Our model only
CC BY 4.0Free for any use, including commercial, if you credit the source.
- Every variant, scored by our model alone
- No third-party predictions
- Gzipped CSV, file name
transporter_variants_native.csv.gz
Download and cite
Each file is one table covering all 349,056 variants. The two files differ in whether AlphaMissense was used, and so in their licences. Choose the one that fits how you will use it.
With AlphaMissense
CC BY-NC-SA 4.0Non-commercial use only. Anything you build from it must be shared under the same licence. These terms come from AlphaMissense.
- Our model combined with AlphaMissense for the 177,464 missense variants it covers; our model alone for the rest
- The more accurate version for missense variants
- Gzipped CSV, file name
transporter_variants_routed.csv.gz
Please keep the two files separate when you share or reuse them, so each keeps its own licence. To get one gene only, use the download button on that gene's page.
Documentation files
- README: what the release contains
- Data dictionary: every column, also described on How to use
- Calibration check: the score cut-offs and the held-out test of the evidence levels
- Evidence levels in 1.1.0: how many variants sit at each level, and how they map to the 1.0.x levels
- Changes between releases
- CITATION.cff and SHA-256 checksums
How to cite
If you use these predictions, please cite:
Abo-Dahab Y, Chun J, Giacomini KM. Calibrated pathogenicity predictions for 783 human transporter and ion-channel genes. Version 1.0.1, 25 October 2026. https://joevondahab.github.io/transporter-variants/
If you use the With AlphaMissense version, please also cite:
Cheng J, Novati G, Pan J, et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science 381, eadg7492 (2023). doi:10.1126/science.adg7492
Data sources
- Variants and classifications from ClinVar.
- Protein sequences, membrane topology and functional sites from UniProtKB (CC BY 4.0).
- Protein structures from AlphaFold DB (CC BY 4.0), loaded from alphafold.ebi.ac.uk when you open a gene page.
- Evidence thresholds from Tavtigian SV et al., Genet Med 20, 1054–1060 (2018).
The site's own code is MIT licensed and available on GitHub.