A Neighborhood Selection Learning Artificial Bee Colony Algorithm Based on Population Backtracking for Detecting Epistatic Interactions

Complex diseases are presently the primary health challenges affecting humanity, with their pathogenesis being intricate and often involving multiple factors such as environmental influences and genetics. Studies have shown that detecting epistatic interactions is crucial for uncovering the pathogenic mechanisms underlying complex diseases. Therefore, this paper proposes a neighborhood selection learning artificial bee colony algorithm based on population backtracking (PBNSLABC) to detect epistatic interactions. Firstly, a neighborhood selection learning strategy and a novel updating mechanism are introduced to enhance the exploitation capability of PBNSLABC. Secondly, a population backtracking strategy is employed to optimize the utilization of search resources. Finally, two objective functions are employed to quantitatively assess the quality of epistatic interactions. To evaluate the performance of PBNSLABC, comparisons were conducted with five advanced epistasis detection methods on simulated datasets, demonstrating its strong detection capability. Most epistatic interactions identified by PBNSLABC in real datasets have been validated as being associated with the target disease. Therefore, PBNSLABC is competitive in detecting epistatic interactions.