Perturbation and Repository based Diversified Cuckoo Search in reconstruction of Gene Regulatory Network: A new Cuckoo Search approach

Abstract

Cuckoo Search (CS), a population based meta-heuristic technique is motivated by the breeding characteristics of cuckoos. The main disadvantage of CS is its local stagnation due to weak exploration capability. To improve the exploration capacity and maintain appropriate balance between exploration and exploitation, this paper proposes a new variant of Cuckoo Search algorithm, namely, Perturbation and Repository based Diversified Cuckoo Search (PRDCS). It incorporates two modifications: Memorisation of the set of best solutions and four single-component successive Perturbations. For Memorisation, PRDCS uses a Repository to store solutions of the population having the same best Cost value. Any member (best solution) from the Repository is selected randomly for applying four single-component successive Perturbations on it. PRDCS has been applied to CEC2005 benchmark suit for performance validation. The computational experiment on CEC 2005 benchmarks suit has established that PRDCS is superior to the state-of-the-art algorithms in respect of the performance metrics used. Furthermore, PRDCS has been applied to Gene Regulatory Network (GRN) reconstruction using Recurrent Neural Network (PRDCS-RNN) model and Pearson Correlation Coefficient has been computed for the reconstructed GRN to remove the false positive edges. Application of the Proposed PRDCS-RNN on two real life gene expression datasets, IRMA-OFF and IRMA-ON reveals that, it is superior to the state-of-the-art algorithms in respect of performance metrics like F-score, Sensitivity, Specificity and Precision in producing well reconstructed GRN.