Accurate variant interpretation remains challenging because most computational methods are tissue-agnostic and typically evaluate a single reference transcript, overlooking tissue- and isoform-specific biology. exTSP addresses this limitation by combining:

  1. IsoPath: likelihood-calibrated posterior probabilities derived from MutPred2 scores, and

  2. Ptse: transcript-level tissue specificity derived from GTEx and/or developmental expression from BrainSpan.

For each variant-transcript-tissue combination, exTSP produces an interpretable score that changes when a variant has high pathogenicity prediction and affects a transcript that is preferentially expressed in the disease-relevant tissue.


Data & Methods

exTSP integrates transcript-level expression with calibrated pathogenicity predictions.

Pathogenicity component (IsoPath): MutPred2 scores are converted to likelihood-calibrated posterior probabilities using distributions of known pathogenic and benign variants and a prior reflecting rare-variant pathogenicity.

Expression component (Ptse): Transcript expression across tissues is normalized into tissue-specific probabilities, then transformed into interpretable likelihood ratios reflecting tissue specificity.

Final score (exTSP): IsoPath and tissue specificity are combined to produce a tissue-aware score for each variant– transcript–tissue triplet, enabling transcript- and tissue-informed prioritization.


Limitations

  1. exTSP depends on the availability and quality of transcript expression estimates; rare isoforms and lowly expressed transcripts may have higher uncertainty.

  2. GTEx reflects adult tissue expression; developmental contexts rely on BrainSpan where available and may not capture all relevant stages or cell types.

  3. Scores are intended for prioritization and hypothesis generation; final interpretation should incorporate genetics, segregation, phenotype, and functional evidence.


Citing exTSP

If you use exTSP in your work, please cite:
Kakati T, Jain S, Sharma H, Pejaver V, Radivojac P, Iakoucheva LM. Improving Missense Variant Effects Predictors using Tissue Context and Splicing Isoform Expression. (Year). (Journal / preprint link forthcoming)