🧬 AlphaGenome-PyTorch

Predict genomic tracks straight from DNA with a PyTorch port of DeepMind's AlphaGenome — 450M parameters, predicting chromatin accessibility, transcription and histone marks at 128 bp resolution across up to 131 kb of sequence.

Sequence is fetched live from the UCSC REST API, so any hg38 / mm10 locus works. Model: gtca/alphagenome_pytorch · code: alphagenome-pytorch

Organism
Window (bp)
Sequence length centred on the locus.
RefSeq gene models from UCSC, drawn under the tracks.
Assays
Biosamples
Each is averaged over its tracks and drawn as its own line. Empty = all tracks.
K562

Tracks averaged

Examples