1Laboratory of Biochemistry and Environmental Toxicology, Department of Biochemistry, Badji Mokhtar University, Annaba, B.P 12, 23000, Algeria
2Collège Boréal, Campus De Toronto, Ontario, Canada
3LISCO Laboratory Department of Computer Science, Badji Mokhtar University, Annaba, B.P 12, 23000, Algeria
G–protein coupled receptors (GPCRs) are key factors in cell-to-cell communication. GPCR activation is necessary for normal physiology of all organisms while dysfunction of GPCR signalling is responsible for many of the diseases. Consequently, GPCRs have a fundamental role in pharmacological research and are targets for many drugs.
The problem is that many GPCRs remain orphans (have unknown function), they are not classified correctly, and new bioinformatics approaches are needed to address this issue. In our work, we focus on bio-inspired approaches, which are increasingly used in recent years because of their interesting inspirations from biological systems mechanisms and their good performances in many research areas.
In this article, we use categories of bio-inspired well-known methods to identify GPCR function, which are swarm-based approaches and immunological computing. The proposed classifiers based on three popular swarm intelligence approaches are Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO) and PSO/ACO hybridization. The classification results are compared with these of the proposed immunological classifier based on the Artificial Immune Recognition System (AIRS), in order to identify the best bio-inspired method for the given problem.
The immune classifier (AIRS2) provided better results than swarm-based classifiers, specifically at the first levels (superfamily and families).
It is interesting to adapt the bio-inspired algorithms in order to increase predictive accuracy at all GPCR hierarchical levels.
Computational Biology, Ant Colony Optimization, Particle Swarm Optimization, PSO/ACO Hybridization, Immunological Computation, AIRS, Classification GPCR
