Enhanced Transcriptomics Analysis by Integration with Large-Scale Knowledge Graphs and large language models — Zenodo Data Deposit

This archive contains tKOI (transcriptomics Knowledge-graph-driven Omics Integration) analysis results generated for the manuscript: Gu W, Bellucci G, Peetoom B, Baranzini S. "Enhanced Transcriptomics Analysis by Integration with Large-Scale Knowledge Graphs and Large Language Models." tKOI integrates differential gene expression data into a heterogeneous biological knowledge graph (a pruned subset of SPOKE) through Personalized PageRank (PPR) network propagation. For each input transcriptomic dataset, tKOI generates enrichment statistics — including network-derived effect sizes (β coefficients) and permutation-based empirical FDR values — across 18 node types, including genes, proteins, pathways, diseases, cell types, anatomical structures, compounds, and other functional entities. Each .rda file contains the complete tKOI result object for a single analysis and can be loaded in R using: load("filename.rda"). The result object contains personalized PageRank scores, β coefficients, empirical p-values, FDR-adjusted q-values, and node-level annotations across all 18 supported biological modalities. The deposit is organized into three folders: e-cigarette/ (4 files, ~1.3 GB): tKOI results from compound association analysis of JUUL e-cigarette flavor aerosol exposure (Figure 3A–C). Transcriptomic data from Moshensky et al. (eLife, 2022), a bulk RNA-seq study in mice exposed to four JUUL pod flavors for 3 months. Includes mango_result.rda, mint_result.rda, menthol_result.rda, and crc_result.rda (colorectal cancer comparison dataset). 2. pathway replication/ (3 files, ~1.0 GB): tKOI results from three bulk RNA-seq datasets benchmarking tKOI enrichment against conventional Gene Ontology enrichment analysis (Figure 2C). Includes ra_result.rda (rheumatoid arthritis, PBMCs), pca_result.rda (prostate cancer), and tobacco_result.rda (tobacco JUUL pod exposure). 3. lung atlas/ (40 files, ~13 GB): tKOI results for 40 individual cell clusters from the Human Lung Cell Atlas single-cell RNA-seq dataset (Figure 3D–F). Covers clusters C0–C39 representing diverse lung cell types, including AT1/AT2 alveolar cells, macrophages, T cells, endothelial cells, fibroblasts, basal cells, goblet cells, and transitional club-AT2 (bronchioalveolar stem) cells.