# ChIP-Atlas

> ChIP-Atlas is a comprehensive database of public epigenomic data (ChIP-seq, ATAC-seq, DNase-seq, Bisulfite-seq) covering over 1M experiments across 10+ genome assemblies. It provides peak browsing, colocalization analysis, target gene prediction, enrichment analysis, and differential analysis.

ChIP-Atlas integrates data from NCBI SRA, GEO, and other public repositories. All experiments are uniformly processed and classified by genome, experiment type, antigen, and cell type.

## Main Features

- [Peak Browser](https://chip-atlas.org/peak_browser): Browse and download assembled peak-call data filtered by genome, experiment type, antigen, cell type, and significance threshold
- [Enrichment Analysis](https://chip-atlas.org/enrichment_analysis): Test whether your genomic regions are enriched for specific ChIP-seq/ATAC-seq signals
- [Differential Analysis](https://chip-atlas.org/diff_analysis): Compare epigenomic profiles between sample groups (DiffBind / DMR)
- [Target Genes](https://chip-atlas.org/target_genes): Predict target genes regulated by specific transcription factors or histone marks
- [Colocalization](https://chip-atlas.org/colo): Analyze genome-wide colocalization between epigenomic factors
- [Experiment Search](https://chip-atlas.org/search): Search all experiments by keyword, accession ID, or classification

## API for Agents

- [Agent Guide](https://chip-atlas.org/agents): Full documentation of MCP tools, HTTP API endpoints, data model, and usage workflows for AI agents
- [OpenAPI 3.1 Spec](https://chip-atlas.org/openapi.yaml): Machine-readable API specification for all JSON endpoints
- [MCP Server Source](https://github.com/inutano/chip-atlas/tree/master/mcp): TypeScript MCP server providing 10 read-only tools for querying ChIP-Atlas programmatically

## API Endpoints

- [List Genomes](https://chip-atlas.org/data/list_of_genome.json): GET — available genome assemblies
- [List Experiment Types](https://chip-atlas.org/data/list_of_experiment_types.json): GET — experiment type categories
- [Experiment Types with Counts](https://chip-atlas.org/data/experiment_types?genome=hg38&clClass=All%20cell%20types): GET — experiment types with counts for a genome and cell type class
- [Sample Types](https://chip-atlas.org/data/sample_types?genome=hg38&agClass=Histone): GET — cell type classes with counts for a genome and experiment type
- [Antigens](https://chip-atlas.org/data/chip_antigen?genome=hg38&agClass=Histone&clClass=undefined): GET — antigen subclasses with counts
- [Cell Types](https://chip-atlas.org/data/cell_type?genome=hg38&agClass=Histone&clClass=undefined): GET — cell type subclasses with counts
- [Experiment Metadata](https://chip-atlas.org/data/exp_metadata.json?expid=SRX018625): GET — detailed metadata for a single experiment by ID
- [Colocalization Index](https://chip-atlas.org/data/colo_analysis.json?genome=hg38): GET — available colocalization analyses for a genome
- [Target Genes Index](https://chip-atlas.org/data/target_genes_analysis.json): GET — genomes and antigens with target gene data

## Data Model

Experiments are classified along these dimensions: genome assembly (hg38, mm10, dm6, etc.), experiment class (Histone, TFs and others, ATAC-Seq, DNase-seq, Bisulfite-Seq), antigen subclass (H3K4me3, CTCF, etc.), cell type class (Blood, Brain, Liver, etc.), and cell type subclass (K-562, HeLa, etc.). Each experiment has an SRX (SRA) or GSM (GEO) accession ID.

## Important Notes for Agents

### Data architecture

ChIP-Atlas consists of two components: the **app server** (chip-atlas.org) serving the web UI and APIs, and the **data backend** (chip-atlas.dbcls.jp) hosting pre-computed result files. The app server generates URLs pointing to the data backend. Download URLs, colocalization results, and target gene results are all served from chip-atlas.dbcls.jp, not the app server itself.

### Peak Browser download

POST /download returns a URL to a pre-computed BED file on the data backend. The `qval` parameter in the condition must match a pre-computed threshold. Use GET /qvalue_range to retrieve valid values (typically "05", "10", "20", "50"). Other qval values will return `{"url": null}`.

The download condition requires all of: `genome`, `agClass`, `agSubClass`, `clClass`, `clSubClass` (use "-" for unspecified), and `qval`.

### Colocalization and Target Genes results

POST /colo and POST /target_genes return URLs to pre-computed result files on the data backend. Not all antigen/cell type combinations have pre-computed results. A 404 from the result URL means that combination has no data available, not that the API is broken. Use the colocalization and target genes index endpoints to check which combinations have data before requesting results.

### Experiment Search

GET /search performs full-text search across the experiment database. The search index is populated at app startup from the experiment metadata JSON. If the index is empty, the endpoint returns `{"total": 0, "returned": 0, "experiments": []}`. This does not indicate an error.

## Optional

- [Publications](https://chip-atlas.org/publications): Papers citing ChIP-Atlas
- [Documentation Wiki](https://github.com/inutano/chip-atlas/wiki): Detailed usage guide
- [GitHub Repository](https://github.com/inutano/chip-atlas): Source code