dbNSFP

Meta-database aggregating 30+ functional prediction scores for all possible non-synonymous SNVs in the human genome. One query → all predictors at once.

What It Does

  • 30+ predictor scores per variant: REVEL, CADD, SIFT, PolyPhen-2, MutationTaster, FATHMM, GERP, PhyloP, PhastCons, AlphaMissense, and more
  • Pre-computed for ALL possible missense variants
  • Tabix-indexed for fast lookups
  • Also includes conservation and regulatory scores

How to Use

Download (large!)

# Download dbNSFP v4.5 (by chromosome, ~5GB each)
wget https://dbnsfp.s3.amazonaws.com/dbNSFP4.5a_variant.chr15.gz
wget https://dbnsfp.s3.amazonaws.com/dbNSFP4.5a_variant.chr15.gz.tbi
 
# Query STRC E1659A
tabix dbNSFP4.5a_variant.chr15.gz 43600551-43600551

Via Ensembl VEP (Easier)

# VEP with dbNSFP plugin gets all scores
vep --input_file variant.vcf --plugin dbNSFP,dbNSFP4.5a.gz,REVEL_score,CADD_phred,AlphaMissense_score

Via VarSome / Franklin

Both show dbNSFP scores on their variant pages.

Verified Status

VERIFIED via Ensembl VEP (returns dbNSFP-equivalent scores). E1659A results:

  • SIFT: deleterious (score: 0)
  • PolyPhen-2: possibly_damaging (0.807)
  • CADD PHRED: 25.5 (note: VEP returns 25.5 vs direct API 27.5 — version difference)
  • Conservation score: 2
  • All predictors concordant: deleterious/damaging.

Full dbNSFP download would add ~25 more scores but key ones already obtained.

STRC Research Usage

  • Individual predictors already used via separate APIs
  • dbNSFP would streamline: one download → all scores for any STRC variant

Connections