A Teaching-Learning-Based Optimization for Operon Prediction

Abstract

Operons are the basic unit of transcription and can be used to understand the transcription regulation in a given prokaryotic genome. Currently, the sequence and gene coordinates of organisms can be rapidly identified, but their operons remain unknown. Moreover, the experimental methods detecting operons are extremely difficult and time-consuming to execute. Operon prediction as pretreatment can greatly reduce the cost of performing an experimental assay. Many algorithms and biological properties have been proposed but the resulting predictions still require improvement in terms of sensitivity, specificity, and accuracy. This study uses a teaching-learning-based optimization (TLBO) algorithm with three biological properties for operon prediction: the intergenic distance, the metabolic pathway, and the cluster of orthologous groups (COG). These properties for the Escherichia coli genome are used to train the evaluation standards of fitness function of gene pairs. The experimental results use the accuracy (ACC), sensitivity (SN) and specificity (SP) to evaluate our prediction method and as a basis for comparison with other methods to validate that the proposed method can effectively solve operon prediction problems.