Biology subjects

Payne, P.

Publications and source records attributed to Payne, P..

12 recordsLinked to original sources

bioRxiv2026

Batch Effect Correction to Enhance Utility of a Publicly Available ALS Proteomics Dataset

Large, publicly available databases of patient data have been successful in identifying associations between neurodegenerative diseases and biological features that might otherwise go unnoticed. The Answer ALS Consortium is one such database that has generated a rich repository of multi-omic data from people with amyotrophic lateral sclerosis (ALS). However, large studies can suffer from batch effects that obscure genuine biological signal and prevent experiment reproducibility. In this study, we identified a batch effect in the Answer ALS 1-2 proteomic dataset (508 samples x 3,643 proteins) and corrected the effect using both the internal controls provided by Answer ALS and the batch effect-correction algorithm (BECA) ComBat. We identified a batch effect in the Answer ALS Dataset 1-2 proteomic dataset that, uncorrected, may obscure biology. Proteins MOS and H4 were outliers of expression in the dataset that correlated with the batch effect. Removing MOS and H4 did not correct the effect fully but did reveal new principal component (PC) contributing proteins related to motor neurons and potentially ALS: AINX, NFL, and NFM. The batch differential controls (BDCs) provided by Answer ALS were used to correct samples in their respective batches. ComBat, a well-characterized BECA, was applied to the dataset both before (ComBat-BDC) and after (BDC-ComBat) correction with the BDCs. BDC-ComBat was the only method that successfully reduced batch effect across the entire dataset. This method also preserved four of the top six PC-contributing proteins from the original dataset, including three neuronal proteins, suggesting that the corrected dataset recapitulates biology as well as the variability of the original dataset. Altogether, we produced a version of the Answer ALS proteomic dataset (354 samples by 3,641 proteins) with globally reduced batch effect and protein-level reduced batch effect. This version of the dataset recapitulates the individual protein variability and important neuronal proteins of the original. We also created a workflow for identifying and correcting batch effects that can be applied to other large data repositories.

bioinformatics↗

bioRxiv2026

Interpreting Omics Data Analysis with Large Language Models for Disease Target and Drug Discovery

In biomedical scientific discovery, synthesizing prior knowledge from the literature is an essential component of interpreting numerical omics data analyses for disease target identification and drug discovery. Large language models (LLMs) alone can rapidly retrieve disease mechanisms from biomedical text, but text-only outputs are general and unreliable for target and drug prioritization without cohort-specific quantitative evidence. Herein, we propose a provenance-aware Text-to-Target framework that couples schema-constrained multi-model LLM retrieval with numeric omics data analysis. The key design is a modality-aware fusion step: candidates are partitioned into overlap-supported anchors, retrieval-only hidden hubs, and network-emergent novelty nodes, then propagated into staged hypothesis and strategy generation under topology constraints. We evaluate the model in Alzheimers disease (AD) and pancreatic ductal adenocarcinoma (PDAC). In PDAC, the workflow produced a balanced 75-gene candidate universe and a 23-strategy portfolio, with significant DepMap support at both target level and strategy level. In AD, stricter candidate controls yielded a compact 34-gene universe and 14 strategies; under an expanded CRISPRbrain registry, both target-level axes were significant, with strong strategy-level enrichment. Across both diseases, final strategies preserved full provenance closure to the candidate pool, enabling end-to-end auditability from retrieval artifacts to validation outputs. These results support a transferable discovery architecture in which omics evidence constrains biological activity, LLM retrieval expands mechanistic search space, and network-aware fusion preserves interpretability. The framework provides a reproducible basis for dual-disease target prioritization and motivates continuous literature-mechanism concordance with agentic evidence-refresh loops.

bioinformatics↗

bioRxiv2025

CRISPR herd immunity against a transducing phage underlies adaptive homologous recombination in resting bacteria

Homologous recombination (HR) is ubiquitous across evolution, driving adaptation by reshuffling standing genetic variation. Although bacteria lack meiotic recombination, HR extensively shapes their genomes. However, the mechanisms and ecological conditions sustaining frequent HR in bacteria remain unclear. Using Escherichia coli, we reveal how frequent recombination emerges from herd immunity to a generalized transducing phage. Herd immunity-established here via CRISPR immunity-maintains genetic polymorphism and enables stable host-phage coexistence, thereby promoting genome-wide gene flow and accelerating adaptation through recombination up to two orders of magnitude relative to de novo mutations. Notably, we show that recombination occurs in stationary phase and is mediated by RecG, which has been previously reported to be regulated by the stringent response - a bacterial reaction to nutrient deprivation and other stress conditions. Bacterial herd immunity thus fulfills an unexpected role of promoting adaptation by HR. This mechanism helps explain the enigmatic high rates of HR across bacterial populations, clarifies how bacteria adapt as resources wane, and suggests a broader evolutionary role for bacterial immune systems beyond individual defense.

evolutionary biology↗

bioRxiv2025

SnakeAltPromoter Facilitates Differential Alternative Promoter Analysis

BackgroundAlternative promoter usage regulates isoform diversity in mammals, playing critical roles in development, disease, and cellular reprogramming. While Cap Analysis of Gene Expression (CAGE) enables precise transcription start site mapping, its high cost and limited coverage hinder scalability. In contrast, RNA-seq is abundant across biological contexts; several algorithms (ProActiv, Salmon, DEXSeq) infer promoter activity from these data, yet no unified, reproducible framework exists to execute, benchmark, and compare them or to scale alternative promoter analyses across large compendia. ResultsWe developed snakeAltPromoter, an end-to-end Snakemake workflow that ingests raw FASTQ files, performs quality control and alignment, quantifies promoter activity using three complementary strategies (junction-based, transcript-based, and first-exon-based), classifies promoters into major, minor/alternative, and inactive categories, and conducts both differential promoter activity and usage analysis. Crucially, snakeAltPromoter integrates a systematic benchmarking module against matched CAGE profiles to reveal method-specific strengths and limitations. ProActiv showed the highest concordance with CAGE in promoter classification, activity and differential analysis, Salmon was robust at low coverage and intronless cases. Overall, the complete workflow recovered a majority of CAGE-validated active promoters and processed a 50 M-read RNA-seq sample in around 2h on a 32-core node, demonstrating both accuracy and scalability. ConclusionssnakeAltPromoter is, to our knowledge, the first reproducible framework that pairs comparative method evaluation with scalable differential alternative promoter analysis. It provides concrete guidance for method selection under different experimental scenarios and enables high-throughput mining of promoter-level regulation from public RNA-seq repositories. Code and example data are freely available at https://github.com/YidanSunResearchLab/SnakeAltPromoter.git.

bioinformatics↗

bioRxiv2025

OmniCellAgent: Towards AI Co-Scientists for Scientific Discovery in Precision Medicine

Real-world biomedical scientific discovery operates as an iterative lifecycle integrating four core pillars: the targeted identification and analysis of question-specific omics datasets; the context-aware interpretation of molecular data using rich biomedical prior knowledge database; comprehensive literature reviews; and the subjective creativity and expert intuition of human scientists. Together, these pillars drive robust evidence synthesis and novel hypothesis generation. The recently reported AI agents support automated omics analysis and literature review, which typically require users to predefine and curate disease-specific datasets, which is a process that remains challenging and time-consuming. In this study, we present OmniCellAgent, a novel multi-agent AI framework built on large-scale single-cell RNA sequencing (scRNA-seq) datasets, and can autonomously retrieve and analyze disease and control-related scRNAseq datasets of diverse cell types across tissues and conditions. Moreover, it incorporates a biomedical prior knowledge and literature review agents, and disease domain-specific expert agents to systematically annotate omic data-derived targets. By aggregating evidence across agents, the framework generates structured analytical reports and potential scientific hypotheses. We evaluated OmniCellAgent across multiple disease settings, demonstrating its ability to identify relevant datasets, generate omic data analysis results, and produce structured reports and scientific hypotheses. Our findings demonstrate that multi-agent AI systems lower the technical barriers to omics-driven research, thereby accelerating scientific discovery and hypothesis generation in biomedical research and precision medicine. The source code is publicly available at: https://github.com/FuhaiLiAiLab/OmniCellAgent.

bioinformatics↗

bioRxiv2024

Multimodal normative modeling in Alzheimer Disease with introspective variational autoencoders

Normative models in neuroimaging learn patterns of healthy brain distributions to identify deviations in disease subjects, such as those with Alzheimers Disease (AD). This study addresses two key limitations of variational autoencoder (VAE)-based normative models: (1) VAEs often struggle to accurately model healthy control distributions, resulting in high reconstruction errors and false positives, and (2) traditional multimodal aggregation methods, like Product-of-Experts (PoE) and Mixture-of-Experts (MoE), can produce uninformative latent representations. To overcome these challenges, we developed a multimodal introspective VAE that enhances normative modeling by achieving more precise representations of healthy anatomy in both the latent space and reconstructions. Additionally, we implemented a Mixture-of-Product-of-Experts (MOPOE) approach, leveraging the strengths of PoE and MoE to efficiently aggregate multimodal information and improve abnormality detection in the latent space. Using multimodal neuroimaging biomarkers from the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset, our proposed multimodal introspective VAE demonstrated superior reconstruction of healthy controls and outperformed baseline methods in detecting outliers. Deviations calculated in the aggregated latent space effectively integrated complementary information from multiple modalities, leading to higher likelihood ratios. The model exhibited strong performance in Out-of-Distribution (OOD) detection, achieving clear separation between control and disease cohorts. Additionally, Z-score deviations in specific latent dimensions were mapped to feature-space abnormalities, enabling interpretable identification of brain regions associated with AD pathology.

neuroscience↗

bioRxiv2024

mosGraphFlow: a novel integrative graph AI model mining disease targets from multi-omic data

Multi-omic data can better characterize complex cellular signaling pathways from multiple views compared to individual omic data. However, integrative multi-omic data analysis to rank key disease biomarkers and infer core signaling pathways remains an open problem. In this study, our novel contributions are that we developed a novel graph AI model, mosGraphFlow, for analyzing multi-omic signaling graphs (mosGraphs), 2) analyzed multi-omic mosGraph datasets of AD, and 3) identified, visualized and evaluated a set of AD associated signaling biomarkers and network. The comparison results show that the proposed model not only achieves the best classification accuracy but also identifies important AD disease biomarkers and signaling interactions. Moreover, the signaling sources are highlighted at specific omic levels to facilitate the understanding of the pathogenesis of AD. The proposed model can also be applied and expanded for other studies using multi-omic data. Model code is accessible via GitHub: https://github.com/FuhaiLiAiLab/mosGraphFlow

bioinformatics↗

bioRxiv2024

mosGraphGPT: a foundation model for multi-omic signaling graphs using generative AI

Generative pretrained models represent a significant advancement in natural language processing and computer vision, which can generate coherent and contextually relevant content based on the pre-training on large general datasets and fine-tune for specific tasks. Building foundation models using large scale omic data is promising to decode and understand the complex signaling language patterns within cells. Different from existing foundation models of omic data, we build a foundation model, mosGraphGPT, for multi-omic signaling (mos) graphs, in which the multi-omic data was integrated and interpreted using a multi-level signaling graph. The model was pretrained using multi-omic data of cancers in The Cancer Genome Atlas (TCGA), and fine-turned for multi-omic data of Alzheimers Disease (AD). The experimental evaluation results showed that the model can not only improve the disease classification accuracy, but also is interpretable by uncovering disease targets and signaling interactions. And the model code are uploaded via GitHub with link: https://github.com/mosGraph/mosGraphGPT

bioinformatics↗

bioRxiv2023

Normative Modeling Using Multimodal Variational Autoencoders to Identify Individual Heterogeneity in Alzheimer Disease

Structured AbstractO_ST_ABSINTRODUCTIONC_ST_ABSPrevious studies have applied normative modeling on a single neuroimaging modality to investigate Alzheimer Disease (AD) heterogeneity. We employed a deep learning-based multimodal normative framework to analyze individual-level variation across ATN (amyloid-tau-neurodegeneration) imaging biomarkers. METHODSWe selected cross-sectional discovery (n = 665) and replication cohorts (n = 430) with available T1-weighted MRI, amyloid and tau PET. Normative modeling estimated individual-level abnormal deviations in amyloid-positive individuals compared to amyloid-negative controls. Regional abnormality patterns were mapped at different clinical group levels to assess intra-group heterogeneity. An individual-level disease severity index (DSI) was calculated using both the spatial extent and magnitude of abnormal deviations across ATN. RESULTSGreater intra-group heterogeneity in ATN abnormality patterns was observed in more severe clinical stages of AD. Higher DSI was associated with worse cognitive function and increased risk of disease progression. DISCUSSIONSubject-specific abnormality maps across ATN reveal the heterogeneous impact of AD on the brain.

neuroscience↗

bioRxiv2023

Cryo-electron tomography of enterovirus cell entry and endosome escape

Enveloped viruses deliver their genomes into the cell cytoplasm by membrane fusion; in contrast, membrane penetration by non-enveloped viruses is more diverse and less well understood. Enteroviruses, one of the largest groups of non-enveloped viruses, cause diseases ranging from the common cold to life-threatening encephalitis. To initiate infection, most enteroviruses enter cells by endocytosis. However, how enterovirus particles or RNA genomes cross the endosome membrane into the cytoplasm remains unknown. Here we used cryo-electron tomography of infected cells to show that endosomes containing rhinovirus 2, echovirus 18, echovirus 30, or enterovirus 71 deform, rupture, and release their content into the cytoplasm. Blocking endosome acidification with bafilomycin A1 reduced the number of enterovirus particles that released their genomes in endosomes, but did not prevent them from reaching the cytoplasm. Inhibiting N-WASP-mediated post-endocytic membrane remodeling with wiskostatin promoted abortive enterovirus genome release in endosomes. We show that the rupture of endosomes also occurs in uninfected cells. In summary, our results indicate that cellular membrane remodeling disrupts enterovirus-containing endosomes and thus releases the virus genomes and particles into the cytoplasm. Since the studied enteroviruses employ different receptors for cell entry but are all delivered into the cytoplasm by cell-mediated endosome disruption, it is possible that many other enteroviruses utilize endosome rupture to infect cells.

microbiology↗

bioRxiv2019

Sympatric ecological divergence with coevolution of niche preference

Reinforcement, the increase of assortative mating driven by selection against unfit hybrids, is conditional on pre-existing divergence. Yet, for ecological divergence to precede the evolution of assortment, strict symmetries between fitnesses in niches must hold, and/or there must be low gene flow between the nascent species. It has thus been argued that conditions favouring sympatric speciation are rarely met in nature. Indeed, we show that under disruptive selection, violating symmetries in niche sizes and increasing strength of the trade-off in selection between the niches quickly leads to loss of genetic variation, instead of evolution of specialists. The region of the parameter space where polymorphism is maintained further narrows with increasing number of loci encoding the diverging trait and the rate of recombination between them. Yet, evolvable assortment and pre-existing assortment both substantially broaden the parameter space within which polymorphism is maintained. Notably, pre-existing niche preference speeds up further increase of assortment, thus facilitating reinforcement in the later phases of speciation. We conclude that in order for sympatric ecological divergence to occur, niche preference must co-evolve throughout the divergence process. Even if populations come into secondary contact, having diverged in isolation, niche preference substantially broadens the conditions for coexistence in sympatry and completion of the speciation process.

evolutionary biology↗

bioRxiv2017

CRISPR-based Herd Immunity Limits Phage Epidemics in Bacterial Populations

Pathogens are a driving force in evolution, giving rise to a diversity of host immune defenses. In order for a pathogen to spread in a population a sufficient number of its members must be susceptible to infection, as resistant individuals can prevent the spread of a pathogen among susceptible hosts in a process known as herd immunity. While herd immunity has been extensively studied in vertebrate populations, little is known about its role, if any, in the dynamics between bacteria and their phage pathogens. Here we explore the dynamics of T7 phage epidemics in structured and unstructured Escherichia coli populations consisting of differing mixtures of susceptible and resistant individuals harboring CRISPR immunity to the phage. Using both experiments and mathematical modelling we describe the conditions under which herd immunity arises in bacterial populations. Notably, the effects of herd immunity depend strongly on the presence of spatial structure in the population, the bacterial growth rate, and phage replication rate. The results of our model can apply to other host-pathogen systems to determine the herd immunity threshold from the relative speed of an epidemic wave in partially resistant populations. In addition, our findings suggest that herd immunity plays an important role in bacterial communities, as seen in other host-pathogen systems, allowing for stable coexistence of bacteria and their phages and the maintenance of polymorphism in bacterial immunity.

evolutionary biology↗