A metric and its derived protein network for evaluation of ortholog database inconsistency

Abstract Ortholog prediction, essential for various genomic research areas, faces growing inconsistencies amidst the expanding array of ortholog databases. The common strategy of computing consensus orthologs introduces additional arbitrariness, underscoring the need to identify proteins prone to ortholog prediction inconsistency. To address this, we introduce the Signal Jaccard Index (SJI), a novel metric based on unsupervised genome context clustering, to assess protein similarity. Utilizing SJI, we construct a protein network, revealing that proteins at the network peripheries primarily contribute to prediction inconsistency. Importantly, we show that a protein's degree centrality can gauge its assignment reliability to a consensus set, facilitating the refinement of ortholog predictions.