RegulomeDB

Regulatory evidence for noncoding SNVs: ENCODE TF ChIP, DNase, motifs, QTLs. Rank 1a (strongest) through 7 (nothing overlapping). Neighbour of encode.md and jaspar.md. Not a consequence scorer — for missense impact use Ensembl VEP (ensembl.md §6, CADD=1).

Search: https://regulomedb.org/regulome-search/ Help: https://regulomedb.org/regulome-help/ All figures verified 2026-09-10.

Request

GET /regulome-search/?regions={rsID|chr:start-end}&genome=GRCh38&format=json

genome is GRCh38 (preferred) or GRCh37. Default format is HTML.

Coordinates are 0-based half-open, same as BED. A single GRCh38 base at 1-based position P is chrN:{P-1}-{P}.

Verified calls

GET ...?regions=rs6983267&genome=GRCh38&format=json

regulome_score.ranking 1a, probability 0.93. This is the example of a regulatory hit.

GET ...?regions=rs28934578&genome=GRCh38&format=json
GET ...?regions=chr17:7675087-7675088&genome=GRCh38&format=json

Both: ranking 4, probability 0.60906. A coding missense with rank 4 is "not what this database is for," not "benign regulatory."

GET ...?regions=chr17:7674220-7674220&genome=GRCh38&format=json

HTTP 200, variants: [], no regulome_score.ranking. Zero-width / 1-based same-number ranges miss the base.

POST /regulome-search/   with a JSON body

HTTP 200, @id /regulome-notfound, @type includes regulome-help. The old POST API is dead. Use GET.

Traps

• Convert 1-based VEP/CADD positions before building regions=.
• Read regulome_score.ranking, not a legacy regulomedb_score field.
• probability is a string.
• Region queries without an rsID return common SNPs (MAF > 1%) in the window, not every possible SNV.
• Rank 7 / missing ranking is "no annotated overlap," not a proof the variant does nothing.