Volume 14, Number 5

GPCR Protein Feature Representation using Discrete Wavelet Transform
and Particle Swarm Optimisation Algorithm


Nor Ashikin Mohamad Kamal1, Azuraliza Abu Bakar2 and Suhaila Zainudin2, 1Universiti Teknologi MARA, Malaysia, 2Universiti Kebangsaan Malaysia, Malaysia


Features play an important role in representing classes in the hierarchy structure, and using unsuitable features will affect classification performance. The discrete wavelet transform (DWT) approach provides the ability to create the appropriate features to represent data. DWT can produce global and local features using different wavelet families and decomposition levels. These two parameters are essential to obtain a suitable representation for classes in the hierarchy structure. This study proposes using a particle swarm optimisation (PSO) algorithm to select the suitable wavelet family and decomposition level for G-protein coupled receptor (GPCR) hierarchical class representation. The results indicate that the PSO algorithm mostly selects Biorthogonal wavelets and decomposition level 2 to represent GPCR protein. Concerning the performance, the proposed method achieved an accuracy of 97.9%, 85.9%, and 77.5% at the family, subfamily, and sub-subfamily levels, respectively.


Decomposition level, GPCR, hierarchical classification, particle swarm optimisation, wavele tfamily.