Related Experiment Video
Updated: Jan 12, 2026
09:43
Identification of Plant Ice-binding Proteins Through Assessment of Ice-recrystallization Inhibition and Isolation Using Ice-affinity Purification
Published on: May 5, 2017
9.1K
AFPDeepPred: A Deep Learning Framework for Accurate Identification of Antifreeze Proteins.
Xingqiao Lin1,2, Jiahui Guan3, Feng Wang2
1School of Informatics, Xiamen University, 361005 Xiamen, China.
Journal of Chemical Information and Modeling
|October 31, 2025
Summary
Antifreeze proteins (AFPs) are crucial for survival in cold environments. A new deep learning model, AFPDeepPred, accurately identifies these proteins by integrating global and local sequence data, achieving 93.46% accuracy.
Area of Science:
- Biochemistry
- Bioinformatics
- Computational Biology
Background:
- Antifreeze proteins (AFPs) are vital for organisms surviving in subzero temperatures.
- AFPs have broad applications in biomedicine and agriculture.
- High sequence diversity in AFPs complicates accurate, high-throughput identification.
Purpose of the Study:
- To develop a novel multimodal deep learning framework, AFPDeepPred, for accurate identification of antifreeze proteins.
- To integrate both global and local sequence information for enhanced AFP prediction.
Main Methods:
- Proposed AFPDeepPred, a multimodal deep learning framework.
- Integrated evolutionary scale modeling (global features) and chaos game representation (local features).
- Utilized a bilinear attention network for fusing heterogeneous sequence features.
Main Results:
- Achieved state-of-the-art performance on Swiss-Prot-based datasets with 93.46% accuracy.
- Demonstrated superior performance across all evaluation metrics compared to existing methods.
- Showcased robustness and generalizability in AFP identification.
Conclusions:
- AFPDeepPred offers a robust and accurate solution for high-throughput AFP identification.
- The multimodal approach effectively leverages both global and local sequence characteristics.
- The framework holds significant potential for advancing AFP research and applications.
