@genebe/gpn-star-hg38-m447-200m

GPN-Star (mammals, 200M) — genome-wide variant effect scores

This annotation contains exhaustive genome-wide single-nucleotide variant scores from GPN-Star, a genomic language model from the Song Lab, for the human reference genome (GRCh38). This specific model (songlab/gpn-star-hg38-m447-200m) was trained on a whole-genome alignment of 447 mammalian species and has 200M parameters.

Every possible single-nucleotide substitution at every position of the genome is scored with two columns:

• llr_calibrated: calibrated log-likelihood ratio for the specific reference-to-alternate substitution.
• abs_llr_calibrated: a separately calibrated score from the model. Despite its name, it is not the absolute value of llr_calibrated.

Refer to the source publication for the exact calibration methodology and the sign convention of both scores.

Citation

Ye C, Benegas G, Albors C, Li JC, Prillo S, Fields PD, Clarke B, Song YS. Predicting genome-wide functional constraints with GPN-Star. Nature. 2026. https://www.nature.com/articles/s41586-026-11005-5

Preprint: https://www.biorxiv.org/content/10.1101/2025.09.21.677619v1

Code: https://github.com/songlab-cal/gpn