Biology subjects

Le Cao, K.-A.

Publications and source records attributed to Le Cao, K.-A..

At least 19 recordsLinked to original sources

bioRxiv2026

MINT Block: multivariate integration across independent studies to identify multi-omics signatures

Integrative multi-omics analyses can identify molecular signatures spanning multiple layers of a biological system. However, signatures derived from datasets with small sample sizes may not generalise. Combining data from multiple independent studies can increase the available sample size, but may also introduce unwanted variation arising from study-specific differences. Existing approaches typically address this variation sequentially by correcting the data before downstream model fitting, which can inadvertently distort the biological signals of interest. We introduce MINT Block, a framework that integrates multi-omics data across independent studies while simultaneously accounting for study-specific unwanted variation. MINT Block supports multivariate regression and classification analyses, enabling the identification of generalisable multi-omics signatures that capture relationships between different types of molecules and predict sample phenotypes. We evaluated MINT Block using bulk and single-cell multi-omics datasets from studies of human disease and marine biology. Across these applications, we showed that MINT Block identified stable, interpretable molecular signatures and improves prediction accuracy compared with existing sequential approaches.

bioinformatics↗

bioRxiv2026

No-take marine reserves promote oligotrophic reef bacterioplankton communities across the Great Barrier Reef

Australias Great Barrier Reef is a biodiversity hotspot critical to ocean health, yet it faces increasing threats from climate change and localised impacts requiring effective conservation and management action. Rezoning of the Great Barrier Reef Marine Park in 2004 expanded No-Take Marine Reserves (NTMRs) to restrict extractive activities like fishing and collecting, creating one of the largest networks of marine reserves globally. Benefits like increased biomass of fisheries-targeted species and improved coral community health metrics have been reported, though the effects of zoning on water chemistry and seawater microbiology remain unexplored. Using data from the Great Barrier Reef Microbial Genomics Database, we investigated the structure of seawater microbiomes on 48 offshore reefs within NTMRs and fished reefs. A supervised classification method (MINT sPLS-DA) identified 350 indicator species that predict zoning with [~]71% accuracy (range 58-85%). Microbial communities broadly reflected reef states, with NTMR zones enriched in streamlined microbial oligotrophs (Pelagibacter and SAR86) correlating with higher cover of hard coral, crustose coralline algae, and herbivore fish abundance under lower nutrient conditions. By contrast, fished reefs harbored opportunists (Flavobacteriales, especially UA16, and Pseudomonadales) associating with elevated nutrients and turf algae cover. Co-occurrence networks revealed stronger competitive interactions in fished reefs, where nutrient-responsive taxa may outcompete other microbes, underscoring the need to investigate how these shifts influence reef nutrient cycling and function. Our findings reveal ecosystem-wide effects of marine zoning beyond fish protection, with distinct seawater microbiomes between fished reefs and NTMRs, which will help build decision tools for more targeted reef health monitoring assessments.

microbiology↗

bioRxiv2026

Multivariate integration of histological images and gene expression data: a comparative review

Integrating histological images with gene expression data offers a promising approach for linking tissue morphologies to molecular signatures and improving disease subtyping. However, such integration remains challenging due to the high dimensionality of these datasets, cross-modal heterogeneity, and limited interpretability. Multivariate methods such as Sparse Canonical Correlation Analysis (Sparse CCA), Joint Nonnegative Matrix Factorisation (Joint NMF), and Angle-based Joint and Individual Variation Explained (AJIVE), have been used to address these challenges by reducing dimensionality while identifying features associated with latent factors, thereby enhancing biological interpretability. Despite increasing application in imaging-omics research, systematic comparisons of their methodological properties remain limited. Consequently, users often lack guidance on how to appropriately select these methods in practice, and these approaches are frequently treated as interchangeable despite differing modelling assumptions. Here, we use paired H&E images and gene expression data from breast cancer as a representative case study to examine the methodological characteristics, interpretability, and complementary properties of these integration approaches. Our results show that each method captures distinct yet complementary aspects of the underlying information. Although the biological findings are derived from the TCGA-BRCA datasets, the methodological insights identified here extend more broadly to imaging-omics integration studies. Overall, this comparative review highlights the strengths and limitations of each approach and outlines considerations for future methodological development.

bioinformatics↗

bioRxiv2026

tensorOmics: Data integration for longitudinal omics data using tensor factorisation

Multi-omics studies capture comprehensive molecular profiles across biological layers to understand complex biological processes. A central challenge is integrating information across heterogeneous data types to identify coordinated molecular responses, particularly when measurements are collected longitudinally. Traditional integration methods can be broadly classified as unsupervised (exploring patterns without phenotypic information) or supervised (discriminating between groups or predicting outcomes). These approaches rely predominantly on matrix-based techniques that concatenate or project data into lower-dimensional spaces. However, matrix methods struggle with longitudinal data, as flattening multi-dimensional structures obscures temporal trajectories and violates independence assumptions. Tensor-based methods preserve the natural multi-way structure of longitudinal data but existing approaches are predominantly unsupervised, cannot incorporate phenotypic responses for discriminant analysis, and lack frameworks for integrating multiple omics layers. We present tensorOmics, a comprehensive framework for longitudinal omics analysis using tensor factorisation. The framework encompasses supervised and unsupervised methods for both single-omic (tensor PCA, tensor PLS discriminant analysis) and multi-omic settings (tensor PLS, block tensor PLS, block tensor PLS discriminant analysis). This unified approach captures coordinated responses across biological layers while preserving temporal structure. We validated tensorOmics through three case studies: antibiotic perturbation experiments, anaerobic digestion systems, and fecal microbiota transplantation. These applications demonstrate tensorOmics differentiates treatment groups, captures time-dependent molecular signatures, and reveals multi-layer coordinated responses that cross-sectional methods miss. Author summaryLongitudinal multi-omics studies track molecular changes over time to understand how biological systems respond to treatments, diseases, or environmental shifts. However, analysing these complex datasets presents significant challenges: traditional methods either flatten the time dimension, losing temporal information, or handle only single omics layers without integration. We developed tensorOmics, a comprehensive computational framework that preserves the natural three-way structure of longitudinal data (samples x features x time) while integrating multiple omics layers. Our approach combines tensor decomposition with multi-block analysis, offering five complementary methods that address both exploratory and supervised analyses across single and multi-omic settings. Through three diverse case studies (antibiotic recovery in humans, anaerobic digestion systems, and fecal microbiota transplantation), we demonstrate that tensorOmics successfully identifies biologically meaningful temporal signatures that distinguish treatment groups and reveal coordinated molecular responses across omic layers. The framework handles the unique statistical properties of different omics types and efficiently manages high-dimensional data through tensor-based data compression. tensorOmics is available as an R package, providing researchers with flexible tools to extract interpretable insights from longitudinal multi-omics experiments while properly accounting for repeated measurements and temporal dynamics.

bioinformatics↗

bioRxiv2026

DIVAS: an R package for identifying shared and individual variations of multiomics data

MotivationMultiomics data integration aims to identify biological patterns shared across molecular modalities. Most existing methods detect either jointly shared variation, across all modalities, or individual variation, unique to a single modality, but overlook partially shared variation, shared by only a subset of modalities. This is a critical limitation, because many biological mechanisms manifest in some but not all molecular modalities. ResultsWe present an open-source R package implementing DIVAS, a framework for systematically identifying jointly shared, partially shared and individual variations across multiple data types. DIVAS combines angle-based subspace analysis with inference through rotational bootstrap, hierarchically searching all combinations of modalities to decompose multiomics data into interpretable components with scores and loadings. In simulations with a known sharing structure, DIVAS recovered every component across a wide range of noise levels, whereas AJIVE and MOFA+ did not. Applied to multi-modal COVID-19 data, it reveals partially shared immune and metabolic dysregulation patterns underpinning disease severity that conventional approaches would miss. Availability and implementationDIVAS is available as an R package on GitHub (https://github.com/ByronSyun/DIVAS), with documentation and vignettes at https://byronsyun.github.io/DIVAS/. A step-by-step vignette for the COVID-19 case study is available at https://byronsyun.github.io/DIVAS_COVID19_CaseStudy/.

bioinformatics↗

bioRxiv2025

A Practical Resource for Multi-Omics Data Integration in Microbial Systems

The increasing availability of microbial multi-omics datasets has created new opportunities to explore complex biological systems. However, exploration remains limited by the lack of accessible, reproducible workflows that integrate multiple omics layers and deliver easily interpretable visualisations of functional and pathway-level insights. Here, we present an R-based workflow for integrated analysis and network-based pathway visualisation of microbial multi-omic data. The workflow enables microbiologists to analyse transcriptomic, proteomic, and metabolomic datasets either individually or in combination, apply univariate and multivariate approaches for biomarker discovery, and generate easily interpretable visualisations of functional and pathway-level signatures. Implemented as multi-step R Markdown, it leverages widely-used open-source tools, including mixOmics for biomarker identification and omics integration and clusterProfiler for pathway and functional enrichment analyses, with a new network-based integration and visualisation. Its flexible design supports a range of experimental structures and facilitates comparisons across strains, omics layers, and conditions, making it suitable for researchers with limited computational expertise. We demonstrate its utility using a publicly available Streptococcus pyogenes dataset, revealing both shared and strain-specific functional responses to human serum. This workflow provides a comprehensive and adaptable framework for systematic multi-omics analysis, improving accessibility and reproducibility and facilitating functional interpretation of microbial responses to diverse environments. Data summaryThe code for this workflow is available on GitHub (https://github.com/warasinee/Multiomics_Case_Study). Datasets from our previously published study (1) were used to showcase the functionality and practical utility of the workflow. The multi-omics Streptococcus pyogenes dataset used in this study is available in the following public repositories: Gene Expression Omnibus (GSE152821; GSE152822; GSE152823; GSE152824, GSE152826), Proteomics Identifications Database (PXD020863), and MetaobLights (MTBLS2324) (1). Impact StatementHigh-throughput omics technologies are transforming our understanding of how microbes adapt to diverse environments and cause disease. The integration of diverse omics layers at a systems level, combining transcriptomics, proteomics, and metabolomics data to identify signature molecules, pathways, and their interactions, remains challenging. Here, we present an R-based bioinformatic workflow designed for microbiology research, which connects existing tools and customised functions to streamline data integration and interpretation. The workflow links biomarkers to functional pathways, visualises results in an interactive network context, and is designed for flexibility and reproducibility. This practical resource lowers technical barriers to microbial multi-omics analysis, providing user-friendly access for dataset exploration and integration and supporting interpretation of system-level microbial adaptation in environmental, clinical, and industrial contexts.

systems biology↗

bioRxiv2025

phylobar: an R package for multiresolution compositional barplots in omics studies

SummaryStacked barplots, though widely used in microbiome studies, can obscure important patterns in microbiome data. They omit rare taxa and can mask shifts that emerge at finer taxonomic levels. To address this issue, we introduce phylobar, an R package that interactively links stacked barplots with overview phylogenetic or taxonomic hierarchies. The interface allows users to collapse or expand subtrees, paint color palettes interactively, and search for specific taxa. This allows comparison across taxonomic resolutions that are hidden in static overviews. phylobar works with any omics data with hierarchical organization, including cell type hierarchies, as we demonstrate in a case study of immune cell composition in COVID-19 patients. Availability and Implementationphylobar is available as an R package on GitHub. It uses the htmlwidgets library to link interactive D3 visualizations with R. The interactive plots can be embedded within R Markdown or Quarto notebooks, and views can be exported as vector graphics files. The package is open source and documented at https://mkdiro-O.github.io/phylobar.

bioinformatics↗

bioRxiv2025

CellDiffusion: a generative model to annotate single-cell and spatial RNA-seq using bulk references

Annotating single-cell and spatial RNA-seq data can be greatly enhanced by leveraging bulk RNA-seq, which remains a cost-effective and well-established benchmark for characterising transcriptional activity in immune cell populations. However, a major technical hurdle lies in the contrasting properties of these data types: single-cell and spatial data are inherently sparse due to its cell-level sampling scheme, leading to much lower sequencing depth compared to bulk RNA-seq. We developed CellDiffusion, a generative machine learning (ML) tool that bridges this gap. CellDiffusion generates realistic virtual cells to augment the sparse single-cell and spatial data, improving signals and the representation of rare cell types. The augmented data are more comparable to bulk references, increasing the accuracy of cell type annotation using bulk references and automated ML classifiers. We benchmarked CellDiffusion on single-cell and spatial datasets from human peripheral blood samples, white adipose tissues, and breast tumours. Our method significantly outperforms state-of-the-art methods such as SingleR, Seurat, and scVI. In addition, CellDiffusion provides critical biological insights, including the identification of novel cell subtypes and their function during cell state transition; the discovery of new marker genes for tissue-resident immune cells, revealing their functional shifts in myeloid populations; and the accurate characterisation of cell subtypes in spatial transcriptomics to decipher tumour microenvironment.

bioinformatics↗

bioRxiv2025

Integrated multi-omics reveals coordinated Staphylococcus aureus metabolic, iron transport and stress responses to human serum

Bloodstream infections caused by Staphylococcus aureus remain a leading cause of mortality worldwide. Our understanding of S. aureus survival and persistence in human serum, a cell-free fraction of blood hostile for bacteria, is still limited. Here, we applied multivariate data integration methods and network analysis to a multi-omic dataset generated from five clinically prevalent S. aureus genotypes exposed to human serum. We observed and then confirmed using isogenic mutants, the significant roles of gapdhB, sucA, sirA, sstD, and perR in bacterial survival in serum. These data show that metabolic versatility in carbon source usage, iron transport and resistance to oxidative stress are interlinked and central to S. aureus fitness in serum, representing potential S. aureus vulnerabilities that could be exploited therapeutically. IMPORTANCEBloodstream infections caused by Staphylococcus aureus are associated with mortality rates of up to 30%. However, the molecular mechanisms that enable this pathogen to survive in human serum, a nutrient-limited and immunologically hostile environment remain poorly understood. By integrating multi-omic data from five clinically relevant S. aureus genotypes and validating key signatures using mutants, we identified conserved genetic determinants critical for bacterial survival in serum. Our findings highlight the interconnected roles of carbohydrate metabolic flexibility, iron acquisition, and oxidative stress resistance in shaping S. aureus adaptation to serum. This work advances our understanding of microbial strategies to survive in the bloodstream and demonstrates the potential of multi-omic integration to uncover therapeutic vulnerabilities in bacterial pathogens.

microbiology↗

bioRxiv2025

Oral microbial signatures of head and neck cancer patients highlight diverse longitudinal patterns of oral mucositis severity

BackgroundOral mucositis is a painful complication commonly observed in head and neck cancer patients receiving cancer treatment. Emerging evidence suggests that changes in the oral microbiome can contribute to oral mucositis development, making microbial signatures potential targets for therapeutic inter-ventions. This study aimed to: (1) characterize longitudinal microbial patterns of oral mucositis severity among head and neck cancer patients; (2) determine clinically relevant patient clusters based on oral mucositis severity trajectories; and (3) identify microbial signatures specific to these clusters. ResultsWe derived a calibrated oral mucositis score by applying non-negative sparse principal component analysis to seven oral mucositis related symptom ratings, using longitudinal microbiome data from 140 head and neck cancer patients. Functional data analysis and hierarchical clustering identified three distinct patient clusters with differing microbial trajectories of oral mucositis progression. One cluster exhibited patients with a rapid increase in oral mucositis severity following treatment initiation, while the other clusters displayed more gradual increase. Demographic comparisons revealed significant differences in age and weight distributions between clusters, with older, lighter patients more common in clusters experiencing more gradual oral mucositis progression. Partial least squares knockoff analysis identified cluster-specific microbial signatures: notably, Prevotella spp. positively associated with calibrated oral mucositis score across all clusters, while Alloprevotella (Alloprevotella0302) was significantly enriched only in patients experiencing rapid oral mucositis progression. Conversely, genera associated with oral health, including Haemophilus, Rothia, and Actinomyces, were negatively correlated with calibrated oral mucositis score. ConclusionsDistinct trajectories of oral mucositis scores in head and neck cancer patients are linked to specific oral microbial profiles and demographic factors. The identification of cluster-specific microbial profiles highlights the potential for microbiome-targeted interventions to manage oral mucositis severity. While most taxa were cluster-specific, Prevotella consistently ranked among the top taxa positively associated with the calibirated oral mucositis score across clusters, suggesting it may not differentiate between patient groups but rather reflects overall disease severity.

bioinformatics↗

bioRxiv2025

Multi-omics integration reveals sex-based differences in the circulating extracellular vesicle lipidome and miRNome of alcohol use disorder patients

Integrated multi-omics and extracellular vesicle (EV) analysis are emerging as powerful, complementary strategies for biomarker discovery. These approaches offer promising tools to enhance early detection, diagnosis, and treatment of alcohol use disorder (AUD). Here we applied an integrated miRNomic and lipidomic approach to analyze plasma EVs from AUD patients and controls of both sexes to gain a comprehensive understanding of the underlying molecular mechanisms. We identified an AUD signature with predictive potential for diagnostic applications. Individual features (e.g., hsa-miR-99b-3p, hsa-miR-556-5p, Cer_NDS-d39:1, and PI18:0_18:2) represented important components; however, the strength of this signature lay in the combined profile rather than isolated markers. We also revealed an AUD-sex signature that provided insight into how biological responses to alcohol differ between females and males (including features such as hsa-miR-1301-3p and PC39:4), which also underscored the power of multi-omic integration. The individual miRNome approach also revealed an opposite functional alteration by sex in various alcohol related systems, such as pathways associated with immunity, oxidative stress, and autophagy. An open-access Shiny web application (https://carpercle.shinyapps.io/SexEVEthOmics/) accompanies this study, providing interactive access to the complete dataset and additional analyses for customized exploration. Together, our findings underscore the added value of multi-omics integration in identifying clinically relevant molecular signatures of disease; this sex-informed approach offers a promising path toward more personalized diagnostic tools and therapeutic strategies in AUD.

bioinformatics↗

bioRxiv2025

Trophic and temporal dynamics of macrophage biology in human inner ear organogenesis

Recent single-cell transcriptomic approaches are uncovering the breadth and depth of cell diversity within the mammalian inner ear. Macrophages, detected from gestational week 7 in the human inner ear, persist into adulthood, and yet remain poorly understood in terms of their origin and function. Using self-generated and public scRNA-seq data, we identify seven distinct macrophage subtypes spanning fetal weeks 7.5 to 16.4 and adulthood. Each macrophage subtype is linked to specific developmental stages and displays a unique gene expression profile. These findings corroborate earlier histological evidence of resident and non-resident macrophages in both the developing and adult human cochlea. We also show that the human inner ear is seeded by macrophages from both embryonic and more definitive sources, corroborating studies in mouse. By analyzing ligand-receptor interactions, we highlight potential macrophage contributions to inner ear organogenesis. This research provides new insights into the diverse roles of human inner ear macrophages.

bioinformatics↗

bioRxiv2025

NeighbourNet: Scalable cell-specific co-expression networks for granular regulatory pattern discovery

Gene regulatory networks (GRNs) provide a fundamental framework for understanding the molecular mechanisms that govern gene expression. Advances in single-cell RNA sequencing (scRNA-seq) have enabled GRN inference at cellular resolution; however, most existing approaches rely on predefined clusters or cell states, implicitly assuming static regulatory programs and potentially missing subtle, dynamic variation in regulation across individual cells. To address these limitations, we introduce NeighbourNet (NNet), a method that constructs cell-specific co-expression networks. NNet first applies principal component analysis to embed gene expression into a low-dimensional space, followed by local regression within each cells k-nearest neighbourhood (KNN) to quantify co-expression. This approach improves computational efficiency and stabilises co-expression estimates, mitigating challenges posed by small sample sizes in KNN regression and the inherent noise and sparsity of scRNA-seq data. Beyond co-expression, NNet supports scalable downstream analyses, including (i) clustering and aggregating cell-specific networks into meta-networks that capture primary co-expression patterns, and (ii) integrating prior knowledge to annotate co-expression and infer active signalling interactions at the individual cell level. All functional modules of NNet are implemented with an efficient algorithm that enables application to large-scale single-cell datasets. We demonstrate NNets effectiveness through three case studies on transcription factor activity prediction, early haematopoiesis, and tumour microenvironments. Provided as an R package, NNet offers a novel framework for exploring cellular variation in co-expression and integrates seamlessly with existing single-cell analysis workflows.

bioinformatics↗

bioRxiv2025

{Phi}-Space ST: a platform-agnostic method to identify cell states inspatial transcriptomics studies

We introduce {Phi}-Space ST, a platform-agnostic method to identify continuous cell states in spatial transcriptomics (ST) data using multiple scRNA-seq references. For ST with supercellular resolution, {Phi}-Space ST achieves interpretable cell type deconvolution with significantly faster computation. For subcellular resolution, {Phi}-Space ST annotates cell states without cell segmentation, leading to highly insightful spatial niche identification. {Phi}-Space ST harmonises annotations derived from multiple scRNA-seq references, and provides interpretable characterisations of disease cell states by leveraging healthy references. We validate {Phi}-Space ST in three case studies involving CosMx, Visium and Stereo-seq platforms for various cancer tissues. Our method revealed niche-specific enriched cell types and distinct cell type co-presence patterns that distinguish tumour from non-tumour tissue regions. These findings highlight the potential of {Phi}-Space ST as a robust and scalable tool for ST data analysis for understanding complex tissues and pathologies.

bioinformatics↗

bioRxiv2024

Global picoplankton biogeography revealed by metagenomic and climatic data integration

Microbial plankton play fundamental roles in biogeochemical cycles, driving nutrient cycling that influences the global climate and supports life on Earth. Picoplankton are the smallest and most abundant planktonic organisms. The distribution and ecology of these organisms is determined by environmental factors and their biogeography is largely shaped by basin-scale patterns of physicochemical composition of ocean waters. The increased availability of high-throughput sequencing data of microbial communities has enabled the description of how the global oceans are partitioned into distinct microbial biogeographical provinces. However, the key attributes associated with such provinces are still unclear. Here we present a model of picoplankton biogeography based on 1454 metagenomes from multiple sampling consortia, resulting in the largest integrated surface ocean metagenome analysis to date. We identify ten distinct groups based on metagenomic dissimilarity, divided into three categories: polar (Arctic and Antarctic), temperate (coastal temperate, temperate/subtropical transition, oceanic temperate, Mediterranean-like) and tropical (tropical low nutrient, tropical high nutrient, subtropical oceanic gyres). Using machine learning and omics data integration techniques, we predict province areas across the surface oceans and describe their environmental, taxonomical, and functional features. We quantify the relationship between environmental factors and each biogeographical province, identify their main representative taxa and the importance of carbon degradation and antimicrobial resistance pathways in functional community composition, and discuss implications for establishing a model for global picoplankton biogeography.

ecology↗

bioRxiv2024

Semisynthetic Simulation for Microbiome Data Analysis

High-throughput sequencing data lie at the heart of modern microbiome research. Effective analysis of these data requires careful preprocessing, modeling, and interpretation to detect subtle signals and avoid spurious associations. In this review, we discuss how simulation can serve as a sandbox to test candidate approaches, creating a setting that mimics real data while providing ground truth. This is particularly valuable for power analysis, methods benchmarking, and reliability analysis. We explain the probability, multivariate analysis, and regression concepts behind modern simulators and how different implementations make trade-offs between generality, faithfulness, and controllability. Recognizing that all simulators only approximate reality, we review methods to evaluate how accurately they reflect key properties. We also present case studies demonstrating the value of simulation in differential abundance testing, dimensionality reduction, network analysis, and data integration. Code for these examples is available in an online tutorial (https://go.wisc.edu/8994yz) that can be easily adapted to new problem settings.

bioinformatics↗

bioRxiv2024

PLSKO: a robust knockoff generator to control false discovery rate in omics variable selection

The knockoff framework, combined with variable selection procedure, controls false discovery rate (FDR) without the need for calculating p-values. Hence, it presents an attractive alternative to differential expression analysis of high-throughput biological data. However, current knockoff variable generators make strong assumptions or insufficient approximations that lead to FDR inflation when applied to biological data. We propose Partial Least Squares Knockoff (PLSKO), an efficient and assumption-free knockoff generator that is robust to varying types of biological omics data. We compare PLSKO with a wide range of existing methods. In simulation studies, we show that PLSKO is the only method that controls FDR with sufficient statistical power in complex non-linear cases. In semi-simulation studies based on real data, we show that PLSKO generates valid knockoff variables for different types of biological data, including RNA-seq, proteomics, metabolomics and microbiome. In preeclampsia multi-omics case studies, we combined PLSKO with Aggregation Knockoff to address the randomness of knockoffs and improve power, and show that our method is able to select variables that are biologically relevant.

bioinformatics↗

bioRxiv2024

A flexible framework for minimal biomarker signature discovery from clinical omics studies without library size normalisation

Application of transcriptomics, proteomics and metabolomics technologies to clinical cohorts has uncovered a variety of signatures for predicting disease. Many of these signatures require the full omics data for evaluation on unseen samples, either explicitly or implicitly through library size normalisation. Translation to low-cost point-of-care tests requires development of signatures which measure as few analytes as possible without relying on direct measurement of library size. To achieve this, we have developed a feature selection method (Forward Selection-Partial Least Squares) which generates minimal disease signatures from high-dimensional omics datasets with applicability to continuous, binary or multi-class outcomes. Through extensive benchmarking, we show that FS-PLS has comparable performance to commonly used signature discovery methods while delivering signatures which are an order of magnitude smaller. We show that FS-PLS can be used to select features predictive of library size, and that these features can be used to normalize unseen samples, meaning that the features in the complete model can be measured in isolation for making new predictions. By enabling discovery of small, high-performance signatures, FS-PLS addresses an important impediment for the further development of precision medical care.

bioinformatics↗