An efficient framework for protein-protein interaction prediction by integrating stacked denoising autoencoders and random ferns

Protein-protein interactions (PPIs) are crucial for understanding disease and discovering drug targets. To overcome the limitations of experimental methods, we propose SDAERFs, a computational framework that predicts PPIs from protein sequences. It leverages evolutionary information in position-specific scoring matrices (PSSMs), employs a stacked denoising autoencoder (SDAE) for feature extraction, and uses a Random Ferns (RFs) classifier for prediction. Extensive validation on benchmark datasets yielded high accuracies of 98.13% and 98.60%. Comprehensive comparisons confirmed the model's superior performance. SDAERFs provides an efficient and reliable tool for advancing PPI prediction and therapeutic development.