Inhaltsangabe: :: Library Catalog

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Tosaki, Taisei, Oshima, Sho, Okamoto, Yuji
Format:	Recurso digital
Sprache:
Veröffentlicht:	Zenodo 2025
Online-Zugang:	https://doi.org/10.5281/zenodo.15496012
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Inhaltsangabe:

<h3>This data is based on a simple Bayesian network from publicly available data on gene expression levels TCGA and LINCS. For more information, please refer to the following.</h3> Producer name withheld due to blind review in progress. <h3> ## Data</h3> ### 1. GRN Dataset & Preprocessing   All graph data objects live under `data/GRN_dataset/` and are generated by the preprocessing script: ```plaintext └── data/     └── GRN_dataset/         ├── Breast/         │   ├── Edge_feature/         │   │   ├── Breast_tcga_ecv.csv      # TCGA patient-specific edge contribution values (ECv)         │   │   └── Breast_lincs_kd_ecv.csv  # LINCS knockdown edge contribution values         │   ├── Node_feature/         │   │   ├── Breast_TCGA_exp.csv       # TCGA patient-specific gene expression levels         │   │   └── Breast_LINCS_KD_exp.csv   # LINCS knockdown gene expression levels         │   └── make_GRN_dataset/         │       └── mk_GRN.py                 # builds per-sample PyG graphs and pickles them         ├── Colorectal/…                       # same structure for colorectal cancer         └── Lung/…                             # same structure for lung cancer ``` - Edge_feature: 1D scalar “edge contribution values” (ECv) per sample - Node_feature: 1D scalar gene expression per sample - mk_GRN.py: combines node + edge features into torch_geometric.data.Data objects and serializes them. ### 2. Label Data for Finetuning All task labels live under `data/labels/`: ```plaintext └── data/     └── labels/         ├── BP_data/         │   └── gene_with_BP_multilabels.csv    # GO-BP multilabels (shared)         ├── CC_data/         │   └── gene_with_CC_multilabels.csv    # GO-CC multilabels (shared)         ├── Cancer_rel_data/         │   └── gene_with_cancer_relation.csv   # Cancer-relation labels (shared)         ├── Subtype_data/         │   └── Breast/subtype.csv              # Breast cancer subtype per patient         └── Survival_data/             ├── Breast/tcga_survival_time.csv   # OS time & event for hazard prediction             ├── Colorectal/…             └── Lung/… ``` ### 3. Metadata Helper files for mapping and filtering samples, under `data/meta_data/`: ```plaintext SupGCL/ └── data/     └── meta_data/         └── Breast/             ├── Breast_LINCS_KD_graphs_metadata.pkl   # Order of LINCS KD graphs             ├── LINCS_sampleID_KDgene_metadata.pkl    # Map LINCS sample → knocked-down gene             └── Breast_tcga_graphs_metadata.pkl       # Order of TCGA patient graphs ``` - LINCS metadata: used by SupGCL pretraining to match teacher (knockdown) graphs - TCGA metadata: used in finetuning to filter only patients with survival/subtype annotations (via --meta) ## About TCGA Datasets The Cancer Genome Atlas (TCGA) Research Network . TCGA TARGET GTEx data were accessed through the UCSC Xena Browser:   Derived data available at: https://xenabrowser.net/datapages/?cohort=TCGA%2520TARGET%2520GTEx&removeHub=https%253A%252F%252Fxena.treehouse.gi.ucsc.edu%253A443 ## About LINCS Datasets Subramanian A, et al. "A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles." Cell 2017. LINCS Program, NIH. L1000 data available at GEO: GSE.92742 ## About Algorithm of Bayesian Network Yoshinori Tamada, Teppei Shimamura, Rui Yamaguchi, Seiya Imoto, Masao Nagasaki, and Satoru Miyano. Sign: Large-Scale Gene Network Estimation Environment for High Performance Computing. Genome Informatics, 25(1):40–52, 2011. Seiya Imoto, Takao Goto, and Satoru Miyano. Estimation of genetic networks and functional struc- tures between genes by using Bayesian networks and nonparametric regression. Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing, pages 175–186, 2002.   ## Acknowledgments This work was supported by JST Moonshot R\&D (JPMJMS2021, JPMJMS2024), JST Research and Development Program for Next-generation Edge AI Semiconductors (JPMJES2511), JSPS KAKENHI (JP25K00148, JP25H02626, JP26K14994), and a project (JPNP14004) commissioned by the New Energy and Industrial Technology Development Organization (NEDO). This work used computational resources of the supercomputer Fugaku provided by RIKEN through the HPCI System Research Project (Project IDs: hp150272, ra000018). Taisei Tosaki received financial support from RIKEN Jr. Research-associated Programs.

Ähnliche Einträge