pangolin

is a Data Source.

Pangolin is a deep learning model that predicts splice site usage from DNA sequence and the effect of genetic variants on splicing. Trained on splice site usage measured in RNA-seq from multiple tissues of human, rhesus macaque, mouse and rat, it reports, for a window around each variant, the maximum increase (gain) and decrease (loss) in splice site usage and their positions, optionally masked using a gene annotation. The Pangolin software is distributed on GitHub under GPL-3.0. Precomputed hg38 scores for all SNVs in protein-coding genes, masked using GENCODE splice site annotations, were computed with default parameters by researchers at the Technical University of Munich and published on Zenodo under CC BY 4.0.

License

GPL-3.0

Homepage

pangolin

Repository

GitHub

External IDs

No external IDs available

Product Summary

Products

From this Resource
ID Name URL Category Format Description
pangolin.software Pangolin Software Pangolin Process python Pangolin source code and trained mode...
pangolin.hg38-snvs-masked Pangolin Precomputed hg38 SNV Scores (Masked) 15649338 (12.1 GB) Product tsv Pangolin precomputed scores for all S...
From other Resources
ID Name URL Category Format Relation Description
biobtree.api BioBTree REST API api Programming Interface http had primary source REST API for searching identifiers an...
spliceai.lookup SpliceAI-lookup spliceailookup.broadinstitute.org Graphical Interface http had primary source SpliceAI-lookup, a web tool maintaine...

Details

Pangolin

Pangolin is a deep learning model for predicting RNA splicing from DNA sequence, developed by Tony Zeng and Yang I. Li at the University of Chicago. It was trained on splice site usage quantified from RNA-seq data across four tissues in four mammalian species (human, rhesus macaque, mouse and rat), and predicts both whether a position is a splice site and how strongly it is used. For a variant, Pangolin reports the largest predicted gain and loss in splice site usage within a window around the variant.

Access

  • Software: the Pangolin Python package on GitHub (GPL-3.0), which scores VCF or CSV input and can mask scores using a GENCODE-based annotation database. A Colab notebook is provided.
  • Precomputed scores: masked Pangolin scores for all hg38 SNVs in protein-coding genes, as a 13 GB zip of per-gene TSV files on Zenodo (CC BY 4.0). These were computed by Nils Wagner and Aleksandr Neverov at the Technical University of Munich, not by the original authors.
  • Pangolin predictions are also available for single variants through the Broad Institute SpliceAI-lookup web tool (see the SpliceAI page).

Is this information incorrect or incomplete? Request an update.

Created: November 12, 2026 | Last modified: November 12, 2026