Prediction of Membrane Protein Interaction Based on Deep Residual Learning

ADH-PPI: An attention-based deep hybrid model for protein-protein interaction prediction

“…From Yeast specie, performance of ADH-PPI is evaluated on a well known public benchmark dataset namely Saccharomyces cerevisiae (S.cerevisiae), which is extensively utilized by several researchers for PPI prediction ( Yu et al., 2021 ; Jiang et al., 2020 ; Wang et al., 2020a ). PPIs of Saccharomyces cerevisiae (S.cerevisiae) were first extracted by ( Guo et al., 2008b ) from Database of Interacting Proteins (DIP): RRID: SCR_003167 ; https://dip.doe-mbi.ucla.edu/dip/Main.cgi ( Xenarios et al., 2002 ).…”

“…In third approach, negative dataset was generated using data augmentation approach. Another widely used PPI prediction dataset ( Yu et al., 2021 ; Jiang et al., 2020 ; Wang et al., 2020a ) Helicobacter pylori belongs to Bacterium specie, which was compiled by ( Martin et al., 2005 ). It contained 2,916 protein pairs out of which 1,458 protein pairs were positive and 1,458 protein pairs were negative.…”

“…It is evident that protein sequences have high length variability. Considering machine learning approaches require fixed length protein sequences, existing PPI prediction approaches transform variable length protein sequences into fixed length sequences using traditional copy padding or sequence truncation approaches ( Yu et al., 2021 ; Jiang et al., 2020 ; Wang et al., 2020a ). We perform experimentation with 6 different variations of copy padding and sequence truncation approaches in order to quantify their efficacy for PPI prediction.…”

“…Following the evaluation criterion of previous PPI predictors ( Yu et al., 2021 ; Jiang et al., 2020 ; Wang et al., 2020a ), to perform a fair performance comparison of proposed ADH-PPI approach with existing PPI predictors, we use 6 different evaluation measures namely accuracy (ACC), precision, recall, matthews correlation coefficient (MCC), F1-score, and area under receiver operating characteristics (AU-ROC). …”

“…To date, several raw sequence-based machine and deep learning-based approaches have also been proposed ( Yu et al., 2021 ; Jiang et al., 2020 ; Wang et al., 2020a ) for PPI prediction. For example, most recently, ( Yu et al., 2021 ) proposed a machine learning-based PPI predictor GcForest-PPI.…”