TI

Tim419/GLMap-panels

Dataset

GLMap Probe Panels

0

10 commits

1 linked in READMEs

updated Jun 2, 2026

See the code

README

GLMap Probe Panels

Prebuilt DNA probe panels for GLMap β€” a training-free framework for profiling genomic language models by their likelihood responses over a fixed panel of DNA sequences.

Files

FileDescription
main_panel.parquet10,000-probe biological panel (14 functional elements Γ— 4 categories)
panel_manifest.jsonPer-element / per-dataset counts and sampling seeds
panel_summary.mdHuman-readable panel composition summary

Usage

import glmap
panel = glmap.load_panel()   # auto-downloads from this dataset if not local

Or directly:

import pandas as pd
from huggingface_hub import hf_hub_download
path = hf_hub_download("Tim419/GLMap-panels", "main_panel.parquet", repo_type="dataset")
panel = pd.read_parquet(path)

License

CC-BY-NC-SA-4.0. The panel was assembled from three upstream benchmarks:

  • GUE (DNABERT-2) β€” Apache-2.0
  • Plant Genomic Benchmark (ArgoNT) β€” CC-BY-NC-SA-4.0
  • DNA Foundation Benchmark (DFB) β€” Apache-2.0

Because the panel contains 1,600 probes drawn from PGB, the combined artefact inherits PGB's CC-BY-NC-SA-4.0 license (ShareAlike). For non-commercial research use with attribution. Please cite all three upstream benchmarks plus the GLMap paper.

benchmark
genomics
TI

Tim419/GLMap-panels

Dataset

GLMap Probe Panels

0

10 commits

1 linked in READMEs

updated Jun 2, 2026

See the code

README

GLMap Probe Panels

Prebuilt DNA probe panels for GLMap β€” a training-free framework for profiling genomic language models by their likelihood responses over a fixed panel of DNA sequences.

Files

FileDescription
main_panel.parquet10,000-probe biological panel (14 functional elements Γ— 4 categories)
panel_manifest.jsonPer-element / per-dataset counts and sampling seeds
panel_summary.mdHuman-readable panel composition summary

Usage

import glmap
panel = glmap.load_panel()   # auto-downloads from this dataset if not local

Or directly:

import pandas as pd
from huggingface_hub import hf_hub_download
path = hf_hub_download("Tim419/GLMap-panels", "main_panel.parquet", repo_type="dataset")
panel = pd.read_parquet(path)

License

CC-BY-NC-SA-4.0. The panel was assembled from three upstream benchmarks:

  • GUE (DNABERT-2) β€” Apache-2.0
  • Plant Genomic Benchmark (ArgoNT) β€” CC-BY-NC-SA-4.0
  • DNA Foundation Benchmark (DFB) β€” Apache-2.0

Because the panel contains 1,600 probes drawn from PGB, the combined artefact inherits PGB's CC-BY-NC-SA-4.0 license (ShareAlike). For non-commercial research use with attribution. Please cite all three upstream benchmarks plus the GLMap paper.

benchmark
genomics