Structural Asymmetries in PPI Networks as a Tool to Improve the Detection

This dataset contains newly constructed protein–protein interaction (PPI) networks for Saccharomyces cerevisiae and Homo sapiens derived from current releases of the BioGRID and STRING databases. The yeast (BioGRID24Y) and human (BioGRID24H) networks include only physical interactions, while a weighted human network (STRING24H) was created using high-confidence STRING interactions (confidence score ≥ 0.700) with confidence scores assigned as edge weights. All networks were analyzed on their giant connected components and are intended for reproducible benchmarking and structural analysis of protein complex detection methods. In addition to these networks, we constructed reference sets derived from Gene Ontology (GO) annotations, restricted exclusively to child terms of the Protein-Containing Complex (PCC). For Saccharomyces cerevisiae, this reference set is denoted SGD24Y, and for Homo sapiens as SGD24H. Only PCC descendant terms were retained to ensure specificity to experimentally and computationally defined protein complexes. These GO-based references provide an independent, ontology-driven benchmark for evaluating the correspondence between detected network structures and curated protein complexes.