Identification of clathrin proteins by incorporating hyperparameter optimization in deep learning and PSSM profiles

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Bioinformatics
• Genomics and Proteomics

Background:

• Clathrin is vital for vesicle formation and cellular processes.
• Clathrin dysfunction is linked to neurodegenerative disorders, cancer, and Alzheimer's disease.
• Accurate identification of clathrin proteins is essential for disease research and drug development.

Purpose of the Study:

• To develop a precise deep learning model for identifying clathrin proteins.
• To utilize a two-dimensional convolutional neural network (CNN) with Position-Specific Scoring Matrix (PSSM) profiles.
• To enhance the understanding of clathrin's role in human diseases.

Main Methods:

• A deep learning model employing a 2D CNN and PSSM profiles was developed.
• PSSM profiles were treated as 20x20 pixel images for CNN input.
• Hyper-parameter optimization and 10-fold cross-validation were used to refine the model.

Main Results:

• The model achieved high performance on an independent dataset: 92.2% sensitivity, 91.2% specificity, 91.8% accuracy, and 0.83 MCC.
• Significant improvements were observed compared to traditional neural network methods.
• The model demonstrates robust predictive ability for clathrin proteins.

Conclusions:

• The developed deep learning tool effectively identifies clathrin proteins.
• This research promotes the application of deep learning in biomedical studies.
• Source code and datasets are freely available for further research.