SpliceTransformer predicts tissue-specific splicing affect

Variant prediction

Input should be in "chromosome_offset_reference_alternation" format. (hg38)

The button reports error if wrong reference allele was given. We will revise the hint messages soon.

VariantΔSpliceAdiposeBloodBlood VesselBrainColonHeartKidneyLiverLungMuscleNerveSmall IntestineSkinSpleenStomach
No data

Predict splicing changes and specifically affected tissues for the variant.

ΔSplice score > 0.27 : the variant is likely to affect splicing with pathogenicity.

ΔSplice score > 0.09 : the variant is less likely to affect splicing but have uncertain significance.

ΔSplice score < 0.09 : the variant is unlikely to have no effect on splicing.

Tissue columns shows predicted tissue specificity for the variant.

If all columns are marked as "No", the variant is likely to affect splicing in all tissues.


You can also directly fetch the api like
The result will be in json format. A string like "score|1011111110" represents the prediction, where the "score" is predicted ΔSplice score, and each digit in the number represents tissue specificity for "Adipose Tissue", "Blood", "Blood Vessel", "Brain", "Colon", "Heart", "Kidney", "Liver", "Lung", "Muscle", "Nerve", "Small Intestine", "Skin", "Spleen", "Stomach", respectively.

SpTransformer source code is available at GithubThe model can be deployed locally to predict massive variants from VCF files. The model output may vary slightly depending on the CUDA version and GPU model. The results in the paper are based on CUDA 11.5 and A100 GPU.