Machine learning on protein–protein interaction prediction: models, challenges and trends

Abstract

Protein–protein interactions (PPIs) carry out the cellular processes of all living organisms. Experimental methods for PPI detection suffer from high cost and false-positive rate, hence efficient computational methods are highly desirable for facilitating PPI detection. In recent years, benefiting from the enormous amount of protein data produced by advanced high-throughput technologies, machine learning models have been well developed in the field of PPI prediction. In this paper, we present a comprehensive survey of the recently proposed machine learning-based prediction methods. The machine learning models applied in these methods and details of protein data representation are also outlined. To understand the potential improvements in PPI prediction, we discuss the trend in the development of machine learning-based methods. Finally, we highlight potential directions in PPI prediction, such as the use of computationally predicted protein structures to extend the data source for machine learning models. This review is supposed to serve as a companion for further improvements in this field.

Leveraging machine learning models for peptide–protein interaction prediction

“…This approach allows for the direct observation of information from protein 3D structures without involving hand-crafted features. 24,90 GCNs 91,92 are a type of neural network that can be used to learn graph embeddings. Similar to CNNs, GCNs take graph embeddings as input and progressively transform them through a series of localized convolutional and pooling layers where each layer updates all vertex features.…”

Machine Learning Empowering Drug Discovery: Applications, Opportunities and Challenges

“…PPIs possess indispensable functions in diverse biological processes. They can contribute to altering protein specificity, regulating protein activity and generating novel binding sites for effector molecules [ 90 ]. Hence, understanding and targeting PPIs offers opportunities to design innovative drugs that can modulate complex biological processes.…”

“…Therefore, ML-based approaches have great potential in enhancing the identification of PPI sites. Compared with sequence-based approaches, structure-based ones are limited by the scarcity of available protein structures and the low quality of familiar protein structures [ 90 , 93 ].…”

TUnA: An uncertainty aware transformer model for sequence-based protein-protein interaction prediction

“…Recently, deep learning (DL) methods have emerged as a promising alternative. 4 While protein structure is critical for protein binding, DL models primarily relying on protein sequence, given its relative abundance over structural data, have achieved impressive performance. 5,6 For example, PIPR 5 utilizes a siamese recurrent convolutional neural network to capture local and sequential features such as co-occurrence similarity of amino acids and electrostaticity and hydrophobicity based features.…”