Biology subjects

Mathe, E.

Publications and source records attributed to Mathe, E..

7 recordsLinked to original sources

bioRxiv2026

GEO-BERT: A Retrieval augmented LLM Framework for Intelligent Dataset Retrieval from the Gene Expression Omnibus

The Gene Expression Omnibus (GEO) is a commonly used data repository of high throughput gene expression and other functional genomics data. This platform is globally used by researchers for submitting experimental datasets, accessing these datasets for hypothesis validation, reusing them in data-driven research, generating ideas for new research directions. Despite global usage, searching and identifying relevant and usable datasets in GEO remains challenging with an underlying keyword based search system needing manual intervention for the retrieval of desired datasets. GEO's inconsistent metadata and varying data annotation practices across submissions, variable data quality are contributing to its challenge for building structured query tools. With the recent advancement of Large Language Models (LLMs), it is timely to revisit the GEO datasets and its query system. Here, we present an AI chat-bot, GEO-BERT, leveraging an underlying retrieval augmented generator framework on the GEO metadata with a novel hierarchical hybrid field aware retrieval system to mine the datasets. GEO-BERT is locally installable, available for programmers and project developers (https://github.com/mrumi/GEOBERTtest) and readily usable by regular users on subset of test data. We demonstrated that GEO-BERT enhances the semantic search performance not currently available by any of the leading existing AI-powered chatbots, while improving the performance of GEO's traditional search performance for project development. Given the global usage of GEO and the latest advancement of the large language models, GEO-BERT fills the gap for context-aware extraction of GEO datasets.

bioinformatics↗

bioRxiv2026

Robust Identification of Drug Mechanism of Action Associated Genes Across Heterogeneous Drug Treated Datasets

Identifying genes associated with the mechanism of action (MOA) of anticancer therapies is essential for understanding drug response. Functional genomics approaches, including genome-wide CRISPR knock out screens, enable systematic gene prioritization following treatment; however, findings from individual datasets are typically interpreted after identifying overlapping genes across datasets to confirm biological hypotheses. Integrating such heterogeneous datasets remains challenging due to variability across experimental conditions. Existing approaches can miss biologically relevant genes due to dataset-specific noise, can fail to capture genes exhibiting consistent but moderate effects across studies. Here, we present ConcuRx, a statistical framework for cross-study integration and gene prioritization across heterogeneous CRISPR drug-response screens. This approach produces more stable gene rankings than single-dataset analyses, improving prioritization of known MOA-associated genes and recovering biologically relevant genes that are overlooked by average- and overlap-based approaches. Applied to CRISPR drug-response screens for the synergistic drug pair of ATR and PARP inhibitors, ConcuRx strengthened support for 15 experimentally validated hypotheses collected from published studies, including mechanisms supported by independent studies not included in the ConcuRx analysis. These findings demonstrate that modelling between-study variability improves identification and prioritization of MOA-associated genes and provides a robust strategy for integrating heterogeneous functional genomics data in precision oncology.

bioinformatics↗

bioRxiv2026

CalFluxTools: An R package for analysis of high-throughput calcium oscillation screening data

Whole plate detection of calcium flux signals using fluorescence is a powerful technology for monitoring oscillatory activity of excitable cells in a high throughput fashion that sees broad applications in neurotoxicity and cardiotoxicity research. However, analysis of calcium oscillation profiles can output more than 60 peak kinetic parameters, which makes manual data analysis challenging. We have created the CalFluxTools R package, an automated pipeline for parsing plate data for 384 well plates, generating quality control metrics and plots, and performing data analysis, including T-tests, Z-score analysis, and machine learning prediction of compound toxicity values. CalFluxTools requires a plate map and experimental manifest that includes parameter filtration specifications, as well as data in the ".statall" format. CalFluxTools runs in less than a minute on a Macbook Pro with a 2.6 GHz 6-Core Intel Core i7 processor.

bioinformatics↗

bioRxiv2025

Commercial Saccharomyces cerevisiae baker's yeasts: strain redundancy, genome plasticity, and colonization of the sourdough environment and the human body

Leavening dough is one of the most widespread applications of fermentative yeast and the most common practice for the general public to come into contact with microbial cultures. Saccharomyces cerevisiae is the typical species used for dough making, but the evolutionary origin of strains isolated from dough is mixed. Here, using 49 newly sequenced and 183 previously described isolates from the bakery environment, we show that the traditional strains used in Europe for sourdough making are more closely related to Chinese Mantou sourdough lineages than to commercially used baker yeast strains. Surprisingly, the expansion of these traditional European strains into other human-associated niches, including the human body, has been very limited. This is in stark contrast to the Mixed-origin clade commercial baking yeasts, which consists of only a few globally distributed clonal lineages that dominate the yeast market and recurrently colonize sourdoughs and human hosts. These clonal lineages consistently maintain their ploidy, unique heterozygous chromosomal rearrangements, and stable aneuploidies in addition to several diverse structural variants. In addition to the previously known diploid and tetraploid groups of commercial isolates, we describe a widely distributed stable triploid aneuploid clonal lineage. We show that company practices and global trade help the distribution of these clonal clusters and that these yeasts are characterized by their exclusively mitotic reproduction.

microbiology↗

bioRxiv2024

Finding Salient Multi-Omic Interactomes Relevant to Multiple Biomedical Outcomes using Graph Ensemble Neural Networks

Although multi-omics integration relevant to patient outcome is typically characterized by an analyte interactome, current multi-omic integration methods either (1) model outcome without directly including associations between analytes, (2) model the interactome without directly evaluating the saliency of the model in the context of outcome, or (3) model outcome in a high-dimensional parameter space not suitable for small sample sizes (which are common in multi-omics studies). We introduce Graph Ensemble Neural Network (GENN), a methodology that learns the interactome most predictive of outcome in a low-dimensional parameter space built on complementary attributes for all possible analyte associations (metafeatures). We show that GENN is robust to noise in measurements using a theoretical model, outperforms the predictive performance of existing methods when evaluated on Tegafur drug response in NCI-60 cancer cell line data, and uncovers potentially novel multi-omic mechanisms driving total serum IgE levels in pediatric asthma and patient survival in glioblastomas.

bioinformatics↗

bioRxiv2023

Exploring NCATS In-House Biomedical Data for Evidence-based Drug Repurposing

AbstractDrug repurposing is a strategy for identifying new uses of approved or investigational drugs that are outside the scope of the original medical indication. Even though many repurposed drugs have been found serendipitously in the past, the increasing availability of large volumes of biomedical data has enabled more systemic, data-driven approaches for drug candidate identification. At National Center of Advancing Translational Sciences (NCATS), we invent new methods to generate new data and information publicly available to spur innovation and scientific discovery. In this study, we aimed to explore and demonstrate biomedical data generated and collected via two NCATS research programs, the Toxicology in the 21st Century program (Tox21) and the Biomedical Data Translator (Translator) for the application of drug repurposing. These two programs provide complementary types of biomedical data from uncovering underlying biological mechanisms with bioassay screening data from Tox21 for chemical clustering, to enrich clustered chemicals with scientific evidence mined from the Translator towards drug repurposing. 129 chemical clusters have been generated and three of them have been further investigated for drug repurposing candidate identification, which is detailed as case studies.

bioinformatics↗

bioRxiv2023

Clustering rare diseases within an ontology-enriched knowledge graph

Structured AbstractO_ST_ABSObjectiveC_ST_ABSIdentifying sets of rare diseases with shared aspects of etiology and pathophysiology may enable drug repurposing and/or platform based therapeutic development. Toward that aim, we utilized an integrative knowledge graph-based approach to constructing clusters of rare diseases. Materials and MethodsData on 3,242 rare diseases were extracted from the National Center for Advancing Translational Science (NCATS) Genetic and Rare Diseases Information center (GARD) internal data resources. The rare disease data was enriched with additional biomedical data, including gene and phenotype ontologies, biological pathway data and small molecule-target activity data, to create a knowledge graph (KG). Node embeddings were used to convert nodes into vectors upon which k-means clustering was applied. We validated the disease clusters through semantic similarity and feature enrichment analysis. ResultsA node embedding model was trained on the ontology enriched rare disease KG and k-means clustering was applied to the embedding vectors resulting in 37 disease clusters with a mean size of 87 diseases. We validate the disease clusters quantitatively by looking at semantic similarity of clustered diseases, using the Orphanet Rare Disease Ontology. In addition, the clusters were analyzed for enrichment of associated genes, revealing that the enriched genes within clusters were shown to be highly related. DiscussionWe demonstrate that node embeddings are an effective method for clustering diseases within a heterogenous KG. Semantically similar diseases and relevant enriched genes have been uncovered within the clusters. Connections between disease clusters and approved or investigational drugs are enumerated for follow-up efforts. ConclusionOur study lays out a method for clustering rare diseases using the graph node embeddings. We develop an easy to maintain pipeline that can be updated when new data on rare diseases emerges. The embeddings themselves can be paired with other representation learning methods for other data types, such as drugs, to address other predictive modeling problems. Detailed subnetwork analysis and in-depth review of individual clusters may lead to translatable findings. Future work will focus on incorporation of additional data sources, with a particular focus on common disease data.

bioinformatics↗