Seeding-inspired chemotaxis genetic algorithm for the inference of biological systems

Inverse problem studies of biochemical systems with structure identification of S-systems by embedding training functions in a genetic algorithm

“…In fact studies in [27,28] , with noisy data (5%) obtained a high number of false positive (FP) interactions. Kikuchi et al [24] and Wu and Wu [32] used coupled ODE form for integration and showed results using toy data without noise. Ho et al [33] added noise in the toy data and similar to the studies with the decomposition approach obtained solutions with a high number of FPs.…”

“…It may be emphasized that handling of noisy datasets during structure identification has not been studied in detail. Preprocessing the noisy data using smoothing methods [20,31,32] or filters [35] before addressing the parameter estimation and structure identification problem have been attempted but it may result in a loss of phenomenological information.…”

“…Using evolving limits on both maximum and minimum connections allowed, Chowdhury et al [29,30] applied a penalty for connections beyond these limits which scaled appropriately following a power law expression. A multi-objective optimization approach by Liu and Wang [31] and Wu and Wu [32] minimized the penalty term used in [19] as one of the objectives. Irrespective of the choice of the penalty function, successful minimal structure identification faced difficulties when noisy data was considered [28,33] .…”

Simulink-Based Analysis for Coupled Metabolic Systems

“…Additionally, we proposed cockroach swarm evolution (CSE) which was based on the cooperatively foraging behaviour of cockroach swarms and the completive behaviour and swarm migration were treated as event-induced operations [5]. Seeding-inspired chemotaxis genetic algorithms (SCGA) were designed to be attracted to an attracter purposely and then jump from it successfully through seeding-inspired strategies and winnerchemotaxis population migration [6]. Kimura and collaborators introduced SVM-based linear programming classification and extracted gene interactive information from the classifiers through a genetic local search with a distanceindependent diversity control [7][8][9].…”