Multivariate Information Fusion for Identifying Antifungal Peptides with Hilbert-Schmidt Independence Criterion

Background: Antifungal Peptides (AFP) have been found to be effective against many fungal infections. Objective: However, it is difficult to identify AFP. Therefore, it is great practical significance to identify AFP via machine learning methods (with sequence information). Methods: In this study, a Multi-Kernel Support Vector Machine (MKSVM) with Hilbert-Schmidt Independence Criterion (HSIC) is proposed. Proteins are encoded with five types of features (188-bit, AAC, ASDC, CKSAAP, DPC), and then construct kernels using Gaussian kernel function. HSIC are used to combine kernels and multi-kernel SVM model is built. Results: Our model performed well on three AFPs datasets and the performance is better than or comparable to other state-of-art predictive models. Conclusion: Our method will be a useful tool for identifying antifungal peptides.