Employing a novel 2-gram subgroup intra pattern (2GSIP) with stacked auto encoder for membrane protein...

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Cell membrane proteins are crucial for cellular behavior.
• Amino acid sequence analysis offers insights into protein structure and function.
• A challenge in amino acid analysis is creating effective digital coding systems that capture amino acid properties and degeneracy.

Purpose of the Study:

• To develop a novel representation for protein sequences.
• To introduce a new feature called '2-gram subgroup intra pattern'.
• To improve the classification of membrane protein functional types for better understanding of biological functions.

Main Methods:

• A novel protein sequence representation incorporating the '2-gram subgroup intra pattern' feature.
• Application of a Stacked Auto Encoder deep learning method for classification.
• Evaluation on two benchmark datasets using various performance metrics.

Main Results:

• The proposed method demonstrated superior performance compared to existing techniques.
• Outperformed other methods in Self-consistency test, Accuracy, Specificity, Sensitivity, Mathew's correlation coefficient, Jackknife test, and Independent data set evaluations.
• The novel representation effectively captures essential information for membrane protein classification.

Conclusions:

• The proposed 2-gram subgroup intra pattern offers an effective digital coding system for amino acid sequences.
• The Stacked Auto Encoder deep learning model, combined with this novel representation, significantly enhances membrane protein classification.
• This approach provides a valuable tool for elucidating the biological functions of membrane proteins.