links anthracycline cardiotoxicity with cancer cell survival. Inhibition with DRP1i2 preserved cardiac structure and function in a chronic doxorubicin cardiotoxicity model without compromising anti-cancer activity, representing mechanism-based cardioprotection, where the heart is protected by directly targeting the molecular processes driving injury. Translation will require pharmacologic profiling and testing in tumour-bearing and comorbid models, followed by early-phase trials to confirm safety and efficacy.

pharmacology and toxicology↗
bioRxiv2026

A 28-year evolution experiment on Burkholderia pseudomallei survival in nutrient-depleted sterile water

Environmental persistence allows opportunistic pathogens to survive a range of harsh conditions, increasing the likelihood of eventual infection. Burkholderia pseudomallei, the causative agent of melioidosis, can endure long-term nutrient-depletion in the environment, but its adaptive mechanisms remain poorly understood. Here, we investigated the evolutionary trajectory of a clinical B. pseudomallei strain maintained in sterile water since 1994. The strain was inoculated into nine individual tubes at nine initial concentrations (102-1010 CFU/mL) and has remained viable to date. Liquid chromatography-mass spectrometry analysis of the water identified potential carbon sources, including phthalic acid -- a plastic degradation product likely leached from inoculation tubes -- which the strain can metabolise via an intact catabolic operon. Genomic variations accumulated between 1994 and 2022 were characterised using both single-colony and plate-sweep sequencing which provided complementary insights. Across all tubes, we identified 249 single-nucleotide polymorphisms (SNPs), 73 indels, and a large-scale deletion. Cultures from each tube displayed a consistently low mutation rate (3.18 x 10- SNPs per site per year), suggesting that cells entered a dormant or slow-growth state. Of 393 genes with mutations, 193 were independently mutated in more than one tube, particularly those involved in signal transduction, cell wall and membrane biogenesis, and secondary metabolite synthesis. These patterns indicate parallel adaptation to long-term nutrient deprivation through modulation of cell-density-related functions and loss of metabolically costly pathways. Remarkably, B. pseudomallei from this experiment remain viable after nearly three decades, providing a rare natural model for understanding how environmental bacteria endure and adapt in extremely nutrient-depleted conditions.

evolutionary biology↗
bioRxiv2026

PRMT5 inhibitors actively promote metastatic progression of lung adenocarcinoma

Epigenetic changes are a major driver of cancer progression, placing considerable focus on epigenetic regulators as therapeutic targets. Protein arginine methyltransferase 5 (PRMT5) is one such regulator, and numerous PRMT5 inhibitors (PRMT5i) in clinical trials. Despite this, the mechanisms and consequences of PRMT5i-resistance are unknown. Here, we demonstrate that aggressive cancer progression is an inbuilt feature of PRMT5i-resistance acquisition in lung adenocarcinoma (LUAD). Independently-generated resistant cell lines gain dedifferentiation signatures that typify late-stage disease and show increased metastatic potential in vivo. We establish that these state shifts are a direct consequence of PRMT5i action; treatment induces rapid and widespread chromatin rewiring, enabling derepression of late-stage disease states that are stably established in resistant cells. Notably, treatment of lung tumor-bearing mice drives rapid disease advancement without decreasing tumor burden, showing that drug-induced disease progression supersedes any benefits from PRMT5 inhibition in vivo. Furthermore, analyses of human cell lines and patient cohorts supports the notion of PRMT5 inhibition-mediated dedifferentiation. Collectively our data show that PRMT5i can actively promote self-resistance and disease progression in different tumor types. This raises serious concerns for the use of PRMT5i in patients, arguing that clinical studies should consider the possibility of drug-induced plasticity, resistance, and disease advancement.

cancer biology↗
bioRxiv2025

A comprehensive AMR genotype-phenotype database (CABBAGE)

Addressing the growing threat of antimicrobial resistance (AMR) requires the development of large-scale resources that link bacterial genomic data with phenotypic antimicrobial resistance profiles. Such datasets are essential for advancing genotype-based predictions of resistance to uncover novel resistance mechanisms, as well as identifying and tracking global trends. Here, we describe the development of the Comprehensive Assessment of Bacterial-Based AMR prediction from GEnotypes (CABBAGE) database, linking bacterial genomes to associated antibiotic susceptibility data and relevant metadata across WHO Bacterial Priority Pathogens, sourced from both publications and existing databases, and curated into a format that is compatible with, and extends, both NCBI and ENA formats. The resulting CABBAGE database, comprising over 170,000 unique sequenced isolates and approximately 1.7 million genome-phenotype pairs linked to extensive metadata, represents the largest database of its kind, consolidating existing AMR phenotype-genotype data into a single unified format. CABBAGE encompasses a broad range of antimicrobials, facilitating the analysis of global resistance trends as well as benchmarks of genotype-to-phenotype predictive methods, and empowering further research uses. The database is freely accessible at https://www.ebi.ac.uk/amr and is currently being integrated with the BioSample database, enabling easy access for the AMR research community.

genomics↗
bioRxiv2025

The origin and molecular evolution of the mammalian liver cell architecture

The liver is a central organ with essential roles in processes such as nutritional metabolism, detoxification, and immune defense1-6. It has been instrumental in the adaptation of mammalian species to diverse environments, as reflected by its rapid molecular evolution7,8. However, the origins and evolutionary dynamics of liver cell types and their structural organization remain largely unexplored. Here we report evolutionary analyses of transcriptome and chromatin accessibility data for liver cells from 17 species, spanning all major feeding strategies (herbivory, omnivory, carnivory, insectivory), great apes (including humans), placental clades (Afrotheria, Xenarthra, Laurasiatheria, Euarchontoglires), major mammalian lineages (placentals, marsupials, monotremes), and a bird as outgroup. Integrated with spatial transcriptomics, our data reveal that liver zonation--the compartmentalization of hepatocyte functions along the lobule, the livers fundamental anatomical and functional unit--is conserved across mammals but absent in other vertebrates. We find that zonation originated in the mammalian ancestor, driven by the emergence of WNT and R-spondin signaling from central vein endothelial cells, which activate central hepatocyte gene expression via the transcription factor TCF7L2. Despite this conserved architecture and signaling, genes with zonated expression exhibit rapid evolutionary turnover. Consistently, hepatocytes evolve fast, likely due to reduced selective constraints, enabling adaptive changes under positive selection. Alongside immune cells, hepatocytes are therefore key drivers of the livers rapid evolution and functional innovations. In great apes, we identify human-specific shifts in zonation and cell-type-specific expression linked to recent cis-regulatory changes, particularly in genes involved in lipid metabolism, likely contributing to human-specific metabolic traits. Our study uncovers the origins of a mammal-specific liver cell architecture, within which reduced constraints facilitated molecular changes underlying ecological adaptations.

evolutionary biology↗
bioRxiv2025

Deacon: fast sequence filtering and contaminant depletion

MotivationRealising the value of large DNA sequence collections demands efficient search and extraction of sequences of interest. Search queries may vary in size from short gene sequences to multiple whole genomes that are too large to fit in computer memory. In microbial genomics, a routine search application involving both large queries and large collections is the removal of contaminating host genome sequences from microbial (meta)genomes. Where the host is human, sensitive classification and excision of host sequences is usually necessary to protect host genetic information. Precise classification is also critical in order to retain microbial sequences and permit accurate microbial genomic analysis. While human pangenomes have been shown to increase sensitivity of human sequence classification, existing bioinformatic host depletion approaches have either limited precision when used with metagenomes or large computing resource requirements. ResultsWe present Deacon, an efficient and versatile sequence filter for raw sequence files and streams. We demonstrate its leading accuracy for the task of host depletion, using less computing resource than existing approaches. By querying a human pangenome index for minimizers contained in each input sequence, Deacon is able to accurately classify and discard diverse human sequences from long reads at over 250Mbp/s with a commodity laptop. We present validation of classification sensitivity, specificity and speed with simulated short and long reads for diverse catalogues of human, bacterial and viral genomes alongside existing methods. Beyond host depletion, Deacon is well suited to common sequence search and filtering applications, particularly those involving large queries. Capable of indexing a human genome in under 30s, Deacon is equipped to rapidly compose custom minimizer indexes using set operations, facilitating efficient search and filtering of massive sequence datasets using gigabase queries. Availability and implementationDeacon is implemented as an MIT-licensed command line tool written in Rust and packaged with Bioconda. Code is available from https://github.com/bede/deacon.

bioinformatics↗
bioRxiv2025

Device encapsulated MSCs for adaptive secretome therapy to effectively target ischaemic heart injury

Effective long-term strategies to protect the ischaemic heart remain a significant challenge. Mesenchymal stem cells (MSCs) offer therapeutic potential primarily through their secretome, a bioactive factor-rich milieu with broad beneficial effects. However, existing delivery methods have not been shown to provide sustained benefits. Herein, we introduce an innovative approach for sustained MSC-secretome delivery for long-term cardioprotection. In a rat model of chronic myocardial infarction, Cymerus MSCs, derived from human induced pluripotent stem cells (iPSCs), were encapsulated in a Procyon immunoisolation device and implanted subcutaneously. A human-iPSC-derived engineered cardiac microtissue model was used to simulate ischaemia-reperfusion injury and assess cardioprotective effects in a human context. The MSC-loaded Procyon device significantly improved cardiac function and reduced adverse left ventricular remodelling over a 12-week period. These positive effects were consistent across both young and middle-aged, male and female rats, indicating broad applicability. The encapsulated MSCs remained viable and continuously released therapeutic secretome for 12 weeks. In vitro, the MSC secretome protected human engineered cardiac microtissues from simulated ischaemia-reperfusion injury, restoring contractile function, improving cell viability, and reducing oxidative stress. Proteomic analysis of MSCs revealed 179 unique cellular proteins post-implantation, linked to adaptive immune and inflammatory responses as well as wound healing. MSC secretome profiling revealed increased protein diversity associated with tissue repair and immune regulation, suggesting MSCs undergo an adaptive response to ischaemic conditions, enhancing their therapeutic potential. This translational study highlights a clinically viable, minimally invasive method for sustained cardioprotection, harnessing the MSC secretome to address a pivotal gap in current treatments for ischaemic heart disease.

pathology↗
bioRxiv2025

Selection-free CRISPR-Cas9 editing protocol for distant Dictyostelid species

Dictyostelids are a species-rich clade of cellular slime molds that are widely found in soils and have been studied for over a century. Most research focusses on Dictyostelium discoideum, which - due to its ease of culturing and genetic tractability - has been adopted as a model species in the fields of developmental biology, cell biology and microbiology. Over decades, genome editing methods in D. discoideum have steadily improved but remain relatively time-consuming and limited in scope, effective in a few species only. Here, we introduce a CRISPR-Cas9 editing protocol that is cloning-free, selection-free, highly-efficient, and effective across Dictyostelid species. After optimizing our protocol in D. discoideum, we obtained knock-out efficiencies of [~]80% and knock-in efficiencies of [~]30% without antibiotic selection. Efficiencies depend on template concentrations, insertion sizes, homology arms and target sites. Since our protocol is selection-free, we can isolate mutants as soon as one day post-transfection, vastly expediting the generation of knock-outs, fusion proteins and expression reporters. Our protocol also makes it possible to generate several knock-in mutations simultaneously in the same cells. Boosted by cell-sorting and fluorescent microscopy, we could readily apply our CRISPR-Cas9 editing protocol to phylogenetically distant Dictyostelid species, which diverged hundreds of millions of years ago and have never been genome edited before. Our protocol therefore opens the door to performing broad-scale genetic interrogations across Dicyostelids.

genetics↗
bioRxiv2024

Primary cilia promote EMT-induced triple-negative breast tumor heterogeneity and resistance to therapy

Tumor heterogeneity and plasticity, driven by Epithelial-Mesenchymal Transition (EMT), enable cancer therapeutic resistance. We previously showed that EMT promotes primary cilia formation, which enables stemness and tumorigenesis in triple-negative breast cancer (TNBC). Here, we establish a role for primary cilia in human TNBC chemotherapeutic resistance. We developed patient-derived organoids, and showed that these recapitulated the cellular heterogeneity of TNBC biopsies. Notably, one of the identified cell states bore a quasi-mesenchymal phenotype, primary cilia, and stemness signatures. We treated our TNBC organoids with chemotherapeutics and observed partial killing. The surviving cells with organoid-reconstituting capacity showed selective enrichment for the quasi-mesenchymal ciliated cell subpopulation. Genomic analyses argue that this enrichment reflects a combination of pre-existing cells and ones that arose through drug-induced cellular plasticity. We developed a family of small-molecule inhibitors of ciliogenesis and show that these, or genetic ablation of primary cilia, suppress chemoresistance. We conclude that primary cilia help TNBC to evade chemotherapy. SignificanceCancer cells that activate EMT to acquire a quasi-mesenchymal state form primary cilia to evade chemotherapy in human triple-negative breast cancer. Pharmacological inhibition of primary ciliogenesis counteracts EMT-induced chemoresistance.

cancer biology↗
bioRxiv2024

CELEBRIMBOR: Pangenomes from metagenomes

SummaryMetagenome Assembled Genomes (MAGs) are often incomplete, with sequences missing due to errors in assembly or low coverage. Incomplete MAGs present a particular challenge for identification of shared genes within a microbial population, known as core genes, as a core gene missing in only a few assemblies will result in it being mischaracterized at a lower frequency. Here, we present CELEBRIMBOR, a snakemake pangenome analysis pipeline which uses a measure of genome completeness to automatically adjust the frequency threshold at which core genes are identified, enabling accurate core gene identification in MAGs. Availability and implementationCELEBRIMBOR is published under open source Apache 2.0 licence at https://github.com/bacpop/CELEBRIMBOR and is available as a Docker container. Supplementary material is available in the online version of the article.

bioinformatics↗
bioRxiv2024

AllTheBacteria - all bacterial genomes assembled, available and searchable

Public microbial genomes encode an immense record of biological diversity, evolution and molecular function, but much of this information remains difficult to reuse because raw sequencing data are not uniformly assembled, quality controlled, annotated or searchable at scale. Here we present AllTheBacteria, an open, community-built resource that transforms public bacterial short-read whole-genome sequencing reads into a uniformly processed discovery platform. The current analysed release contains 2,440,377 high-quality bacterial and archaeal genomes from 11,273 species, together with standardized taxonomic assignments, genome annotations, antimicrobial resistance calls, antiphage-defence annotations, protein structure predictions and AI-ready sequence tables. We show that this infrastructure enables applications that would otherwise be impractical, from global sequence search and outbreak contextualization to pangenome method development, antimicrobial resistance reservoir mapping and antiphage-defence ecology. As a stringent experimental demonstration, we mined 3,919,096 encrypted peptide fragments from AllTheBacteria proteomes using our deep learning model APEX 1.1, identifying 1,867 candidates with predicted antimicrobial activity. We synthesized 24 representative peptides and tested them against 20 clinically relevant bacterial strains, including antibiotic-resistant pathogens. Multiple peptides showed low-micromolar activity, membrane-responsive conformational transitions and selective envelope perturbation. A lead molecule, ATB20, reduced Acinetobacter baumannii burden in a murine skin abscess model with efficacy comparable to polymyxin B and no overt toxicity. Together, these results establish AllTheBacteria as both a foundational community resource for microbiology and a renewable engine for AI-guided antimicrobial discovery.

bioinformatics↗
bioRxiv2023

Optimising machine learning prediction of minimum inhibitory concentrations in Klebsiella pneumoniae

Minimum Inhibitory Concentrations (MICs) are the gold standard for quantitatively measuring antibiotic resistance. However, lab-based MIC determination can be time-consuming and suffers from low reproducibility, and interpretation as sensitive or resistant relies on guidelines which change over time. Genome sequencing and machine learning promise to allow in-silico MIC prediction as an alternative approach which overcomes some of these difficulties, albeit the interpretation of MIC is still needed. Nevertheless, precisely how we should handle MIC data when dealing with predictive models remains unclear, since they are measured semi-quantitatively, with varying resolution, and are typically also left- and right-censored within varying ranges. We therefore investigated genome-based prediction of MICs in the pathogen Klebsiella pneumoniae using 4367 genomes with both simulated semi-quantitative traits and real MICs. As we were focused on clinical interpretation, we used interpretable rather than black-box machine learning models, namely, Elastic Net, Random Forests, and linear mixed models. Simulated traits were generated accounting for oligogenic, polygenic, and homoplastic genetic effects with different levels of heritability. Then we assessed how model prediction accuracy was affected when MICs were framed as regression and classification. Our results showed that treating the MICs differently depending on the number of concentration levels of antibiotic available was the most promising learning strategy. Specifically, to optimise both prediction accuracy and inference of the correct causal variants, we recommend considering the MICs as continuous and framing the learning problem as a regression when the number of observed antibiotic concentration levels is large, whereas with a smaller number of concentration levels they should be treated as a categorical variable and the learning problem should be framed as a classification. Our findings also underline how predictive models can be improved when prior biological knowledge is taken into account, due to the varying genetic architecture of each antibiotic resistance trait. Finally, we emphasise that incrementing the population database is pivotal for the future clinical implementation of these models to support routine machine-learning based diagnostics. Data SummaryThe scripts used to run and fit the models can be found at https://github.com/gbatbiff/Kpneu_MIC_prediction. The Illumina sequences from Thorpe et al. are available from the European Nucleotide Archive under accession PRJEB27342. All the other genomes are available on https://www.bv-brc.org/ database. Impact statementKlebsiella pneumoniae is a leading cause of hospital and community acquired infections worldwide, highly contributing to the global burden of antimicrobial resistance (AMR). Ordinary methods to assess antibiotic resistance are not always satisfactory, and may not be effective in terms of costs and delays, so robust methods able to accurately predict AMR are increasingly needed. Genome-based prediction of minimum inhibitory concentrations (MICs) through machine learning methods is a promising tool to assist clinical diagnosis, also offsetting phenotypic MIC discordance between the different culture-based assays. However, benchmarking predictive models against phenotypic data is problematic due to inconsistencies in the way these data are generated and how they should be handled remains unclear. In this work, we focused on genome-based prediction of MIC and evaluated the performance of interpretable machine learning models across different genetic architectures and data encodings. Our workflow highlighted how MICs need to be treated as different types of data depending on the method used to measure them, in particular considering each antibiotic separately. Our findings shed further light on the factors affecting model performance, paving the way to future improvements of antibiotic resistance prediction.

bioinformatics↗
bioRxiv2023

fastlin: an ultra-fast program for Mycobacterium tuberculosis complex lineage typing

Lineage typing of the Mycobacterium tuberculosis complex (MTBC) has evolved from traditional phenotypic methods to advanced molecular and genomic techniques. In this study we present fastlin, a bioinformatics tool designed for rapid MTBC lineage typing. Fastlin utilises an ultra-fast alignment-free approach to detect previously identified barcode single nucleotide polymorphisms (SNPs) associated with specific MTBC lineages directly from fastq files. In a comprehensive benchmarking against existing tools, fastlin demonstrated high accuracy and significantly faster running times. Analysis of large MTBC datasets revealed fastlins capability not only to predict MTBC lineages, but also to detect mixed-lineage strain mixtures and estimate their proportions. Fastlin offers a user-friendly and efficient solution for MTBC lineage typing, complementing existing tools and facilitating large-scale analysis.

bioinformatics↗