---
license: mit
task_categories:
  - text-classification
  - token-classification
  - tabular-classification
  - other
tags:
  - bioinformatics
  - genomics
  - variant-calling
  - machine-learning
  - fantom5
size_categories:
  - 1G<n<2G
---

# Dataset Card for seqedge-data

## Dataset Description

This repository serves as the official public data backend for **GalibierHub**, an interactive web platform for genomic cohort analytics. The dataset hosts reference bundles, release archives, and sample-level files designed for academic research and downstream bioinformatics pipelines. It currently contains approximately 1.3 GB of open-access data published under the MIT license.

## Dataset Structure

The repository is organized into distinct directories to support both traditional bioinformatics workflows and machine learning applications. 

**Core Academic Data (GalibierHub Records)**
*   **Records/Reference_Genomes/**: Contains the global reference files, including `reference.fasta` and `annotation.gff3`.
*   **Records/Variant_Calling_VCF/**: Contains 12 sample-specific `.vcf.gz` files tailored for variant and population genetics analysis.
*   **Records/ML_Ready_FASTA/**: Contains 12 full `.fasta` files optimized for machine learning sequence extraction and promoter prediction.

**Additional Resources**
*   **SARS-CoV-2 Reference Files**: Root-level files including `scov2.fa`, `scov2.gb`, `scov2.genes.bed`, and `scov2.genes.gff3`.
*   **Learning-Resources/**: A collection of educational materials and bioinformatics reference PDFs.
*   **Graduate-student-textbooks/**: Supplemental textbooks for graduate-level study.

## Source Data Details

The 12 primary demo samples located in the `Records/` directory (e.g., CNhs13076, CNhs13080, CNhs13195, CNhs13216) are derived from public test files.
*   **Source**: FANTOM5 human primary-cell CAGE data (hg19)
*   **Cohort**: FANTOM5 human.primary_cell.hCAGE
*   **Platform**: HeliScope CAGE
*   **Project URL**: https://fantom.gsc.riken.jp/5/
*   **Reference Publication**: https://www.nature.com/articles/sdata2017112

## How to Use This Data

Files can be downloaded directly from this repository using the Hugging Face CLI, `wget`, or `curl`. However, for the optimal experience, we recommend using the **GalibierHub Downloads Interface**, which provides:
*   Automated generation of Python, SLURM, and `.bat` batch download scripts for recursive folder downloads.
*   MD5 and SHA-256 integrity verification commands to ensure data is uncorrupted after transfer.
*   Seamless routing from the embedded JBrowse 2 genome visualization directly to the corresponding file downloads.

## About GalibierHub

GalibierHub is a comprehensive scientific data portal built for modern researchers. Key end-user capabilities include:
*   **Search & Discovery**: Advanced filtering of promoter records by locus, gene, score, tissue, cohort, and BMI class.
*   **Data Visualization**: An embedded genome browser with a distraction-free fullscreen zen mode and floating promoter detail panels.
*   **Community & Moderation**: A rich discussion board featuring Markdown support, image lightboxes, real-time WebSocket notifications, and a gamified badge system to reward community contributions.
*   **User Profiles**: Persistent profiles synced via Supabase with activity feeds, online status, and secure authentication (GitHub OAuth or Turnstile-protected email).