WUREN: Whole-modal fUsion Representation for protein interaction interfacE predictioN
Xiaodong, W.; Xiangrui, G.; Xuezhe, F.; Zhe, H.; Mengcheng, Y.; Tianyuan, W.; XIaolu, H.; Lipeng, L.
Show abstract
Proteins are one of the most important components in life, and the research on protein complex and the development of protein or antibody drugs relies on effective representation of proteins. Both experimental methods like cryo-electron microscopy and computational methods like molecular dynamic simulation suffer from high cost, long time investment and low throughput, and cannot be used in large-scale studies. Some examples of artificial intelligence for protein complex prediction tasks show that different representations of proteins have their own limitations. This paper constructs a multimodal model named WUREN (Whole-modal fUsion Representation for protein interaction interfacE predictioN), which effectively fuses sequence, graph, and structural features. WUREN has achieved state-of-the-art performance on both the antigen epitope prediction task and the protein-protein interaction interface prediction task, with AUC-PR reaching 0.462 and 0.516, respectively. Our results show that WUREN is a general and effective feature extraction model for protein complex, which can be used in the development of protein-based drugs. Furthermore, the general framework in WUREN can be potentially applied to model similar biologics to proteins, such as DNA and RNA.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- GraphGPSM: a global scoring model for protein structure using graph neural networks 97%
- OPUS-Rota4: A Gradient-Based Protein Side-Chain Modeling Framework Assisted by Deep Learning-Based Predictors 97%
- Improved protein contact prediction using dimensional hybrid residual networks and singularity enhanced loss function 96%
Similar papers in this journal
- To Improve Protein Sequence Profile Prediction through Image Captioning on Pairwise Residue Distance Map 98%
- E2EDA: Protein domain assembly based on end-to-end deep learning 96%
- ProAffinity-GNN: A Novel Approach to Structure-based Protein-Protein Binding Affinity Prediction via a Curated Dataset and Graph Neural Networks 96%
Similar papers in this journal
- Accurate prediction of protein torsion angles using evolutionary signatures and recurrent neural network 97%
- Deep Learning for Protein Peptide bindingPrediction: Incorporating Sequence, Structural andLanguage Model Features 97%
- ProteinGLUE: A multi-task benchmark suite for self-supervised protein modeling. 96%
Similar papers in this journal
- Integration of molecular coarse-grained model into geometric representation learning framework for protein-protein complex property prediction 96%
- Pretrainable Geometric Graph Neural Network for Antibody Affinity Maturation 96%
- A Fast Approach for Structural and Evolutionary Analysis Based on Energetic Profile Protein Comparison 95%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.