MCANet: shared-weight-based MultiheadCrossAttention network for drug–target interaction prediction

Abstract

Accurate and effective drug–target interaction (DTI) prediction can greatly shorten the drug development lifecycle and reduce the cost of drug development. In the deep-learning-based paradigm for predicting DTI, robust drug and protein feature representations and their interaction features play a key role in improving the accuracy of DTI prediction. Additionally, the class imbalance problem and the overfitting problem in the drug–target dataset can also affect the prediction accuracy, and reducing the consumption of computational resources and speeding up the training process are also critical considerations. In this paper, we propose shared-weight-based MultiheadCrossAttention, a precise and concise attention mechanism that can establish the association between target and drug, making our models more accurate and faster. Then, we use the cross-attention mechanism to construct two models: MCANet and MCANet-B. In MCANet, the cross-attention mechanism is used to extract the interaction features between drugs and proteins for improving the feature representation ability of drugs and proteins, and the PolyLoss loss function is applied to alleviate the overfitting problem and the class imbalance problem in the drug–target dataset. In MCANet-B, the robustness of the model is improved by combining multiple MCANet models and prediction accuracy further increases. We train and evaluate our proposed methods on six public drug–target datasets and achieve state-of-the-art results. In comparison with other baselines, MCANet saves considerable computational resources while maintaining accuracy in the leading position; however, MCANet-B greatly improves prediction accuracy by combining multiple models while maintaining a balance between computational resource consumption and prediction accuracy.

MSI-DTI: predicting drug-target interaction based on multi-source information and multi-head self-attention

“…MCANet [ 48 ]: This model proposes a shared weights-based PolyLoss cross-attention algorithm, which extracts the interaction features of drug-target pairs to highlight the features of binding regions for a more robust feature representation. The PolyLoss loss function is applied to alleviate the overfitting problem and category imbalance problem in the drug-target dataset.…”

MolLoG: A Molecular Level Interpretability Model Bridging Local to Global for Predicting Drug Target Interactions

“…These challenges compel even the most advanced models to continue using methods such as the CNN, RNN, 23 and LSTM 24 to extract protein features from sequences, as seen in models like MolTrans, 25 DeepPurpose, 26 DrugBAN, 27 and MCANet. 28 Although the CNN can extract local aggregated features akin to graph structures, the emergence of the gradient vanishing problem results in some shallow prior knowledge often getting lost with the deepening of the model. Balancing the incorporation of global semantic and intrinsic biological background feature embeddings while considering richer local aggregated features has become a complex issue.…”

“…Furthermore, despite the significant progress AlphaFold from DeepMind has made in predicting protein 3D structures, recently generating predictions of 2 billion protein 3D structures from a million species, the reality is that the number of high-accuracy 3D structure proteins only account for a small fraction of known protein sequences. These challenges compel even the most advanced models to continue using methods such as the CNN, RNN, and LSTM to extract protein features from sequences, as seen in models like MolTrans, DeepPurpose, DrugBAN, and MCANet . Although the CNN can extract local aggregated features akin to graph structures, the emergence of the gradient vanishing problem results in some shallow prior knowledge often getting lost with the deepening of the model.…”

Attention-based approach to predict drug-target interactions across seven target superfamilies

“…Some of the existing methods do not restrict their use of attention to only generating embeddings. DTITR [17] and MCANet [18] successfully demonstrated how cross-attention can be applied across the descriptors of protein targets and compounds for improved interaction predictions. Nevertheless, these methods rely solely on SMILES and AA sequences to generate descriptors for compounds and protein targets, respectively.…”