Biology subjects

Lund, O.

Publications and source records attributed to Lund, O..

16 recordsLinked to original sources

bioRxiv2026

Resolving Genome-to-Phenotype Links in Bacteria: Machine-Learned Inference from Downsampled k-mer Representations

Standard approaches to bacterial phenotyping often treat the entire genome as the fundamental unit of information, resulting in high-dimensional inputs that may contain significant redundancy. Consequently, current bacterial phenotyping techniques typically rely on the assumption that entire sequences are required for accurate predictions. While downsampling based on min-hashing or prefix filtering has been used for clustering, its utility as a direct input for predictive machine learning remains underexplored. Here, we show that a novel prefix-based downsampling algorithm can reduce the size of genomes while maintaining relatively high predictive accuracy on phenotype prediction tasks. By combining a prefix reduction strategy with the specificity of short k-mers, we developed a method to downsample entire genomes into k-mer frequency matrices and k-mer-on-a-string representations. We found that ensemble models, such as Random Forest and Gradient Boosting, trained on k-mer frequency matrices from downsampled genome representations outperformed more complex deep learning architectures with the same downsampled representation, particularly on datasets with limited data or highly similar genomes. We were able demonstrate explainability by tracing back the k-mers with the most impact on the models to genes coding for the specific phenotype. Our results demonstrate that downsampling genomic data can yield models with good predictive power thus establishing an alternative when using full genomes is infeasible. We present an approach that offers relatively high performance on bacterial phenotyping tasks and demonstrates a path forward towards lightweight Genome Language Models that will enable analysis of entire genomes.

bioinformatics↗

bioRxiv2025

Multiple instance learning on tile level-pathologyimages provides accurate and interpretableclassification for breast cancer molecular subtypes

Accurate breast cancer molecular subtyping is critical for treatment decisions, yet standard methods such as immunohistochemistry and gene expression profiling are costly and labor intensive. Deep learning classification approaches using Hematoxylin and Eosin-stained whole slide images are an active area of research. However, many existing methods rely on large, high-quality annotated datasets where tumor regions are manually outlined for segmentation. This process is costly, does not scale well, depends on expert pathologists, and may ignore relevant tumor microenvironment features or reflect subjective labelling decisions. Here, we present an annotation-free, weakly supervised pipeline and web-based tool for breast cancer molecular subtyping using a computational pathology foundation model. A total of 1433 WSIs from three public cohorts (TCGA-BRCA, CPTAC-BRCA, and the Warwick HER2 cohort) were tiled into 224x224 patches without overlap at 20x magnification. Tile-level embeddings were extracted with a foundation model, and slide-level representations were obtained by mean pooling. We evaluated one-vs-rest classifiers including cosine similarity, logistic regression, and attention-based multiple instance learning. On a held-out test set of 287 WSIs, calibrated logistic regression achieved a macro F1 score of 0.75 using slide embeddings, while attention-based MIL reached 0.83 using tile embeddings. Luminal A and Basal subtypes were predicted reliably, whereas Luminal B remained challenging. Novel attention and probability heatmaps highlights spatial regions most informative for predictions, supporting qualitative interpretability. These results demonstrate accurate and interpretable breast cancer subtyping without tumor annotations, and we provide a web server to support pathology diagnostics.

bioinformatics↗

bioRxiv2025

When Not to Kill Your Host in Competitive Bust-Boom Environments

Temperate phages, which can either kill their host cells or integrate into them, struggle to compete with purely virulent phages in environments with plenty of available hosts. This suggests that their survival strategy is fine-tuned for unstable conditions, where they hedge their bets between immediate replication and long-term persistence as an integrated prophage. In this study, we explored how temperate phages make these life-or-death decisions, both in isolation and when competing with other phages. We found that when temperate phages compete with each other, those with relatively stable lysogens survive better. Environments with competitive temperate phages further select for lower lysogeny frequency. Our findings suggest that dosage-dependent lysogeny choice is adapted to competing phages with overlapping immunity. In environments where phages can disperse between separate sub-systems that fluctuate independently, temperate phages struggle to survive against virulent phages.

microbiology↗

bioRxiv2023

MAGinator enables strain-level quantification of de novo MAGs

MotivationMetagenomic sequencing has provided great advantages in the characterization of microbiomes, but currently available analysis tools lack the ability to combine strain-level taxonomic resolution and abundance estimation with functional profiling of assembled genomes. In order to define the microbiome and its associations with human health, improved tools are needed to enable comprehensive understanding of the microbial composition and elucidation of the phylogenetic and functional relationships between the microbes. ResultsHere, we present MAGinator, a freely available tool, tailored for the profiling of shotgun metagenomics datasets. MAGinator provides de novo identification of subspecies-level microbes and accurate abundance estimates of metagenome-assembled genomes (MAGs). MAGinator utilises the information from both gene- and contig-based methods yielding insight into both taxonomic profiles and the origin of genes as well as genetic content, used for inference of functional content of each sample by host organism. Additionally, MAGinator facilitates the reconstruction of phylogenetic relationships between the MAGs, providing a framework to identify clade-level differences within subspecies MAGs. Availability and implementationMAGinator is available as a Python module at https://github.com/Russel88/MAGinator ContactTrine Zachariasen, trine_zachariasen@hotmail.com

bioinformatics↗

bioRxiv2021

PlasmidHostFinder: Prediction of plasmid hosts using random forest

Plasmids play a major role facilitating the spread of antimicrobial resistance between bacteria. Understanding the host range and dissemination trajectories of plasmids is critical for surveillance and prevention of antimicrobial resistance. Identification of plasmid host ranges could be improved using automated pattern detection methods, compared to homology-based methods due to the diversity and genetic plasticity of plasmids. In this study, we developed a method for predicting the host range of plasmids based on the random forest machine learning method. We trained the models with 8,519 plasmids from 359 different bacterial species per taxonomic level, where the models achieved 0.662 and 0.867 Matthews correlation coefficients at the species and order levels, respectively. Our results suggest that despite the diverse nature and genetic plasticity of plasmids, our random forest model can accurately distinguish between plasmid hosts. This tool can be used online through Center for Genomic Epidemiology (https://cge.cbs.dtu.dk/services/PlasmidHostFinder/). ImportanceAntimicrobial resistance is a global health threat to humans and animals causing high mortality and morbidity, and effectively ending decades of success in fighting against bacterial infections. Plasmids confer extra genetic capabilities to the host organisms through accessory genes, which can encode antimicrobial resistance and virulence factors. In addition to lateral inheritance, plasmids can be transferred horizontally between bacterial taxa. Therefore, detecting the host range of plasmids is crucial for understanding and predicting the dissemination trajectories of extrachromosomal genes and bacterial evolution, as well as for taking effective counter measures against antimicrobial resistance.

bioinformatics↗

bioRxiv2021

Deep integrative models for large-scale human genomics

Polygenic risk scores (PRSs) are expected to play a critical role in achieving precision medicine. Currently, PRS predictors are generally based on linear models using summary statistics, and more recently individual-level data. However, these predictors mainly capture additive relationships and are limited in data modalities they can use. Here, we developed a deep learning framework (EIR) for PRS prediction which includes a model, genome-local-net (GLN), specifically designed for large scale genomics data. The framework supports multi-task (MT) learning, automatic integration of other clinical and biochemical data, and model explainability. When applied to individual level data in the UK Biobank, we found that GLN outperformed LASSO for a wide range of diseases and in particularly autoimmune diseases. Furthermore, we show that this was likely due to modelling epistasis, and we showcase this by identifying widespread epistasis for Type 1 Diabetes. Furthermore, we trained PRS by integrating genotype, blood, urine and anthropometrics and found that this improved performance for 93% of 290 diseases and disorders considered. Finally, we found that including genotype data provided better calibrated PRS models compared to using measurements alone. EIR is available at https://github.com/arnor-sigurdsson/EIR.

bioinformatics↗

bioRxiv2020

MINTyper: A method for generating phylogenetic distance matrices with long read sequencing data

For detection of clonal outbreaks in clinical settings, we present a complete pipeline that generates a SNP-distance matrix from a set of sequencing reads. Importantly, the program is able to handle a separate mix of both short reads from the Illumina sequencing platforms and long reads from Oxford Nanopore Technologies (ONT) platforms as input. MINTyper performs automated reference identification, alignment, alignment trimming, optional methylation masking and pairwise distance calculations. With this approach, we could rapidly and accurately cluster a set of DNA sequenced isolates, with a known epidemiological relationship to confirm the clustering. Functions were built to allow for both high-accuracy methylation-aware base-called MinION reads (hac_m Q10) and fast generated lower-quality reads (fast Q8) to be used, also in combination with Illumina data. With fast Q8 reads a higher number of base pairs were excluded from the calculated distance matrix, compared to the high-accuracy methylation-aware Q10 base-calling of ONT data. Nonetheless, when using different qualities of ONT data with corresponding input parameters, the clustering of isolates were nearly identical.

bioinformatics↗

bioRxiv2019

Understanding and predicting ciprofloxacin minimum inhibitory concentration in Escherichia coli with machine learning

2.A possible way to tackle the crisis of antimicrobial resistance development is a strict policy when prescribing antibiotics. Thus, it is important that prescriptions are based on antimicrobial susceptibility data to ensure effective treatment outcomes. The increasing availability of next-generation sequencing (NGS), bacterial whole genome sequencing (WGS) can facilitate a more reliable and faster alternative to traditional phenotyping for the detection and surveillance of AMR. This work proposes a machine learning approach that can predict the minimum inhibitory concentration (MIC) for a given antibiotic, here ciprofloxacin, on the basis of both genome-wide mutation profiles and profiles of acquired antimicrobial resistance genes (ARG). We analysed 704 Escherichia coli genomes combined with their respective MIC measurements for ciprofloxacin originating from different countries. The four most important predictors found by the model, mutations in gyrA residues Ser83 and Asp87, a mutation in parC residue Ser80 and presence of any qnrS gene, have been experimentally validated before. Using only these four predictors in a linear regression model, 65% and 92% of the test samples MIC were correctly predicted within a two- and a four-fold dilution range, respectively. The presented work goes further than the typical predictions that use machine learning as a black box model concept. The recent progress in WGS technology in combination with machine learning analysis approaches indicates that in the near future WGS of bacteria might become cheaper and faster than a MIC measurement. 3. Impact statementWhole genome sequencing has become the standard approach to study molecular epidemiology of bacteria. However, the application of WGS in the clinical microbiology laboratory as part of individual patient diagnostics still requires significant steps forward, in particular with respect to prediction of antibiotic susceptibility based on DNA sequence. Whilst the majority of studies of prediction of susceptibility have used a binary outcome (susceptible/resistant), a quantitative prediction of susceptibility, such as the MIC, will allow for earlier detection of trends in increasing resistance as well as the flexibility to follow potential adjustments in definitions of susceptible (wild type) and resistant (non-wild type) categories (breakpoints/ epidemiological cut-off values). 4. Data summaryIn this study, 704 E. coli genomes combined with MIC measurement for ciprofloxacin were analysed (24). Paired-end sequencing was performed on all isolates and the results were stored in FASTQ format. The isolates originated from five countries, Denmark, Italy, USA, UK, and Vietnam. The MIC distribution for these isolates is depicted in Table 1. Out of 704, 266 E. coli genomes had no country metadata available and were used as an independent test set. All data were deposited in the AMR Data Hub (24) which consists of raw sequencing data, ciprofloxacin minimum inhibitory concentrations, and additional metadata such as the origin of the samples. O_TBL View this table: org.highwire.dtl.DTLVardef@5a31d2org.highwire.dtl.DTLVardef@17cfa2eorg.highwire.dtl.DTLVardef@ce3ccorg.highwire.dtl.DTLVardef@3f6721org.highwire.dtl.DTLVardef@8bb604_HPS_FORMAT_FIGEXP M_TBL O_FLOATNOTABLE 1C_FLOATNO O_TABLECAPTIONThe collected and used data in the analysis grouped by country and MIC values. C_TABLECAPTION C_TBL Publicly available sequencing data was used from projects PRJEB21131, PRJNA266657, PRJNA292901, PRJNA292904, PRJNA292902, PRJDB7087, PRJEB21880, PRJEB21997, PRJEB14086 and PRJEB16326. Download and analysis scripts are available at https://github.com/patbaa/AMR_ciprofloxacin. iTOL phylogenetic tree is available at https://itol.embl.de/tree/14511722611491391569485969. The authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files.

microbiology↗

bioRxiv2019

CCMetagen: comprehensive and accurate identification of eukaryotes and prokaryotes in metagenomic data

High-throughput sequencing of DNA and RNA from environmental and host-associated samples (metagenomics and metatranscriptomics) is a powerful tool to assess which organisms are present in a sample. Taxonomic identification software usually align individual short sequence reads to a reference database, sometimes containing taxa with complete genomes only. This is a challenging task given that different species can share identical sequence regions and complete genome sequences are only available for a fraction of organisms. A recently developed approach to map sequence reads to reference databases involves weighing all high scoring read-mappings to the data base as a whole to produce better-informed alignments. We used this novel concept in read mapping to develop a highly accurate metagenomic classification pipeline named CCMetagen. Using simulated fungal and bacterial metagenomes, we demonstrate that CCMetagen substantially outperforms other commonly used metagenome classifiers, attaining a 3 - 1580 fold increase in precision and a 2 - 922 fold increase in F1 scores for species-level classifications when compared to Kraken2, Centrifuge and KrakenUniq. CCMetagen is sufficiently fast and memory efficient to use the entire NCBI nucleotide collection (nt) as reference, enabling the assessment of species with incomplete genome sequence data from all biological kingdoms. Our pipeline efficiently produced a comprehensive overview of the microbiome of two biological data sets, including both eukaryotes and prokaryotes. CCMetagen is user-friendly and the results can be easily integrated into microbial community analysis software for streamlined and automated microbiome studies.

bioinformatics↗

bioRxiv2019

Specific Gut Microbiome Members are Associated with Distinct Immune Markers in Allogeneic Hematopoietic Stem Cell Transplantation

BackgroundIncreasing evidence reveals the importance of the microbiome in health and disease and inseparable host-microbial dependencies. Host-microbe interactions are highly relevant in patients receiving allogeneic hematopoietic stem cell transplantation, (HSCT), i.e. a replacement of the cellular components of the patients immune system with that of a foreign donor. HSCT is employed as curative immunotherapy for a number of non-malignant and malignant hematologic conditions, including acute lymphoblastic leukemia. The procedure can be accompanied by severe side effects such as infections, acute graft-versus-host disease (aGvHD), and death. Here, we performed a longitudinal analysis of immunological markers, immune reconstitution and gut microbiota composition in relation to clinical outcomes in children undergoing HSCT. Such an analysis could reveal biomarkers, e.g. at the time point prior to HSCT, that in the future could be used to predict which patients are of high risk in relation to side effects and clinical outcomes and guide treatment strategies accordingly.\n\nResultsIn two multivariate analyses (sparse partial least squares regression and canonical correspondence analysis), we identified three consistent clusters: (1) High concentrations of the antimicrobial peptide human beta-defensin 2 (hBD2) prior to the transplantation in patients with high abundances of Lactobacillaceae, who later developed moderate or severe aGvHD and exhibited high mortality. (2) Rapid reconstitution of NK and B cells in patients with high abundances of obligate anaerobes such as Ruminococcaceae, who developed no or mild aGvHD and exhibited low mortality. (3) High inflammation, indicated by high levels of C-reactive protein, in patients with high abundances of facultative anaerobic bacteria such as Enterobacteriaceae. Furthermore, we observed that antibiotic treatment influenced the bacterial community state.\n\nConclusionsWe identify multivariate associations between specific microbial taxa, host immune markers, immune cell reconstitution and clinical outcomes in relation to HSCT. Our findings encourage further investigations into establishing longitudinal surveillance of the intestinal microbiome and relevant immune markers, such as hBD2, in HSCT patients. Profiling of the microbiome may prove useful as a prognostic tool that could help identify patients at risk of poor immune reconstitution and adverse outcomes, such as aGvHD and death, upon HSCT, providing actionable information in guiding precision medicine.

microbiology↗

bioRxiv2019

The COMPARE Data Hubs

Data sharing enables research communities to exchange findings and build upon the knowledge that arises from their discoveries. Areas of public and animal health as well as food safety would benefit from rapid data sharing when it comes to emergencies. However, ethical, regulatory, and institutional challenges, as well as lack of suitable platforms which provide an infrastructure for data sharing in structured formats often lead to data not being shared, or at most shared in form of supplementary materials in journal publications. Here, we describe an informatics platform that includes workflows for structured data storage, managing and pre-publication sharing of pathogen sequencing data and its analysis interpretations with relevant stakeholders.

bioinformatics↗

bioRxiv2019

Large scale automated phylogenomical analysis of bacterial whole-genome isolates and the Evergreen platform

Public health authorities whole-genome sequence thousands of pathogenic isolates each month for microbial diagnostics and surveillance of pathogenic bacteria. The computational methods have not kept up with the deluge of data and need for real-time results. We have therefore created a bioinformatics pipeline for rapid subtyping and continuous phylogenomic analysis of bacterial samples, suited for large-scale surveillance. To decrease the computational burden, a two level clustering strategy is employed. The data is first divided into sets by matching each isolate to a closely related reference genome. The reads then are aligned to the reference to gain a consensus sequence and SNP based genetic distance is calculated between the sequences in each set. Isolates are clustered together with a threshold of 10 SNPs. Finally, phylogenetic trees are inferred from the non-redundant sequences and the clustered isolates are placed on a clade with the cluster representative sequence. The method was benchmarked and found to be accurate in grouping outbreak strains together, while discriminating from non-outbreak strains. The pipeline was applied in Evergreen Online, which processes publicly available sequencing data from foodborne bacterial pathogens on a daily basis, updating the phylogenetic trees as needed. It has so far placed more than 100,000 isolates into phylogenies, and has been able to keep up with the daily release of data. The trees are continuously published on https://cge.cbs.dtu.dk/services/Evergreen

bioinformatics↗

bioRxiv2019

Accelerating surveillance and research of antimicrobial resistance - an online repository for sharing of antimicrobial susceptibility data associated with whole genome sequences

Antimicrobial resistance (AMR) is an emerging threat to modern medicine. Improved diagnostics and surveillance of resistant bacteria require the development of next generation analysis tools and collaboration between international partners. Here, we present the "AMR data hub", an online infrastructure for storage and sharing of structured phenotypic AMR data linked to bacterial genome sequences. Leveraging infrastructure built by the European COMPARE Consortium and structured around the European Nucleotide Archive (ENA), the AMR data hub already provides an extensive data collection for some 500 isolates with linked genome and AMR data. Representing these data in standardized formats, we provide tools for the validation and submission of new data and services supporting search, browse and retrieval. The current collection was created through a collaboration by several partners from the European COMPARE Consortium, demonstrating the capacities and utility of the AMR data hub and its associated tools. We anticipate growth of content and offer the hub as a basis for future research into methods to explore and predict AMR.

microbiology↗

bioRxiv2018

Proof of concept: Malaria rapid diagnostic tests and massively parallel sequencing for surveillance of molecular markers of antimalarial resistance in Bissau, Guinea-Bissau during 2014-2017

Real-time and large-scale surveillance of molecular markers of antimalarial drug resistance is a potential method of resistance monitoring, to complement therapeutic efficacy studies in settings where the latter are logistically challenging. This study investigates whether routinely used malaria rapid diagnostic tests (RDTs) can be used for massive parallel amplicon sequencing. RDTs used for malaria diagnosis were routinely collected together with patient age and sex between 2014 and 2017, from two health centres in Bissau, Guinea-Bissau. A subset of positive RDTs (n=2,184) were tested for Plasmodium DNA content. Those containing sufficient Plasmodium DNA (n=1,390) were used for library preparation, consisting of amplification of gene fragments from pfcrt, pfmdr1, pfdhfr, pfdhps and pfK13. A total of 5532 gene fragments were successfully analysed on a single Illumina Miseq flow cell. Pre-screening of samples for Plasmodium DNA content proved necessary and the nested PCR protocol applied for library preparation varied notably in PCR-positivity from 13-87%. We found a high frequency of the pfmdr1 codon 86N at 88%-97%, a significant decrease of the pfcrt wildtype CVMNK haplotype and elevated levels of the pfdhfr/pfdhps quadruple mutant ranging from 33%-51% between 2014-2017. No polymorphisms indicating artemisinin tolerance were discovered. Lastly, the demographic data indicate a large proportion of young adults (66%, interquartile range 11-28 years) presenting with P. falciparum infections. With some caution, our findings suggest that routine collection of RDTs could facilitate large-scale molecular surveillance of antimalarial resistance.\n\nImportance (word count: 147)Continuous spread and repeated emergence of Plasmodium falciparum parasites resistant towards one or more antimalarials represents an enormous threat to current treatment efficacy levels, especially in sub-Saharan Africa, where 90% of malaria infections occur. In order to prevent substantial treatment failure, it is therefore recommended to monitor treatment efficacy every 2-3 years. Therapeutic efficacy studies, however, can present insurmountable logistical and financial challenges in some settings in sub-Saharan Africa. Molecular surveillance of antimalarial resistance is therefore an important proxy for treatment efficacy. However, the scale by which such studies can be performed depends on the development of high-throughput protocols and the accessibility of samples. If RDTs can be used in the high-throughput protocols available with Next Generation Sequencing (NGS)-technology, surveillance can be performed efficiently for any setting in which RDTs are already used for malaria diagnosis. The majority of settings in sub-Saharan Africa have access to RDTs.

molecular biology↗

bioRxiv2017

Abundance and diversity of the fecal resistome in slaughter pigs and broilers in nine European countries

EFFORT groupHaitske Graveland, Alieda van Essen, Bruno Gonzalez-Zorn, Gabriel Moyano, Pascal Sanders, Claire Chauvin, Julie David, Antonio Battisti, Andrea Caprioli, Jeroen Dewulf, Thomas Blaha, Katharina Wadepohl, Maximiliane Brandt, Dariusz Wasyl, Magdalena Skarzynska, Magdalena Zajac, Hristo Daskalov, Helmut W Saatkamp, Katharina D.C. Stark.\n\nAbstractAntimicrobial resistance (AMR) in bacteria and associated human morbidity and mortality is increasing. Use of antimicrobials in livestock selects for AMR that can subsequently be transferred to humans. This flow of AMR between reservoirs demands surveillance in livestock as well as in humans. As part of the EFFORT project (www.effort-against-amr.eu), we have quantified and characterized the acquired resistance gene pools (resistomes) of 181 pig and 178 poultry farms from nine European countries, generating more than 5,000 gigabases of DNA sequence, using shotgun metagenomics. We quantified acquired AMR using the ResFinder database and a database constructed for this study, consisting of AMR genes identified through screening environmental DNA. The pig and poultry resistomes were very different in abundance and composition. There was a significant country effect on the resistomes, more so in pigs than poultry. We found higher AMR loads in pigs, while poultry resistomes were more diverse. We detected several recently described, critical AMR genes, including mcr-1 and optrA, the abundance of which differed both between host species and countries. We found that the total acquired AMR level, was associated with the overall country-specific antimicrobial usage in livestock and that countries with comparable usage patterns had similar resistomes. Novel, functionally-determined AMR genes were, however, not associated with total drug use.

microbiology↗

bioRxiv2017

KmerFinderJS: A Client-Server Method For Fast Species Typing Of Bacteria Over Slow Internet Connections

MotivationKmerFinder is a program based on K-mer statistics for identifying bacterial species in whole genome data, that as a web server that have been used more than 10.000 times. Kmer-FinderJS is a development of the KmerFinder that benefits from the downsampling of data using a prefix filtering used by KmerFinder, to minimize amount of data that needs to be transferred between the client and the server.\n\nResultsKmerFinderJS replaces the python based hash structure for holding the databases with a true Key-value database. These improvements are shown to lead to a many-fold speed up of species identification with the internet transfer speeds that are realistic to expect today. It is also shown that the method can find the true content of an artificial metagenomic cocktail with no false positives.\n\nAvailabilityThe method is freely available at https://cge.cbs.dtu.dk/services/KmerFinderJS/ and as a source code at https://bitbucket.org/genomicepidemiology/kmerfinderjs\n\nContactcisneros@cbs.dtu.dk\n\nSupplementary informationSupplementary data are available at biorxiv online.

bioinformatics↗