A GHKNN model based on the physicochemical property extraction method to identify SNARE proteins

Area of Science:

• Biochemistry
• Molecular Biology
• Bioinformatics

Background:

• SNARE proteins are essential membrane fusion proteins critical for vesicle transport.
• Dysfunctional SNARE proteins are implicated in various human diseases.
• Accurate identification methods for SNARE proteins are vital for research and diagnostics.

Purpose of the Study:

• To develop and validate a highly accurate computational model for identifying SNARE proteins.
• To establish a robust method for analyzing protein sequences and their functions.

Main Methods:

• Utilized a graph-regularized k-local hyperplane distance nearest neighbor (GHKNN) binary classification model.
• Employed physicochemical property extraction for protein sequence feature engineering.
• Applied the SMOTE (Synthetic Minority Over-sampling Technique) method for feature upsampling.

Main Results:

• The GHKNN model demonstrated superior accuracy in identifying all tested protein sequences.
• Comparative analysis showed GHKNN significantly outperformed other classification models.
• Performance was rigorously evaluated using SN, SP, ACC, and MCC metrics.

Conclusions:

• The developed GHKNN model offers a highly accurate and effective approach for SNARE protein identification.
• This method holds promise for advancing research in membrane trafficking and disease-associated studies.
• Physicochemical properties and SMOTE-based feature engineering enhance classification performance.