The paper presents a Bayesian network approach to predict protein-protein interactions (PPIs) from genomic data in yeast. The method combines weakly associated genomic features, such as mRNA co-expression, co-essentiality, and co-localization, to generate reliable predictions. The approach can integrate noisy experimental interaction datasets and predict PPIs de novo. The authors validate their predictions using TAP-tagging experiments and show that their method outperforms existing high-throughput experimental datasets in terms of accuracy. The Bayesian network allows for the probabilistic combination of multiple datasets and naturally weights each information source according to its reliability. The method is applied to yeast and provides a comprehensive view of known and putative protein complexes, with a focus on large complexes and potential new interactions. The procedure is also adaptable to other organisms and can handle various types of genomic data, making it a powerful tool for predicting PPIs.The paper presents a Bayesian network approach to predict protein-protein interactions (PPIs) from genomic data in yeast. The method combines weakly associated genomic features, such as mRNA co-expression, co-essentiality, and co-localization, to generate reliable predictions. The approach can integrate noisy experimental interaction datasets and predict PPIs de novo. The authors validate their predictions using TAP-tagging experiments and show that their method outperforms existing high-throughput experimental datasets in terms of accuracy. The Bayesian network allows for the probabilistic combination of multiple datasets and naturally weights each information source according to its reliability. The method is applied to yeast and provides a comprehensive view of known and putative protein complexes, with a focus on large complexes and potential new interactions. The procedure is also adaptable to other organisms and can handle various types of genomic data, making it a powerful tool for predicting PPIs.