Related Experiment Video
Updated: May 24, 2026
09:47
DeepOmicsAE: Representing Signaling Modules in Alzheimer's Disease with Deep Learning Analysis of Proteomics, Metabolomics, and Clinical Data
Published on: December 15, 2023
Opt-DMA-trans: Explainable optimized residual dense-assisted multi-scale attention transformer assisted multi-class
Paritosh Kumar1, Akshay Deepak1, Prabhat Kumar1
1Department of Computer Science & Engineering, National Institute of Technology, Patna, Bihar 800005, India.
Computational Biology and Chemistry
|May 22, 2026
Summary
This study introduces an Explainable AI model to accurately classify multi-class Nucleic Acid Binding Proteins (NABP). The novel Opt-DMA-Trans model achieves high accuracy, improving upon existing methods for protein sequence analysis.
Area of Science:
- Bioinformatics
- Computational Biology
- Artificial Intelligence in Genomics
Background:
- Nucleic acid binding proteins (NABP), including DNA binding proteins (DBP) and RNA binding proteins (RBP), are crucial for gene regulation and disease pathways.
- Experimental determination of NABP is time-consuming and costly, struggling to keep pace with vast protein sequence data.
- Existing computational models often lack interpretability and struggle with complex relationships in protein sequences due to high-dimensional, redundant features.
Purpose of the Study:
- To develop an interpretable and accurate Explainable AI (XAI) model for the multi-class classification of NABP.
- To address limitations of existing models in handling complex protein sequence relationships and feature redundancy.
- To enhance the efficiency and scalability of NABP identification using advanced machine learning techniques.
Main Methods:
- Data sourced from the UniProtKB dataset with pre-processing to remove duplicates and irrelevant information.
- Feature extraction using the Python ProtLearn library, followed by feature selection via Improved Orangutan Optimization Algorithm (IOOA) and Autoencoder (AE).
- Classification performed using an optimized Residual Dense assisted Multi-Scale Attention Transformer (Opt-DMA-Trans) model, with Shapley Additive Explanations (SHAP) for interpretability.
Main Results:
- The proposed Opt-DMA-Trans model achieved high performance metrics: 98.95% accuracy, 98.95% precision, and 98.75% F1-score.
- Feature selection methods (IOOA and AE) effectively reduced redundancy and improved prediction.
- The SHAP technique provided insights into the model's decision-making process, enhancing interpretability.
Conclusions:
- The developed Explainable AI model demonstrates high efficacy and accuracy in predicting multi-class NABP.
- The integration of advanced feature selection and attention-based transformer architecture offers a robust solution for NABP classification.
- This approach provides a scalable, interpretable, and cost-effective alternative to experimental methods for identifying NABP.
