Disease Research with ENCODE Functional Genomics

When to Use

• User wants to connect GWAS variants to ENCODE regulatory elements for disease mechanism research
• User asks about "disease", "pathology", "therapeutic targets", "GWAS interpretation", or "clinical variants"
• User needs to annotate disease-associated loci with functional genomics data from ENCODE
• User wants to identify drug targets from epigenomic evidence using Open Targets integration
• Example queries: "find enhancers disrupted by diabetes GWAS hits", "identify drug targets from ChIP-seq data", "connect my disease variants to regulatory elements"

Leverage ENCODE's 926,535 cCREs and multi-layer functional data to understand disease mechanisms, interpret disease-associated variants, identify therapeutic targets, and connect genomic findings to clinical applications.

Scientific Rationale

The question: "How can ENCODE functional genomics help me understand a disease's molecular mechanisms and identify actionable targets?"

Over 90% of disease-associated variants from GWAS fall in non-coding regions (Maurano et al. 2012). They disrupt regulatory elements controlling gene expression, not protein sequences. ENCODE provides the most comprehensive catalog of these elements across hundreds of cell types and tissues. This skill connects (1) genetic association data, (2) ENCODE functional annotations, and (3) clinical/pharmacological databases for druggable targets.

Literature Foundation

Reference	Year	Journal	Key Contribution	Citations	DOI
ENCODE Phase 3	2020	Nature	926,535 human cCREs across 400+ biosamples, SCREEN portal	~1,656	10.1038/s41586-020-2493-4
Maurano et al.	2012	Science	Disease variants enriched in regulatory DNA; DNase hotspots explain 76.6% of GWAS SNPs	~3,500	10.1126/science.1222794
Finucane et al.	2015	Nat Genet	S-LDSC partitions heritability into functional annotations; ENCODE categories explain disproportionate heritability	~2,253	10.1038/ng.3404
Nasser et al.	2021	Nature	ABC model links enhancers to genes in 131 cell types; connected 5,036 GWAS signals to 2,249 genes	~468	10.1038/s41586-021-03446-x
Roadmap Epigenomics	2015	Nature	111 reference epigenomes; tissue-specific chromatin states; disease variant enrichment in tissue-specific marks	~5,810	10.1038/nature14248
Visscher et al.	2017	Am J Hum Genet	GWAS review — 10 years of discoveries, statistical frameworks, shift toward functional interpretation	~2,500	10.1016/j.ajhg.2017.06.005
Buniello et al.	2019	Nucleic Acids Res	GWAS Catalog — curated repository; >250,000 SNP-trait associations	~3,000	10.1093/nar/gky1120
Ochoa et al.	2021	Nucleic Acids Res	Open Targets Platform — integrates GWAS, functional genomics, drugs for systematic target identification	~600	10.1093/nar/gkaa1027

Step 1: Map Disease to Relevant Tissues

ENCODE regulatory elements are highly tissue-specific. Correct tissue mapping is the single most important decision.

Disease-Tissue Mapping Table

Disease Category	Primary Tissues	Key Cell Types	ENCODE Cell Lines	Example Diseases
Neurological	brain (cortex, hippocampus, cerebellum)	neurons, astrocytes, microglia	SK-N-SH, SK-N-DZ, BE2C	Alzheimer's, Parkinson's, schizophrenia
Cardiovascular	heart, aorta, blood vessels	cardiomyocytes, endothelial, smooth muscle	HUVEC, HCASMC	coronary artery disease, heart failure
Metabolic	pancreas, liver, adipose, muscle	beta cells, hepatocytes, adipocytes	HepG2, Panc1	type 2 diabetes, NAFLD, obesity
Cancer	tissue of origin	tumor cells, microenvironment	K562, HepG2, MCF-7, A549, HCT116, PC-3	leukemia, breast cancer, lung cancer
Autoimmune	blood, immune organs, thymus	T cells, B cells, macrophages	GM12878, Jurkat	RA, lupus, MS, type 1 diabetes
Respiratory	lung, trachea	alveolar epithelial, bronchial	A549, IMR-90	asthma, COPD, pulmonary fibrosis
Renal	kidney	podocytes, tubular epithelial	HEK293	CKD, IgA nephropathy, FSGS
Hepatic	liver, bile duct	hepatocytes, cholangiocytes	HepG2, Hep3B	NAFLD, cirrhosis, hepatitis
Endocrine	thyroid, adrenal, pituitary, pancreas	thyrocytes, adrenal cortical, beta cells	—	hypothyroidism, Cushing's
Musculoskeletal	bone, cartilage, skeletal muscle	osteoblasts, chondrocytes, myocytes	—	osteoarthritis, osteoporosis
Gastrointestinal	intestine, colon, stomach	epithelial, goblet, Paneth cells	HCT116, Caco-2	IBD, Crohn's, celiac
Hematological	blood, bone marrow	HSCs, erythrocytes, megakaryocytes	K562, GM12878, CD34+	sickle cell, thalassemia, AML

Check ENCODE availability:

encode_get_facets(organ="pancreas")
encode_get_facets(organ="brain")

If tissue has limited data: Use Tier 1 cell lines (K562, GM12878, H1-hESC) or Roadmap Epigenomics as proxies. Document the mismatch explicitly.

Step 2: Find Disease-Relevant ENCODE Data

For GWAS / Genetic Diseases

Open chromatin and active enhancers in disease tissue (Maurano et al. 2012):

encode_search_experiments(assay_title="ATAC-seq", organ="...", biosample_type="tissue")
encode_search_experiments(assay_title="Histone ChIP-seq", target="H3K27ac", organ="...")
encode_search_experiments(assay_title="DNase-seq", organ="...", biosample_type="tissue")

Then use the variant-annotation skill to overlap variants with functional elements.

For Cancer Research

ENCODE cancer cell lines (NOT tumors — see Cancer Epigenomics section below):

encode_search_experiments(biosample_term_name="K562")    # CML
encode_search_experiments(biosample_term_name="HepG2")   # Hepatocellular carcinoma
encode_search_experiments(biosample_term_name="MCF-7")   # ER+ breast cancer
encode_search_experiments(biosample_term_name="A549")     # Lung adenocarcinoma

For Perturbation Studies

encode_search_experiments(perturbed=True, organ="...")
encode_search_experiments(assay_title="CRISPR screen", organ="...")

For Rare Diseases

1. Search for closest available tissue; 2. Check Roadmap Epigenomics; 3. Use Tier 1 cell lines as baseline; 4. Document tissue proxy limitations.

Step 3: Cross-Reference with Disease Databases

PubMed — Disease Literature

search_articles(query="[DISEASE] AND (ENCODE OR regulatory element OR enhancer)")

Track experiments and link papers:

encode_track_experiment(accession="ENCSR...", notes="Disease research - [disease]")
encode_get_citations(accession="ENCSR...")
encode_link_reference(experiment_accession="ENCSR...", reference_type="pmid", reference_id="12345678")

ClinicalTrials.gov — Active Trials

search_trials(condition="[DISEASE]", intervention="[TARGET_GENE or DRUG]", status=["RECRUITING"])

Link trials: encode_link_reference(experiment_accession="ENCSR...", reference_type="nct_id", reference_id="NCT...")

Open Targets — Target-Disease Associations

search_entities(query_strings=["[GENE_NAME]"])
query_open_targets_graphql(
    query_string="query target($ensemblId: String!) { target(ensemblId: $ensemblId) { approvedSymbol knownDrugs { rows { drug { name } phase mechanismOfAction } } } }",
    variables={"ensemblId": "ENSG..."}
)

bioRxiv — Recent Preprints

search_preprints(category="genetics", recent_days=90)
encode_link_reference(experiment_accession="ENCSR...", reference_type="preprint_doi", reference_id="10.1101/...")

Step 4: Annotate Disease Variants with ENCODE Functional Data

Variant-to-Regulatory-Element Annotation

Overlap Category	ENCODE Data Type	Interpretation
cCRE-PLS	DNase + H3K4me3 near TSS	Variant in promoter-like element
cCRE-dELS	DNase + H3K27ac >2kb from TSS	Variant in distal enhancer
cCRE-pELS	DNase + H3K27ac within 2kb of TSS	Variant in proximal enhancer
cCRE-CTCF	DNase + CTCF	Variant may disrupt insulator/boundary
TF ChIP peak	TF ChIP-seq	Variant at TF binding site
No overlap	—	Not active in tested tissues

Retrieve functional data files:

encode_list_files(experiment_accession="ENCSR...", file_format="bed", output_type="IDR thresholded peaks", assembly="GRCh38", preferred_default=True)

GWAS Variant Interpretation Workflow (Maurano 2012 + Finucane 2015)

From GWAS hits to candidate causal variants to target genes:

1. Obtain GWAS loci from the GWAS Catalog (Buniello et al. 2019) or published study
2. Expand to LD proxies (r2 > 0.8) — lead SNPs are NOT necessarily causal
3. Map to disease tissue using the tissue mapping table (Step 1)
4. Overlay with ENCODE cCREs active in disease tissue
5. Layer additional annotations: TF binding, chromatin state, 3D genome
6. Prioritize: Variants in tissue-specific enhancers with TF binding = highest priority
7. Link to target genes via ABC model (Nasser et al. 2021), Hi-C, or eQTL colocalization

This reduces a typical locus from 50-200 LD variants to 1-5 high-confidence candidates.

Step 5: Build the Disease Regulatory Model

6-layer model combining ENCODE data types:

1. Active regulatory elements: ATAC-seq + H3K27ac identify enhancers/promoters in disease tissue
2. TF binding: TF ChIP-seq maps which factors bind at regulatory elements
3. Enhancer-gene linkage: Hi-C/ChIA-PET + ABC model (Nasser et al. 2021) connect enhancers to target genes
4. Gene expression: RNA-seq measures baseline expression of candidate targets
5. Disease variant overlay: Intersect GWAS/eQTL variants with layers 1-4
6. Druggable target assessment: Cross-reference target genes with Open Targets and ClinicalTrials.gov

encode_search_experiments(assay_title="ATAC-seq", organ="...", biosample_type="tissue")
encode_search_experiments(assay_title="Histone ChIP-seq", target="H3K27ac", organ="...")
encode_search_experiments(assay_title="TF ChIP-seq", organ="...")
encode_search_experiments(assay_title="Hi-C", organ="...")
encode_search_experiments(assay_title="total RNA-seq", organ="...", biosample_type="tissue")

Step 6: Identify Therapeutic Targets

Connect ENCODE regulatory elements to druggable targets (Ochoa et al. 2021).

The variant-to-drug pipeline:

• GWAS variant disrupts an enhancer (ENCODE annotation)
• Enhancer regulates Gene X (ABC model / Hi-C)
• Gene X is druggable (Open Targets tractability)
• Existing drug targets Gene X (Open Targets knownDrugs)
• Clinical trials test that drug (ClinicalTrials.gov)

search_entities(query_strings=["[GENE_NAME]"])
query_open_targets_graphql(
    query_string="query target($ensemblId: String!) { target(ensemblId: $ensemblId) { approvedSymbol tractability { label modality value } knownDrugs { count rows { drug { name } phase status } } } }",
    variables={"ensemblId": "ENSG..."}
)

Targets with both ENCODE regulatory evidence AND GWAS genetic association are the strongest candidates — orthogonal lines of support.

Step 7: Validate Findings

Cross-Study Replication

encode_track_experiment(accession="ENCSR...", notes="Validation - independent replicate")
encode_compare_experiments(accession1="ENCSR...", accession2="ENCSR...")

Functional Validation Hierarchy

Level	Method	Strength	In ENCODE?
Strongest	CRISPRi/CRISPRa perturbation	Direct causal test	Yes (ENCODE 4)
Strong	MPRA / STARR-seq	High-throughput activity	Yes (ENCODE 4)
Moderate	Reporter assays / eQTL coloc	Element activity / statistical	External
Supportive	Chromatin annotation / conservation	Biochemical / evolutionary	Yes / External

encode_search_experiments(perturbed=True, organ="...")
encode_search_experiments(assay_title="CRISPR screen", organ="...")

Step 8: Clinical Trial Cross-Reference

search_trials(condition="[DISEASE]", intervention="[GENE or DRUG]", status=["RECRUITING"])
search_trials(condition="[DISEASE]", phase=["PHASE2", "PHASE3"])
search_by_sponsor(sponsor_name="[PHARMA]", condition="[DISEASE]", phase=["PHASE3"])
analyze_endpoints(condition="[DISEASE]", phase=["PHASE3"])

Link trials to tracked experiments:

encode_link_reference(experiment_accession="ENCSR...", reference_type="nct_id", reference_id="NCT...", description="Trial targeting [gene] from ENCODE analysis")

Step 9: Document for Publication

encode_track_experiment(accession="ENCSR...", notes="Disease research - [disease] - [layer]")
encode_log_derived_file(
    file_path="/path/to/disease_model.bed",
    source_accessions=["ENCSR...", "ENCSR..."],
    description="Disease regulatory model - active enhancers overlapping GWAS loci",
    file_type="disease_model", tool_used="bedtools intersect + ABC model",
    parameters="GRCh38, IDR thresholded peaks, ABC score > 0.015"
)
encode_get_citations(export_format="bibtex")
encode_export_data(format="csv")

Heritability Enrichment with ENCODE Annotations

S-LDSC (Finucane et al. 2015) quantifies how much disease heritability is attributable to ENCODE-defined annotation categories. This is the most rigorous validation that ENCODE elements are disease-relevant.

Why it matters: If heritability is enriched in tissue-specific enhancers (e.g., 15x in pancreatic islet enhancers for T2D), this confirms correct tissue mapping and that regulatory variation drives disease risk.

Requirements: Full GWAS summary statistics, LD score files, ENCODE BED annotations, ldsc software.

Interpretation: Enrichment > 1 = more heritability than expected by annotation size. Compare across tissues to confirm disease-relevant tissue. Significant enrichment (p < 0.05/n_annotations) validates the ENCODE-based disease model.

Cancer Epigenomics with ENCODE

Cell Line	Cancer Type	Key Features	Caveats
K562	CML	Tier 1 — most ENCODE data; BCR-ABL	Transformed hematopoietic
HepG2	Hepatocellular carcinoma	Tier 2 — extensive liver regulatory data	Well-differentiated
MCF-7	ER+ breast cancer	Hormone-responsive model	ER+ only
A549	Lung adenocarcinoma	Epithelial lung cancer model	KRAS-mutant
HCT116	Colorectal carcinoma	MSI-H model	Not MSS CRC
PC-3	Prostate adenocarcinoma	Androgen-independent	Late-stage only

Cell line vs. tumor caveats: Cell lines are clonal, immortalized, and lack tumor heterogeneity. Regulatory landscapes diverge from primary tumors. High-passage lines accumulate epigenetic drift. Use for hypothesis generation; validate against TCGA/primary tumor data. Comparing cancer cell line vs. normal tissue counterpart (e.g., HepG2 vs. primary liver) can reveal cancer-specific regulatory changes.

Pitfalls and Common Mistakes

1. Tissue Relevance Mismatch

Annotating variants with wrong-tissue data is the most common error. Verify tissue selection against published heritability enrichment (Finucane 2015; Roadmap 2015).

2. Cell Line vs. Primary Tissue

Tier 1 cell lines have deepest data but are transformed. Prioritize: primary tissue > primary cells > cell lines. State cell line limitations explicitly.

3. LD Confounding

A lead SNP is NOT the causal variant. Always expand to LD proxies (r2 > 0.8) or use fine-mapped credible sets. Annotating only lead SNPs misses the causal variant in most cases.

4. Functional Does Not Equal Causal

Regulatory element overlap is necessary but not sufficient. The variant must DISRUPT the element. CRISPRi/MPRA validation is required for proof. Chromatin overlap is context, not mechanism.

5. Publication Bias and Ascertainment

ENCODE is biased toward well-studied tissues. Rare disease-relevant tissues may lack coverage. Absence of data does NOT mean absence of regulatory elements.

6. Over-Interpreting Regulatory Overlap

Multiple LD variants may overlap regulatory elements by chance. Enrichment testing (S-LDSC, GARFIELD) and fine-mapping distinguish real signal from LD-driven overlap.

Integration

This skill produces...	Feed into...	Using tool/skill
GWAS hits in ENCODE regulatory elements	Drug target identification	cross-reference → Open Targets
Disease-tissue regulatory map	Enhancer annotation	regulatory-elements skill
Candidate therapeutic targets	Clinical trial search	cross-reference → ClinicalTrials.gov
Variant-regulatory element intersections	Variant annotation	variant-annotation + clinvar-annotation
Tissue-specific regulatory network	Expression correlation	gtex-expression skill

Walkthrough: ENCODE-Driven Disease Mechanism Discovery for Alzheimer's Disease

Goal: Use ENCODE regulatory data combined with disease databases to identify candidate regulatory mechanisms underlying Alzheimer's disease risk variants. Context: Alzheimer's GWAS has identified 75+ risk loci, most in non-coding regions. ENCODE maps the regulatory landscape to interpret these variants.

Step 1: Survey ENCODE data for brain tissue

encode_get_facets(organ="brain", organism="Homo sapiens")

Expected output:

{
  "assay_title": [
    {"term": "Histone ChIP-seq", "count": 120},
    {"term": "TF ChIP-seq", "count": 45},
    {"term": "ATAC-seq", "count": 32},
    {"term": "total RNA-seq", "count": 28},
    {"term": "DNase-seq", "count": 22},
    {"term": "Hi-C", "count": 8}
  ]
}

Interpretation: Extensive brain epigenomic data available. Histone ChIP-seq (120 experiments) and ATAC-seq (32) provide the strongest regulatory element maps.

Step 2: Find brain enhancer experiments

encode_search_experiments(assay_title="Histone ChIP-seq", organ="brain", target="H3K27ac", organism="Homo sapiens")

Expected output:

{
  "results": [
    {"accession": "ENCSR100BRN", "assay_title": "Histone ChIP-seq", "target": "H3K27ac", "biosample_summary": "brain tissue male adult (54 years)"}
  ],
  "total": 24,
  "limit": 25,
  "offset": 0,
  "has_more": false,
  "next_offset": null
}

Step 3: Cross-reference with Alzheimer's GWAS variants

Using → gwas-catalog skill:

# Query GWAS Catalog for Alzheimer's disease associations
GET https://www.ebi.ac.uk/gwas/rest/api/associations?efoTrait=EFO_0000249

Intersect GWAS variants with ENCODE brain H3K27ac peaks to identify risk variants in active brain enhancers.

Step 4: Annotate candidate regulatory variants

Using → variant-annotation pipeline:

1. VEP consequence prediction (→ ensembl-annotation)
2. ClinVar pathogenicity (→ clinvar-annotation)
3. gnomAD population frequency (→ gnomad-variants)
4. Brain-specific gene expression (→ gtex-expression)

Step 5: Identify disease mechanism hypotheses

For each Alzheimer's risk variant in a brain enhancer:

• Which gene? → Peak-annotation assigns target genes
• Which cell type? → scATAC-seq reveals neuron vs. microglia specificity
• What pathway? → Open Targets provides pathway context
• Existing drugs? → Open Targets identifies therapeutic opportunities

Step 6: Generate disease research summary

encode_summarize_collection()

Interpretation: A comprehensive disease research report links ENCODE regulatory data → GWAS variants → annotated variants → candidate genes → therapeutic targets, creating an end-to-end discovery pipeline.

Integration with downstream skills

• Brain enhancer peaks from → histone-aggregation provide the regulatory landscape
• GWAS variants from → gwas-catalog define the disease-associated search space
• Variant annotation from → variant-annotation provides functional predictions
• Expression validation from → gtex-expression confirms tissue-specific gene activity
• Single-cell resolution from → cellxgene-context identifies disease-relevant cell types
• Drug targets from Open Targets connect to therapeutic opportunities

Code Examples

1. Survey ENCODE data for a disease-relevant tissue

encode_get_facets(organ="brain", organism="Homo sapiens")

Expected output:

{
  "assay_title": [
    {"term": "Histone ChIP-seq", "count": 120},
    {"term": "ATAC-seq", "count": 32},
    {"term": "total RNA-seq", "count": 28},
    {"term": "Hi-C", "count": 8}
  ]
}

2. Find experiments for disease-relevant cell types

encode_search_experiments(assay_title="ATAC-seq", biosample_term_name="microglia", organism="Homo sapiens")

Expected output:

{
  "results": [
    {"accession": "ENCSR200MIC", "assay_title": "ATAC-seq", "biosample_summary": "microglia", "status": "released"}
  ],
  "total": 4,
  "limit": 25,
  "offset": 0,
  "has_more": false,
  "next_offset": null
}

3. Get citations for disease research

encode_get_citations(accession="ENCSR100BRN")

Expected output:

{
  "publications": [
    {
      "id": 1,
      "experiment_accession": "ENCSR100BRN",
      "pmid": "31234567",
      "doi": "10.1038/s41586-023-06345-2",
      "title": "Brain enhancer atlas reveals Alzheimer regulatory mechanisms",
      "authors": "Li YE, Preissl S, Hou X",
      "journal": "Nature",
      "year": "2023",
      "abstract": ""
    }
  ],
  "count": 1
}

Related Skills

• variant-annotation — Detailed variant-level functional annotation with ENCODE data
• regulatory-elements — Discovering and characterizing regulatory elements independent of disease context
• quality-assessment — Evaluating quality of ENCODE experiments used in disease research
• cross-reference — Connecting ENCODE data to PubMed, bioRxiv, ClinicalTrials.gov
• data-provenance — Tracking the full chain from ENCODE source data to derived disease models
• epigenome-profiling — Build tissue epigenomic profiles as the foundation for disease regulatory models
• single-cell-encode — Cell type-resolved data for dissecting disease mechanisms in heterogeneous tissues
• histone-aggregation — Aggregate histone peaks across donors before disease variant annotation
• accessibility-aggregation — Aggregate ATAC-seq peaks across donors for improved sensitivity
• multi-omics-integration — Combine RNA, ATAC, histone, and TF data for comprehensive disease regulatory landscapes
• pipeline-guide — Guidance for S-LDSC, fine-mapping, and other disease genomics pipelines
• gnomad-variants — Population frequency filtering and gene constraint for disease gene prioritization
• ensembl-annotation — VEP annotation and phenotype associations from Ensembl
• ucsc-browser — UCSC Genome Browser tracks for disease locus visualization
• geo-connector — Find complementary disease expression datasets in GEO
• clinvar-annotation — Clinical significance of disease variants from ClinVar
• gwas-catalog — GWAS trait associations and risk alleles from NHGRI-EBI catalog
• gtex-expression — Tissue expression context for disease genes across 54 GTEx tissues
• publication-trust — Verify literature claims backing analytical decisions

Presenting Results

• Present disease-relevant findings as: gene/variant | disease | evidence_type | source | confidence. Show ENCODE regulatory support. Suggest: "Would you like to find clinical trials related to these findings?"

For the request: "$ARGUMENTS"