: issues or feature requests for this website
for faster variant consequence lookup on both hg19 and hg38 as well as HGVS conversion on hg38, while still using
for HGVS on hg19.
icon in the results shows every position in the scoring window that has a non-zero SpliceAI or Pangolin score.
- Can now search for chrom:position to see REF scores (eg.
July 1st, 2026
- Improved SpliceAI-10k output when SpliceAI predicts multiple acceptor/donor shifts (examples:
NM_024675.4(PALB2):c.47A>G,
NM_000546.6(TP53):c.673-1G>A)
May, 2026
- Added
SpliceAI-10k Calculator predictions
- Updated list of Related web tools
December 14, 2025
- In the 'Other scores' section, only populate the 'Points' column for missense variants since the calibration was only done for missense
September 22, 2025
- updated to Gencode v49 (was previously on v47)
- added a new optional visualization track called
"genes used for precomputed scores". It shows the Gencode v24 transcript models originally used by the SpliceAI authors to generate the precomputed SpliceAI score tables available from Illumina. This track can help explain some of the differences in scores between SpliceAI-lookup and these precomputed tables.
June 27, 2025
- added PromoterAI scores for promoters of all protein-coding transcripts in Gencode v39 (not just the canonical transcripts)
- added PromoterAI scores for hg19
December 5, 2024
- added
PrimateAI-3D,
PromoterAI,
AlphaMissense and other scores
- provided a table with detailed SpliceAI scores for insertion variants whenever donor or acceptor gain is predicted to occurs within the inserted sequence itself. This addresses
issue #84.
MAJOR BUG FIX: Between April and November of 2024,
Pangolin scores were incorrectly computed for ~28% of hg38 variants and ~5% of hg19 variants due to a bug related to multiple requests being processed in parallel. This issue has now been fixed.
SpliceAI scores were fine and were not affected by this bug.
November 3, 2024
- added choice of Gencode
basic or comprehensive transcripts
- updated to Gencode v47
June 20, 2024
- fixed SpliceAI visualizations to clarify positions of predicted splicing changes. See
issue #70 for details.
June 3, 2024
- show the input variant's
VEP consequence in the results table
- the Gencode gene track is now optional in the visualizations
- updated everything to
Gencode v46
March 7, 2024
- added warning for insertion variants with delta scores ≥ 0.2 and position = 0bp saying that they may be difficult to interpret due to
issue #67. Thanks to @SophieCandille for the issue report and example variant:
2:47790924 C>CAGTTG
- moved server to Google Cloud Run to better support higher usage
November 30, 2023
- changed default search settings to "not masked" due to
confusion and
issues with how masked scores are computed.
- when looking at hg19 variants, the visualization checkboxes now make it clear which IGV.js tracks are or aren't available on hg19.
October 5, 2023
- fixed Pangolin bug identified by
@kazmiekr where variants that overlap more than one gene showed incorrect masked scores
-
@pj-sullivan added controls to show/hide transcripts
- added MANE Select transcript labels
- updated to gencode v44 "basic" annotations (instead of comprehensive). These have fewer transcripts per gene.
- added igv.js visualizations
May 22, 2023
- changed defaults to 'masked' and max distance to 500bp
- retired
Illumina precomputed scores since they were created using Gencode v24
and max distance = 50bp and so can differ from computed scores which are based on the latest Gencode.
May 12, 2023
- added REF & ALT score columns to show SpliceAI's underlying predictions for each haplotype
(using
SpliceAI PR by @h-joshi).
The Δ score column is the difference between these two scores.
- added support for MNPs (see
issue #1 &
SpliceAI PR by @kdahlo)
- updated to
Gencode v43 (was previously on v42)
May 10, 2023
- fixed bug where "raw" Pangolin scores were shown regardless of whether the user selected "raw" or "masked".
Feb 5, 2023
- show
Pangolin scores in addition to SpliceAI scores.
June 2, 2021
- show RefSeq ids for the subset of Ensembl ENST ids (~30%) that have one or more matching
RefSeq NM ids according to Ensembl. See
issue #8 for details.
April 11, 2021
- show gray background for non-coding transcripts
- switch to showing all non-coding transcripts. (Previously, transcripts with Gencode biotypes
like
lncRNA,
processed_pseudogene,
processed_transcript,
retained_intron, and
nonsense_mediated_decay were filtered out if they overlapped
protein_coding transcripts).
See
issue #6 for details.
[hide older updates]