Skip to content

examples

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

examples

README.md

Examples

Quick reference for all biosieve splitting and reduction strategies.

Directory layout

examples/
├── configs/                    # YAML parameter files (one per strategy)
├── raw_data_examples/          # Demo data files (CSV, npy)
├── generating_random_datasets.py          # Generate base demo dataset
├── generate_demo_descriptors.py           # Add numerical descriptors columns
├── generate_demo_distances_structures.py  # Generate structural distance edges
├── prepare_cluster_mapping.py             # Extract cluster mapping CSV
├── prepare_embedding_for_reductions.py    # Convert ESM2 CSV → npy + ids CSV
├── testing_splitters.sh        # Run all split strategies
├── testing_reducers.sh         # Run all reduction strategies
└── using_sylphy_to_get_demo_embedding.sh  # Fetch ESM2 embeddings via Sylphy

Setup: generate demo data

All scripts must be run from the raw_data_examples/ directory.

cd raw_data_examples/

# 1. Base dataset (1000 peptides)
python ../generating_random_datasets.py

# 2. Dataset with descriptor columns (needed for descriptor-based strategies)
python ../generate_demo_descriptors.py

# 3. Structural distance edges (needed for structural_distance reducer)
python ../generate_demo_distances_structures.py

# 4. Cluster mapping CSV (needed for cluster_aware splitter with external mapping)
python ../prepare_cluster_mapping.py

# 5. Embeddings (needed for embedding-based strategies)
#    Option A: use Sylphy to generate real ESM2 embeddings
bash ../using_sylphy_to_get_demo_embedding.sh
python ../prepare_embedding_for_reductions.py

Running the examples

cd raw_data_examples/

# All split strategies
bash ../testing_splitters.sh

# All reduction strategies
bash ../testing_reducers.sh

Configs (configs/)

File Strategy
params_split_random.yaml random
params_split_random_kfold.yaml random_kfold
params_split_stratified.yaml stratified
params_split_stratified_kfold.yaml stratified_kfold
params_split_stratified_numerical.yaml stratified_numeric
params_split_stratified_num_kfold.yaml stratified_numeric_kfold
params_split_group.yaml group
params_split_group_kfold.yaml group_kfold
params_split_time.yaml time
params_split_distance_aware.yaml distance_aware (embeddings)
params_split_distance_aware_descriptors.yaml distance_aware (descriptors)
params_split_distance_aware_embedding_kfold.yaml distance_aware_kfold (embeddings)
params_split_distance_aware_desc_kfold.yaml distance_aware_kfold (descriptors)
params_split_homology_aware.yaml homology_aware (runs MMseqs2)
params_split_homology_aware_precomputed.yaml homology_aware (precomputed clusters)
params_split_cluster_aware.yaml cluster_aware
params_split_cluster_aware_with_mapping.yaml cluster_aware (external mapping file)
params_reducer.yaml embedding_cosine reducer
params_reducer_minhash_jaccard.yaml minhash_jaccard reducer

CLI reference

# Split
biosieve split -i <input.csv> -o <outdir/> --strategy <name> --params <params.yaml>

# Reduce
biosieve reduce -i <input.csv> -o <output.csv> --strategy <name> \
  [--mapping-output map.csv] [--report-output report.json] [--params <params.yaml>]

# Override a param inline (no YAML needed)
biosieve split -i data.csv -o out/ --strategy random --set random.seed=42