Biology subjects

Hall, M. B.

Publications and source records attributed to Hall, M. B..

12 recordsLinked to original sources

bioRxiv2026

Evaluating Nanopore read trimming and quality filtering in bacterial genomics

Oxford Nanopore Technologies (ONT) sequencing is widely used in bacterial genomics, particularly for reference-based variant calling and de novo assembly. Both analyses routinely start with quality control (QC) preprocessing, but as ONT basecalling accuracy has improved, its necessity and downstream trade-offs remain poorly defined. We benchmarked 17 preprocessing combinations, built from three adapter trimmers (Dorado, Porechop_ABI, Barbell) and three quality-based processing tools (Filtlong, Chopper, SeqKit) with untrimmed and unprocessed controls, plus Fastplong, which performs both. Each combination was evaluated across 13 bacterial isolates at 100x and 20x depth under the HAC and SUP basecalling models. Variant calling was assessed by F1 score. Assemblies were scored on sequence accuracy, contiguity, replicon recovery and residual adapter contamination, then aggregated into a weighted arithmetic mean using weights from a survey of the microbial genomics community. With SUP basecalling, raw reads with no trimming or quality-based processing produced accurate SNP calls and outperformed several pipelines that applied both operations; adapter trimming alone left SNP accuracy unchanged. Subsetting reads by combined length and quality with Filtlong gave the highest SNP F1 scores at both depths, except with Barbell at 20x, where it ranked second lowest. As is often the case in benchmarking studies, assembly results involved trade-offs, and the best tool depends on which aspects of performance matter most for a given use. Filtlong produced the lowest error rates, but every Filtlong pipeline lost three small plasmids at both depths. Barbell trimming raised error rates and reduced contiguity at 20x, while skipping adapter trimming left residual adapter and barcode sequences in assemblies. Under the community-derived weights, Porechop_ABI followed by Chopper had the highest overall score at both depths, without winning any single metric. QC tool choice should therefore be guided by the downstream analysis goal rather than applied indiscriminately across all workflows.

bioinformatics↗

bioRxiv2026

Interrogating the Escherichia coli epitranscriptome via CRISPR interference and Nanopore native RNA sequencing

Epitranscriptomics has recently gained significant momentum due to technological advances and translational applications, however, studies on bacterial RNA modifications remain limited. Bacterial RNA remains notoriously prone to degradation and methodologies to investigate the epitranscriptome are challenging. Prior research has shown RNA modifications modulate antimicrobial resistance, virulence and pathogenicity. This research employed CRISPR interference to knock down five known Escherichia coli rRNA modification genes (rlmF, rlmJ, rluD, rsmF and rsmG) in three E. coli strains. These isolates underwent growth curves, proteome analysis and native RNA sequencing CRISPRi adequately silenced the majority of RNA modification genes in E. coli (>80% reduction). Significant growth delays were associated with rlmF, rsmF and rsmG repression. Unique protein pathways corresponding with RNA modification loss were found for rlmJ (TreB, XylF), rluD (CysH, HycB, PutP, TrpB), rsmF (EvgA) and rsmG (OppC). Known rRNA modification sites for rluD ({Psi}) and rsmG (m7G) were detected from analysis of nanopore electrical signal, however, only a weak signal was apparent for m6A (rlmF, rlmJ) and m5C (rsmF) modifications. The inhibition of rRNA modifications resulted in mRNA modification changes including for genes ompC, cspC, dbhA, dbhB and secY. Our work provides an approach for unravelling the epitranscriptome of E. coli and gain insight into its functional role.

microbiology↗

bioRxiv2026

Standalone nanopore sequencing for foodborne pathogen surveillance: a large-scale evaluation and quality control framework

Whole-genome sequencing (WGS) is central to foodborne pathogen surveillance and cross-border outbreak detection. Long-read sequencing using Oxford Nanopore Technologies (ONT) promises rapid, complete, and cost-effective genome assemblies in a single workflow. However, the adoption of standalone ONT sequencing of native DNA has been slowed by concerns that DNA modifications can compromise per-base sequencing accuracy and downstream genotyping. In this study, we evaluated ONT-only sequencing performance across 294 genetically diverse isolates representing ten major foodborne pathogens. Using the SUP@v5.2 basecalling model at 50x coverage, 97.3% (286/294) of the ONT assemblies produced identical or near-identical cgMLST profiles ([≤]3 allelic differences) as Illumina-polished hybrid assemblies. Elevated error rates were observed in four Salmonella enterica serovar Kentucky and four Listeria monocytogenes isolates and were associated with the presence of specific DNA phosphorothioation or methylation systems. Re-basecalling the same dataset with the newly released HAC@v6.0 model revealed a different error profile: although 93.5% (275/294) of assemblies remained highly accurate, all 13 isolates carrying dnd (DNA phosphorothioation) or dpd (7-deazaguanine modification) systems, including isolates of S. enterica, Cronobacter sakazakii, and Vibrio parahaemolyticus, exhibited high error rates, suggesting that such atypical modifications were not adequately represented in the models training dataset. To enable rapid identification of unreliable assemblies, we developed alpaqa, a lightweight computational tool that detects systematic nanopore assembly errors without requiring supplemental short-read data or reference genomes. By identifying affected assemblies, alpaqa provides a quality safeguard for ONT-only workflows. Masking low-quality bases in assemblies flagged by alpaqa improved cgMLST accuracy, although this reduced the number of callable loci and therefore genotyping resolution. Our findings demonstrate that standalone ONT sequencing of native DNA is sufficiently accurate for routine foodborne pathogen surveillance when combined with appropriate quality control, supporting its use in harmonised genomic surveillance frameworks. Data summaryAll sequencing data generated in this study have been submitted to the NCBI Sequence Read Archive. Accession numbers for Illumina and ONT (SUP@v5.2) reads are listed in Supplementary Table S1. Raw pod5 files from error-prone isolates have been deposited in SquiDBase (SQB000021). Alpaqa is available at github.com/MBiggel/alpaqa/. An automated ONT assembly and quality control pipeline integrating alpaqa is available at github.com/MBiggel/boap/. Impact statementThis study demonstrates that standalone Oxford Nanopore sequencing of native DNA can achieve highly accurate genotyping for routine foodborne pathogen surveillance across diverse species. We show that the remaining inaccuracies are linked to specific DNA modification systems, including phosphorothioation and 7-deazaguanine modifications, which are identified here as previously unrecognised sources of systematic sequencing errors. To address this limitation, we introduce alpaqa, a reference-free method for detecting such error-prone assemblies, providing a practical quality-control framework for ONT-only workflows. Together, these results support the reliable use of nanopore sequencing in routine genomic surveillance.

genomics↗

bioRxiv2025

Genome graphs reveal the importance of structural variation in Mycobacterium tuberculosis evolution and drug resistance

Structural variants (SVs) are increasingly recognized as key drivers of bacterial evolution, yet their role has not been explored thoroughly. This is due to limitations in traditional short-read sequencing and linear reference-based analyses, which can miss complex structural changes. Tuberculosis (TB), a disease caused by Mycobacterium tuberculosis (Mtb), remains a major global health concern. In this study, we harness long-read sequencing technologies and genome graph tools to construct a Mtb pangenome reference graph (PRG) from 859 high-quality, diverse, long-read assemblies. To enable accurate genotyping of SVs leveraging the PRG, we developed miniwalk, a tool that outperforms a traditional linear genome-based approach in precision for SV detection. We characterize patterns of structural variation genome-wide, revealing a virulence-associated ESX-5 deletion to be recurrent across the phylogeny, and fixed in a sub-lineage of L4. Systematic screens for additional genes that are recurrently affected by SVs implicated those related to metal homeostasis, including a copper exporter fixed in the widely distributed L1.2.1 sub-lineage. Lastly, we genotyped 41,134 isolates and found SVs putatively associated with resistance to various first and second-line drugs. These findings underscore the broader role of SVs in shaping Mtb diversity, highlighting their importance in both understanding evolution and designing strategies to combat drug-resistant TB.

genomics↗

bioRxiv2024

Genome size estimation from long read overlaps

SummaryAccurate genome size estimation is an important component of genomic analyses, though existing tools are primarily optimised for short-read data. We present LRGE, a novel tool that uses read-to-read overlap information to estimate genome size in a reference-free manner. LRGE calculates per-read genome size estimates by analysing the expected number of overlaps for each read, considering read lengths and a minimum overlap threshold. The final size is taken as the median of these estimates, ensuring robustness to outliers such as reads with no overlaps. Additionally, LRGE provides an expected confidence range for the estimate. LRGE outperforms k-mer-based methods in both accuracy and computational efficiency and produces genome size estimates comparable to those from assembly-based approaches, like Raven, while using significantly less computational resources. We validate LRGE on a large, diverse bacterial dataset and confirm it generalises to eukaryotic datasets. Availability and implementationOur method, LRGE (Long Read-based Genome size Estimation from overlaps), is implemented in Rust and is available as a precompiled binary for most architectures, a Bioconda package, a prebuilt container image, and a crates.io package as a binary (lrge) or library (liblrge). The source code is available at https://github.com/mbhall88/lrge under an MIT license.

bioinformatics↗

bioRxiv2024

Benchmarking reveals superiority of deep learning variant callers on bacterial nanopore sequence data

Variant calling is fundamental in bacterial genomics, underpinning the identification of disease transmission clusters, the construction of phylogenetic trees, and antimicrobial resistance prediction. This study presents a comprehensive benchmarking of SNP and indel variant calling accuracy across 14 diverse bacterial species using Oxford Nanopore Technologies (ONT) and Illumina sequencing. We generate gold standard reference genomes and project variations from closely-related strains onto them, creating biologically realistic distributions of SNPs and indels. Our results demonstrate that ONT variant calls from deep learning-based tools delivered higher SNP and indel accuracy than traditional methods and Illumina, with Clair3 providing the most accurate results overall. We investigate the causes of missed and false calls, highlighting the limitations inherent in short reads and discover that ONTs traditional limitations with homopolymer-induced indel errors are absent with high-accuracy basecalling models and deep learning-based variant calls. Furthermore, our findings on the impact of read depth on variant calling offer valuable insights for sequencing projects with limited resources, showing that 10x depth is sufficient to achieve variant calls that match or exceed Illumina. In conclusion, our research highlights the superior accuracy of deep learning tools in SNP and indel detection with ONT sequencing, challenging the primacy of short-read sequencing. The reduction of systematic errors and the ability to attain high accuracy at lower read depths enhance the viability of ONT for widespread use in clinical and public health bacterial genomics.

bioinformatics↗

bioRxiv2023

Comparison of computational methods for host removal and classification of mycobacteria from clinical metagenomic data

BackgroundCulture-free real-time sequencing of clinical metagenomic samples promises both rapid pathogen detection and antimicrobial resistance profiling. However, this approach introduces the risk of patient DNA leakage. To mitigate this risk, we need near-comprehensive removal of human DNA sequence at the point of sequencing, typically involving use of resource-constrained devices. Existing benchmarks have largely focused on use of standardised databases and largely ignored the computational requirements of depletion pipelines as well as the impact of human genome diversity. ResultsWe benchmarked host removal pipelines on simulated Illumina and Nanopore metagenomic samples. We found that construction of a custom kraken database containing diverse human genomes results in the best balance of accuracy and computational resource usage. In addition, we benchmarked pipelines using kraken and minimap2 for taxonomic classification of Mycobacterium reads using standard and custom databases. With a database representative of the Mycobacterium genus, both tools obtained near-perfect precision and recall for classification of Mycobacterium tuberculosis. Computational efficiency of these custom databases was again superior to most standard approaches, allowing them to be executed on a laptop device. ConclusionsNanopore sequencing and a custom kraken human database with a diversity of genomes leads to superior host read removal from simulated metagenomic samples while being executable on a laptop. In addition, constructing a taxon-specific database provides excellent taxonomic read assignment while keeping runtime and memory low. We make all customised databases and pipelines freely available.

bioinformatics↗

bioRxiv2023

Drug resistance prediction for Mycobacterium tuberculosis with reference graphs

2.The dominant paradigm for analysing genetic variation relies on a central idea: all genomes in a species can be described as minor differences from a single reference genome. However, this approach can be problematic or inadequate for bacteria, where there can be significant sequence divergence within a species. Reference graphs are an emerging solution to the reference bias issues implicit in the "single-reference" model. Such a graph represents variation at multiple scales within a population - e.g., nucleotide- and locus-level. The genetic causes of drug resistance in bacteria have proven comparatively easy to decode compared with studies of human diseases. For example, it is possible to predict resistance to numerous anti-tuberculosis drugs by simply testing for the presence of a list of single nucleotide polymorphisms and insertion/deletions, commonly referred to as a catalogue. We developed DrPRG (Drug resistance Prediction with Reference Graphs) using the bacterial reference graph method Pandora. First, we outline the construction of a Mycobacterium tuberculosis drug resistance reference graph, a process that can be replicated for other species. The graph is built from a global dataset of isolates with varying drug susceptibility profiles, thus capturing common and rare resistance- and susceptible-associated haplotypes. We benchmark DrPRG against the existing graph-based tool Mykrobe and the haplotype-based approach of TBProfiler using 44,709 and 138 publicly available Illumina and Nanopore samples with associated phenotypes. We find DrPRG has significantly improved sensitivity and specificity for some drugs compared to these tools, with no significant decreases. It uses significantly less computational memory than both tools, and provides significantly faster runtimes, except when runtime is compared to Mykrobe on Nanopore data. We discover and discuss novel insights into resistance-conferring variation for M. tuberculosis - including deletion of genes katG and pncA - and suggest mutations that may warrant reclassification as associated with resistance. 3. Impact statementMycobacterium tuberculosis is the bacterium responsible for tuberculosis (TB). TB is one of the leading causes of death worldwide; before the coronavirus pandemic it was the leading cause of death from a single pathogen. Drug-resistant TB incidence has recently increased, making the detection of resistance even more vital. In this study, we develop a new software tool to predict drug resistance from whole-genome sequence data of the pathogen using new reference graph models to represent a reference genome. We evaluate it on M. tuberculosis against existing tools for resistance prediction and show improved performance. Using our method, we discover new resistance-associated variations and discuss reclassification of a selection of existing mutations. As such, this work contributes to TB drug resistance diagnostic efforts. In addition, the method could be applied to any bacterial species, so is of interest to anyone working on antimicrobial resistance. 4. Data summaryThe authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files. The software method presented in this work, DrPRG, is freely available from GitHub under an MIT license at https://github.com/mbhall88/drprg. We used commit 9492f25 for all results via a Singularity[1] container from the URI docker://quay.io/mbhall88/drprg:9492f25. All code used to generate results for this study are available on GitHub at https://github.com/mbhall88/drprg-paper. All data used in this work are freely available from the SRA/ENA/DRA and a copy of the datasheet with all associated phenotype information can be downloaded from the archived repository at https://doi.org/10.5281/zenodo.7819984 or found in the previously mentioned GitHub repository. The Mycobacterium tuberculosis index used in this work is available to download through DrPRG via the command drprg index --download mtb@20230308 or from GitHub at https://github.com/mbhall88/drprg-index.

bioinformatics↗

bioRxiv2022

Genomic Sequencing from Sputum for Tuberculosis Disease Diagnosis, Lineage Determination and Drug Susceptibility Prediction

BackgroundUniversal access to drug susceptibility testing for newly diagnosed tuberculosis patients is recommended. Access to culture-based diagnostics remains limited and targeted molecular assays are vulnerable to emerging resistance conferring mutations. Improved sample preparation protocols for direct-from-sputum sequencing of Mycobacterium tuberculosis would accelerate access to comprehensive drug susceptibility testing and molecular typing. MethodsWe assessed a thermo-protection buffer-based direct-from-sample M. tuberculosis whole-genome sequencing protocol. We prospectively processed and analyzed 60 acid-fast bacilli smear-positive sputum samples from tuberculosis patients in India and Madagascar. A diversity of semi-quantitative smear positivity level samples were included. Sequencing was performed using Illumina and MinION (monoplex and multiplex) technologies. We measured the impact of bacterial inoculum and sequencing platforms on M. tuberculosis genomic mean read depth, drug susceptibility prediction performance and typing accuracy. ResultsM. tuberculosis was identified from 88% (Illumina), 89% (MinION-monoplex) and 83% (MinION-multiplex) of samples for which sufficient DNA could be extracted. The fraction of M. tuberculosis reads from MinION sequencing was lower than from Illumina, but monoplexing grade 3+ sputum samples on MinION produced higher read depth than Illumina (p<0.05) and MinION multiplex (p<0.01). No significant difference in overall sensitivity and specificity of drug susceptibility predictions was seen across these sequencing modalities or within each sequencing technology when stratified by smear grade. Lineage typing agreement percentages between direct and culture-based sequencing were 85% (MinION-monoplex), 88% (Illumina) and 100% (MinION-multiplex) ConclusionsM. tuberculosis direct-from-sample whole-genome sequencing remains challenging. Improved and affordable sample treatment protocols are needed prior to clinical deployment.

microbiology↗

bioRxiv2021

Minos: variant adjudication and joint genotyping of cohorts of bacterial genomes

Short-read variant calling for bacterial genomics is a mature field, and there are many widely-used software tools. Different underlying approaches (eg pileup, local or global assembly, paired-read use, haplotype use) lend each tool different strengths, especially when considering non-SNP (single nucleotide polymorphism) variation or potentially distant reference genomes. It would therefore be valuable to be able to integrate the results from multiple variant callers, using a robust statistical approach to "adjudicate" at loci where there is disagreement between callers. To this end, we present a tool, Minos, for variant adjudication by mapping reads to a genome graph of variant calls. Minos allows users to combine output from multiple variant callers without loss of precision. Minos also addresses a second problem of joint genotyping SNPs and indels in bacterial cohorts, which can also be framed as an adjudication problem. We benchmark on 62 samples from 3 species (Mycobacterium tuberculosis, Staphylococcus aureus, Klebsiella pneumoniae) and an outbreak of 385 M. tuberculosis samples. Finally, we joint genotype a large M. tuberculosis cohort (N{approx}15k) for which the rifampicin phenotype is known. We build a map of non-synonymous variants in the RRDR (rifampicin resistance determining region) of the rpoB gene and extend current knowledge relating RRDR SNPs to heterogeneity in rifampicin resistance levels. We replicate this finding in a second M. tuberculosis cohort (N{approx}13k). Minos is released under the MIT license, available at https://github.com/iqbal-lab-org/minos.

bioinformatics↗

bioRxiv2020

Nucleotide-resolution bacterial pan-genomics with reference graphs

BackgroundBacterial genomes follow a U-shaped frequency distribution whereby most genomic loci are either rare (accessory) or common (core); the union of these is the pan-genome. The alignable fraction of two genomes from a single species can be low (e.g. 50-70%), such that no single reference genome can access all single nucleotide polymorphisms (SNPs). The pragmatic solution is to choose a close reference, and analyse SNPs only in the core genome. Given much bacterial adaptability hinges on the accessory genome, this is an unsatisfactory limitation. ResultsWe present a novel pan-genome graph structure and algorithms implemented in the software pandora, which approximates a sequenced genome as a recombinant of reference genomes, detects novel variation and then pan-genotypes multiple samples. The method takes fastq as input and outputs a multi-sample VCF with respect to an inferred data-dependent reference genome, and is available at https://github.com/rmcolq/pandora. Constructing a reference graph from 578 E. coli genomes, we analyse a diverse set of 20 E. coli isolates. We show pandora recovers at least 13k more rare SNPs than single-reference based tools, achieves equal or better error rates with Nanopore as with Illumina data, 6-24x lower Nanopore error rates than other tools, and provides a stable framework for analysing diverse samples without reference bias. We also show that our inferred recombinant VCF reference genome is significantly better than simply picking the closest RefSeq reference. ConclusionsThis is a step towards comprehensive cohort analysis of bacterial pan-genomic variation, with potential impacts on genotype/phenotype and epidemiological studies.

bioinformatics↗

bioRxiv2017

Chiron: Translating nanopore raw signal directly into nucleotide sequence using deep learning

Sequencing by translocating DNA fragments through an array of nanopores is a rapidly maturing technology which offers faster and cheaper sequencing than other approaches. However, accurately deciphering the DNA sequence from the noisy and complex electrical signal is challenging. Here, we report Chiron, the first deep learning model to achieve end-to-end basecalling: directly translating the raw signal to DNA sequence without the error-prone segmentation step. Trained with only a small set of 4000 reads, we show that our model provides state-of-the-art basecalling accuracy even on previously unseen species. Chiron achieves basecalling speeds of over 2000 bases per second using desktop computer graphics processing units.

bioinformatics↗