Sequence Data

Over the years, several methods have been proposed for the computational PPI prediction with different performance evaluation strategies. These methods have limitations over different issues such as ill-treated cross-validation design, positive and negative sample selection strategies etc. To address these issues, we propose a multi-level feature based approach (<i>JUPPI</i>) using sequence, domain and GO information for PPI prediction, and an evaluation strategy that, 1) filters high quality negative data, and 2) introduces a pair-input based cross validation strategy with three difficulty levels for prediction of test sets. Our proposed evaluation strategy reduces the component-level overlapping issue in test sets. The proposed method has shown a strong confidence for PPI prediction in both test and train sets in strict categories of test sets. Performance of <i>JUPPI </i>is compared with <i>state-of-the-art </i>approaches and tested on six independent PPI datasets. In almost all the datasets, <i>JUPPI </i>outperforms the <i>state-of-the-art </i>not only at human proteome level for PPI prediction, but also for prediction of interactors for intrinsic disordered human proteins.