STRC Research

SpliceAI

2 min read

SpliceAI

Deep learning model for predicting splicing effects of genetic variants. Developed by Illumina. Predicts whether a variant creates or disrupts splice sites.

What It Does

  • Predicts splice site creation/disruption from DNA sequence
  • Four scores per variant: acceptor gain, acceptor loss, donor gain, donor loss
  • Score 0-1: >0.5 likely affects splicing, >0.8 high confidence
  • Distance to nearest affected splice site
  • Works for any variant type (SNV, indel) within 10kb of splice sites

How to Use

Web Lookup

  1. Go to https://spliceailookup.broadinstitute.org/
  2. Enter variant: chr15-43600551-A-C (GRCh38)
  3. View delta scores for each splice effect

Python Package

pip install spliceai
from spliceai import predict
# Requires reference genome FASTA
# predict(variant, reference_fasta, annotation)

Pre-computed Scores

# Download pre-computed (SNVs only)
# Available from Illumina: https://basespace.illumina.com/s/otSPW8hnhaZR
tabix spliceai_scores.raw.snv.hg38.vcf.gz chr15:43600551-43600551

Via VEP Plugin

vep --input_file variants.vcf --plugin SpliceAI,snv=spliceai_scores.raw.snv.hg38.vcf.gz

Verified Status

PARTIAL — pip install works (v1.3.1) but runtime fails due to deprecated pkg_resources. API (spliceailookup-api.broadinstitute.org) TIMES OUT. Web lookup at spliceailookup.broadinstitute.org works for manual queries only. Pre-computed scores available via Illumina BaseSpace. For E1659A: must use web interface manually (API times out). As a missense in exon 29, major splice effects unlikely. AlphaGenome SPLICE output showed no significant splice junction changes at this position (all quantile scores ~0.99 for nearby genes, not STRC itself).

STRC Research Usage

  • Not yet used for E1659A
  • Critical for: evaluating any intronic STRC variants, assessing cryptic splice sites

Connections

Graph View

Backlinks