AllenInstitute/BG-human-enformer

Enformer fine-tuned on human (Homo sapiens) basal ganglia chromatin accessibility across 60 cell-type tracks. Long-context (196,608 bp) companion to the dilated-CNN basal ganglia models.

Background

A transformer predicting basal ganglia cell-type accessibility, complementing the shorter-context dilated CNNs trained on the same tracks.

Availability

Currently private. RegRex (formerly enhancer-designer) is an internal Allen Institute tool with a public release planned. For early access contact Kasia Kedzierska (kasia.kedzierska@alleninstitute.org).

Model Details

• Architecture: enformer
• Species: Homo_sapiens
• Number of tracks: 60
• Framework: PyTorch Lightning
• Checkpoint format: Lightning .ckpt

Target Tracks

• AMY-SLEA-BNST_D1_GABA
• AMY-SLEA-BNST_GABA
• Astrocyte
• BAM
• BF_SKOR1_Glut
• B_cells
• COP
• Endo
• Ependymal
• GPe-NDB-SI_LHX6-LHX8-GBX1_GABA
• ... (50 more)

Usage

from regrex.models import load_model

model = load_model("AllenInstitute/BG-human-enformer")