Bioinformatics and Genomic Networks

BiomiX automates multiomics data analysis and integration, enabling efficient comparison of two cohorts through MOFA and providing interpretable results using literature and clinical data.

The paper introduces a weighted variance components framework to integrate metagenomic and metabolomic data for pathway analysis, showing improved power while controlling type I error.

The paper develops a framework to identify transcription factors mediating signaling convergence from multiple receptors during cell-cell interactions using single-cell RNA-seq data.

HCNetlas links disease genes to cell types by analyzing cell type-specific gene networks in healthy tissues and comparing them to disease-affected samples.

This paper maps disease relationships using six data dimensions to reveal novel insights beyond traditional classifications, identifying shared features between Type 2 diabetes and Alzheimer's disease.

The paper evaluates ssGSEA's performance on scRNA-seq data, finds it unreliable due to sparsity, and identifies UCell as more robust for pseudobulk analysis.

This paper evaluates multiple gene co-expression network inference methods for Plasmodium falciparum to improve functional annotation of its genes.

The paper introduces mousiPLIER, a mouse pathway-level information extractor model trained on brain RNA-seq data, which reduces dimensionality and identifies aging-related latent variables.

MetalinksDB provides a comprehensive database of metabolite-protein interactions, enabling context-specific analysis of metabolite-mediated cell-cell communication.

PathFinder infers multi-cell signaling pathways and communications using a graph transformer model, identifying novel interactions in AD mouse models.

The paper introduces the CM4AI project, which aims to create AI-ready datasets of human cell architecture using multimodal data from cell lines.

Domain2GO predicts protein domain-GO term associations using statistical resampling and co-annotation patterns, achieving high performance in molecular function and biological process prediction with low computational cost.

This study evaluates GSEA performance on RNA-seq data using TCGA and curated pathways, finding the classic permutation method offers better sensitivity-specificity tradeoffs than complex alternatives.

The paper constructs a tissue-specific protein interaction atlas using proteomic data from 11 human tissues, identifying key drivers of inter-tissue variation and prioritizing disease genes in brain disorders.

The paper integrates scRNA-seq time series data with molecular networks to identify PD-related genes and potential drug targets.

The paper develops classification models to identify host genes crucial for HIV and Sars-CoV-2 life cycles, using graph embeddings and Siamese networks to predict gene interactions.

The paper identifies common differentially expressed genes between cutaneous and uveal melanoma and links two genes to poor prognosis in cutaneous melanoma through computational analysis.

This study reconstructs the metabolic pathways of the amphipod Gammarus fossarum using multi-omics data, revealing organ-specific metabolic profiles and validating enzyme expression through proteomics.

The paper evaluates an erroneous KEGG-derived dataset used in machine learning for predicting metabolite-to-pathway mappings and shows how duplicates inflated model performance.

DMPA is a de novo multi-omics method that discovers signaling pathways and regulatory relationships across omics data, outperforming existing methods in low-sample and zero-inflated datasets.

Decipher identifies context-specific mechanistic insights from single-cell data by building cell signalling networks and uncovering disease-related pathways.

RESET is a single sample gene set testing method that improves accuracy and computational efficiency for detecting differential abundance and correlation patterns in scRNA-seq data.

The paper shows cancer disrupts gene regulatory networks by altering co-expression between genes of unicellular and multicellular origins, with changes more pronounced in advanced tumors.

ProLLM uses a large language model to predict protein-protein interactions by modeling biological pathways as natural language prompts, improving accuracy and generalizability over existing methods.

phuEGO reconstructs active signalling pathways from phosphoproteomics data by combining network propagation and decomposition to improve functional interpretation.

CoVar identifies coordinated regulators driving gene expression changes by analyzing network connectivity in perturbed data, revealing key genes missed by differential expression alone.

The paper introduces Stress Knowledge Map, a resource with two knowledge graphs for plant stress responses, and demonstrates its use in hypothesis generation and analysis.

This study introduces DISHyper, a hypergraph neural network that integrates annotated gene sets to predict cancer genes, identifying 44 novel ones associated with multiple cancers.

This paper introduces a supervised Bayesian factor model to improve multi-omics signature discovery by integrating dimensionality reduction and predictive modeling.

This study introduces the clGENE framework to identify shared molecular mechanisms across diseases by integrating normalization, cosine similarity, and PCA, validated on a pan-cancer dataset and immune cell infiltration analysis.