Skip to content

README.md

Latest commit

 

History

History
118 lines (101 loc) · 4.34 KB

README.md

File metadata and controls

118 lines (101 loc) · 4.34 KB

Simplified Scripts (embeddings)

These commands assume:

  • Python env: source /path/to/StructTokenBench/bin/activate
  • Python: python
  • H5 embeddings: ./inference_embed_results/.../embeddings
  • Data roots:
    • Functional: ./pst_evaluation/struct_token_bench_release_data/data/functional/local
    • Structural: ./pst_evaluation/struct_token_bench_release_data/data/structural
    • Physicochemical (ATLAS): ./pst_evaluation/struct_token_bench_release_data/data/physicochemical
    • Sensitivity (Apo/Holo): ./pst_evaluation/struct_token_bench_release_data/data/sensitivity

Adjust paths if your local layout is different.

Optional shorthand:

export EMB_DIR=./inference_embed_results/.../embeddings

Datasets

  • StructTokenBench evaluation data (labels/splits used by eval scripts via --data-root): download and extract into ./pst_evaluation/struct_token_bench_release_data/
  • PST raw CIF/H5 structures (used to generate embeddings): download and extract into, then run inference_embed.py (repo root) to produce H5 embeddings for the eval scripts (--h5)
  • For PST evaluation, embeddings should exclude missing residues (no -1 rows). This matches per-residue labels, and some tasks (e.g., remote homology) do not filter -1 rows.

BioLIP2 binding

python pst_evaluation/biolip2_binding_eval.py \
  --h5 "$EMB_DIR/vq_embed_biolip2_binding_lite_model.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

BioLIP2 catalytic

python pst_evaluation/biolip2_catalytic_eval.py \
  --h5 "$EMB_DIR/vq_embed_biolip2_catalytic_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

ProteinShake binding site

python pst_evaluation/proteinshake_binding_eval.py \
  --h5 "$EMB_DIR/vq_embed_proteinshake_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

InterPro binding

python pst_evaluation/interpro_binding_eval.py \
  --h5 "$EMB_DIR/vq_embed_interpro_binding_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

InterPro active site

python pst_evaluation/interpro_activesite_eval.py \
  --h5 "$EMB_DIR/vq_embed_interpro_activesite_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

InterPro conserved site

python pst_evaluation/interpro_conserved_eval.py \
  --h5 "$EMB_DIR/vq_embed_conserved_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

InterPro repeats

python pst_evaluation/interpro_repeats_eval.py \
  --h5 "$EMB_DIR/vq_embed_interpro_repeats.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

ProteinGLUE epitope region

python pst_evaluation/proteinglue_epitope_region_eval.py \
  --h5 "$EMB_DIR/vq_embed_proteinglue_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/functional/local

ATLAS RMSF

python pst_evaluation/atlas_rmsf_eval.py \
  --h5 "$EMB_DIR/vq_embed_atlas_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/physicochemical

ATLAS B-factor

python pst_evaluation/atlas_bfactor_eval.py \
  --h5 "$EMB_DIR/vq_embed_atlas_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/physicochemical

ATLAS NEQ

python pst_evaluation/atlas_neq_eval.py \
  --h5 "$EMB_DIR/vq_embed_atlas_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/physicochemical

Remote homology

python pst_evaluation/remote_homology_eval.py \
  --h5 "$EMB_DIR/vq_embed_remote_homology_train_tst_valid.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/structural \
  --progress

Apo/Holo + Fold Switching (unsupervised)

python pst_evaluation/apolo_unsupervised_eval.py \
  --h5 "$EMB_DIR/vq_embed_apolo_lite.h5" \
  --data-root ./pst_evaluation/struct_token_bench_release_data/data/sensitivity \
  --target-field tm_score