CAT-CPI: Combining CNN and transformer to learn compound image features for predicting compound-protein interactions

Abstract

Compound-protein interaction (CPI) prediction is a foundational task for drug discovery, which process is time-consuming and costly. The effectiveness of CPI prediction can be greatly improved using deep learning methods to accelerate drug development. Large number of recent research results in the field of computer vision, especially in deep learning, have proved that the position, geometry, spatial structure and other features of objects in an image can be well characterized. We propose a novel molecular image-based model named CAT-CPI (combining CNN and transformer to predict CPI) for CPI task. We use Convolution Neural Network (CNN) to learn local features of molecular images and then use transformer encoder to capture the semantic relationships of these features. To extract protein sequence feature, we propose to use a k-gram based method and obtain the semantic relationships of sub-sequences by transformer encoder. In addition, we build a Feature Relearning (FR) module to learn interaction features of compounds and proteins. We evaluated CAT-CPI on three benchmark datasets—Human, Celegans, and Davis—and the experimental results demonstrate that CAT-CPI presents competitive performance against state-of-the-art predictors. In addition, we carry out Drug-Drug Interaction (DDI) experiments to verify the strong potential of the methods based on molecular images and FR module.

Multilevel Attention Network with Semi-supervised Domain Adaptation for Drug-Target Prediction

“…The simplest approach is to concatenate the features ( Öztürk, Özgür, and Ozkirimli 2018;Lee, Keum, and Nam 2019;Zheng et al 2020;Nguyen et al 2021) and pass them through a Fully-Connected Network (FCN) to obtain the prediction results. Another approach (Qian, Wu, and Zhang 2022) is to overlap the feature maps and use CNN to extract interaction features. However, these methods lack interpretability and overlook the inherent structure of interactions.…”

“…These models commonly utilize 1D-CNN ( Öztürk, Özgür, and Ozkirimli 2018;Lee, Keum, and Nam 2019;Zhao et al 2022;Bai et al 2023) or transformer architectures (Chen et al 2020;. Secondly, drug molecules can be represented as graphs (Nguyen et al 2021;Tsubaki, Tomii, and Sese 2019; or images (Qian, Wu, and Zhang 2022). Similarly, protein distance maps can serve as a 2D abstraction of their 3D structural information (Zheng et al 2020), enabling the use of Graph Neural Networks (GNNs) (Scarselli et al 2008), Graph Convolutional Networks (GCNs) (Kipf and Welling 2016), and Convolutional Neural Networks (CNNs).…”

Revolutionizing Medicinal Chemistry: The Application of Artificial Intelligence (AI) in Early Drug Discovery

“…For example, in the ligand-based virtual screening approach, the L3D-PLS model employs CNN to extract crucial interaction features from grids surrounding aligned ligands, outperforming traditional methods in the lead optimization of small datasets [ 69 ]. In another application, the CAT–CPI model combines CNN with transformers to improve the prediction of compound-protein interactions, accelerating drug development [ 70 ]. Additionally, the FRSite method uses a faster R-CNN-based approach to accurately predict protein binding sites, introducing multi-source 3D data and RPN-3D networks to simultaneously predict the center and size of the binding site [ 71 ].…”

Attention is all you need: utilizing attention in AI-enabled drug discovery

“…One research endeavor introduces DISAE , a deep learning framework that leverages evolutionary insights and self-supervised learning to predict chemical binding onto poorly annotated proteins [ 129 ]. The CAT-CPI model combines CNN and transformer encoders to enhance molecular image learning and protein sequence representation [ 130 ]. CAT-CPI utilizes Feature Relearning to capture interaction features, achieving optimal outcomes and extending its application to DDI tasks.…”