HSIC-H-FLapSVM: A Kernel Entropy Component Analysis and Multiple Kernel Learning-based Fuzzy Laplacian SVM Model for...

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Exosomal proteins are vital in biological processes and hold significant potential for clinical diagnosis and prognosis.
• Accurate identification of exosomal proteins is challenging due to low sequence similarity in existing datasets.
• Developing robust computational methods for exosomal protein prediction is an urgent need.

Purpose of the Study:

• To propose an effective algorithmic model for accurately identifying proteins secreted by exosomes.
• To enhance the accuracy of exosomal protein prediction using integrated sequence features and machine learning.

Main Methods:

• Extraction and selection of six protein sequence feature types: PSSM-DWT, PSSM-AB, and PsePSSM.
• Integration of selected features using Multiple Kernel Learning based on the Hilbert-Schmidt Independence Criterion (MKL-HSIC).
• Derivation of fuzzy membership scores using Kernel Entropy Component Analysis (KECA) for training samples.

Main Results:

• The proposed HSIC-H-FLapSVM model demonstrated superior performance on the testing set.
• Achieved an accuracy (ACC) of 0.8623 and a Matthews Correlation Coefficient (MCC) of 0.6347.
• Outperformed all competing methods in exosomal protein prediction.

Conclusions:

• HSIC-H-FLapSVM is an effective computational tool for predicting exosomal proteins.
• The integration of PSSM-DWT, PSSM-AB, PsePSSM features with MKL-HSIC and KECA significantly improves prediction accuracy.
• This method offers a promising approach for advancing clinical diagnosis and prognosis through exosomal protein analysis.