API.md

API Reference

Every public class and function in syndrumnet, grouped by subpackage in the order the pipeline uses them: load and parse data, build the network and modules, compute metrics, propagate, score, evaluate, visualise.

This page is a map. Full parameter and return documentation lives in the docstrings themselves, in NumPy style, and is the authoritative source:

from syndrumnet.propagation.prince import PRINCE
help(PRINCE)

Regenerate this file after changing the public surface:

python scripts/gen_api_docs.py > docs/API.md

---

syndrumnet.io

syndrumnet.io.downloaders

• DataDownloader - Centralized data downloader for SyndrumNET pipeline.
  • methods: download_file(), download_huri(), download_corum(), download_phosphositeplus(), download_creeds(), download_lincs(), download_disease_genes(), download_id_mapping(), download_all(), get_file_paths()

syndrumnet.io.id_mapping

• IDMapper - Gene/protein ID mapping service.
  • methods: cache_file, load_cache(), save_cache(), to_hgnc(), to_entrez(), harmonize_gene_list(), batch_convert()

syndrumnet.io.parsers

• parse_huri(filepath) - Parse HuRI protein-protein interaction data.
• parse_corum(filepath) - Parse CORUM protein complex data.
• parse_phosphositeplus(filepath) - Parse PhosphoSitePlus kinase-substrate data.
• parse_kegg_rpair(filepath) - Parse KEGG RPair reaction data.
• parse_creeds(filepath) - Parse CREEDS disease signatures.
• parse_lincs(sig_filepath, meta_filepath, top_pct, aggregate) - Parse LINCS L1000 drug signatures.

syndrumnet.data

syndrumnet.data.modules

• ModuleBuilder - Construct disease and drug modules.
  • methods: build_disease_modules(), build_drug_modules(), save_modules(), load_modules()

syndrumnet.data.network_builder

• NetworkBuilder - Build integrated human molecular interaction network.
  • methods: add_source(), build(), get_network_stats(), save(), load()

syndrumnet.metrics

syndrumnet.metrics.distances

• shortest_path_distance(G, source_set, target_set, infinity_value, exclude_self) - Compute average shortest path distance from source set to target set.
• module_proximity(G, module_a, module_b) - Compute bidirectional proximity between two modules.
• separation_score(G, module_a, module_b) - Compute network separation s_AB between two modules.
• compute_all_pairwise_distances(G, gene_set) - Compute all pairwise shortest path distances within a gene set.

syndrumnet.metrics.null_models

• degree_preserving_randomization(G, module, n_random, seed) - Generate degree-preserving random gene sets.
• compute_zscore(observed, null_distribution) - Compute z-score of observed value against null distribution.
• compute_normalized_proximity(G, disease_module, drug_module, n_random, seed) - Compute z-score normalized proximity between disease and drug modules.

syndrumnet.metrics.transcription

• compute_correlation(signature_a, signature_b, method) - Compute correlation between two gene expression signatures.
• transcriptional_similarity(disease_signature, drug_signature_up, drug_signature_down, inverse_correlation) - Compute transcriptional similarity between disease and drug.
• aggregate_transcriptional_scores(scores, method) - Aggregate multiple transcriptional scores.

syndrumnet.propagation

syndrumnet.propagation.prince

• PRINCE - PRINCE network propagation algorithm.
  • methods: propagate(), propagate_multiple(), get_top_genes()

syndrumnet.propagation.similarity_layers

• compute_disease_similarity(disease_modules, method) - Compute pairwise disease similarity matrix.
• compute_drug_similarity(drug_fingerprints, method) - Compute pairwise drug similarity matrix.
• kcf_fingerprint_similarity(smiles_a, smiles_b) - Compute KCF-S fingerprint similarity between two molecules.
• jaccard_similarity(set_a, set_b) - Jaccard similarity coefficient.
• overlap_coefficient(set_a, set_b) - Overlap coefficient.
• tanimoto_similarity_matrix(fingerprints) - Compute pairwise Tanimoto similarity for binary fingerprints.
• build_similarity_matrix(entities, similarity_func) - Build pairwise similarity matrix for a list of entities.

syndrumnet.scoring

syndrumnet.scoring.cqab

• compute_cqab(disease_signature, drug_a_signature_up, drug_a_signature_down, drug_b_signature_up, drug_b_signature_down) - Compute transcriptional correlation score CQAB.
• compute_cqab_batch(disease_signature, drug_signatures, drug_pairs) - Compute CQAB for multiple drug pairs.

syndrumnet.scoring.pqab

• module_seed(base_seed, module) - Derive a deterministic null-model seed for one gene module.
• compute_pqab(G, disease_module, drug_a_module, drug_b_module, n_randomizations, seed) - Compute proximity score PQAB.
• proximity_zscore(G, disease_module, drug_module, n_randomizations, seed) - Compute the z-scored disease-drug proximity P_QA for a single drug.
• compute_pqab_batch(G, disease_module, drug_modules, drug_pairs, n_randomizations, seed, proximity_zscores) - Compute PQAB for multiple drug pairs.

syndrumnet.scoring.predictor

• SynergyPredictor - Complete SyndrumNET synergy prediction pipeline.
  • methods: set_disease_modules(), set_drug_modules(), set_disease_signatures(), predict_all(), predict_multiple_diseases(), save_predictions()

syndrumnet.scoring.tqab

• TopologyClass - The six drug-drug-disease classes of Cheng et al. (2019), Figure 2.
• classify_topology(z_qa, z_qb, s_ab) - Assign a drug pair to one of the six topological classes.
• compute_tqab(z_qa, z_qb, s_ab) - Compute the topological class score T_QAB.
• compute_tqab_batch(G, disease_module, drug_modules, drug_pairs, proximity_zscores, n_randomizations, seed) - Compute TQAB for multiple drug pairs.

syndrumnet.eval

syndrumnet.eval.benchmarks

• load_known_synergies(filepath, disease_filter) - Load known synergistic drug combinations.
• load_drugcombdb(filepath) - Load DrugCombDB database.

syndrumnet.eval.metrics

• compute_auc(y_true, y_score) - Compute AUC-ROC.
• compute_pr(y_true, y_score) - Compute AUC-PR (average precision).
• compute_roc_curve(y_true, y_score) - Compute ROC curve.
• compute_precision_recall_curve(y_true, y_score) - Compute precision-recall curve.
• evaluate_predictions(predictions, known_synergies) - Evaluate predictions against known synergies.

syndrumnet.eval.reporting

• generate_evaluation_report(results, output_path) - Generate evaluation summary report.

syndrumnet.viz

syndrumnet.viz.plots

• plot_degree_distribution(G, output_path, log_scale, dpi) - Plot network degree distribution.
• plot_roc_curve(fpr, tpr, auc, output_path, title, dpi) - Plot ROC curve.
• plot_pr_curve(precision, recall, auc_pr, output_path, title, dpi) - Plot precision-recall curve.
• plot_score_distributions(predictions, output_path, dpi) - Plot distributions of TQAB, PQAB, CQAB scores.
• plot_top_predictions(predictions, k, output_path, dpi) - Plot top-k predictions as a signed decomposition of the total score.
• plot_auc_comparison(results, output_path, dpi) - Plot AUC comparison across diseases.

syndrumnet.utils

syndrumnet.utils.config

• Config - Configuration container with nested access support.
  • methods: get(), to_dict()
• load_config(config_path) - Load configuration from YAML file.
• merge_configs(base, override) - Merge override dictionary into base config.
• save_config(config, output_path) - Save configuration to YAML file.

syndrumnet.utils.logging

• setup_logger(name, log_dir, level, console) - Setup structured logger with file and console handlers.
• LoggerMixin - Mixin to add logger attribute to classes.
  • methods: logger

syndrumnet.utils.seeds

• set_random_seed(seed) - Set random seed for all libraries to ensure reproducibility.
• get_random_state(seed) - Create a NumPy RandomState object for isolated random operations.

---

Entry points

The four scripts in scripts/ wire the above together. Each takes --config <path> and accepts dotted overrides such as --propagation.alpha 0.7.

Script	Stage
build_all_data.py	Download sources, map IDs, build the network and modules
run_pipeline.py	Compute TQAB, PQAB and CQAB for every drug pair
evaluate.py	Score predictions against known synergies (AUC-ROC, AUC-PR)
make_figures.py	Render figures into reports/figures/