Prediction of key responsive genes to environmental stresses in Thermus thermophilus by using SVM-RFE

Various environmental agents such as redox, low and high temperature, nutrient availability, heavy metal, hydrogen peroxide, and salt stresses have negative effects on the growth and development of all organisms. Bacteria adapt and illicit a quick response against these environmental changes by changing transcriptome content. They are armed with variety of receptors and signal molecules, enabling them to protect themselves against harsh environmental changes. The identification of responsive genes to these changes is one of the main steps in better understanding of defense mechanisms. To reach this aim, we performed SVM-RFE (support vector machine-recursive feature elimination) algorithms based on meta-analysis of ۲۴ samples of microarray data from four di↵erent types of stress conditions namely cold, heat, salt, and hydrogen peroxide stresses. We modified the SVM-RFE by using bootstrapping and leave-one-out cross-validation to overcome small sample size. We analyzed ۱۵ key genes (TTHA۰۹۰۲, TTHA۰۲۶۰, TTHA۰۲۳۰, TTHB۱۵۰, TTHA۱۶۶۰, TTHB۱۶۰, TTHA۱۲۸۵, TTHA۱۲۳۳, TTHA۰۳۵۰, TTHA۱۱۵۲, TTHY۷۰۸۰, TTHC۰۱۳, TTHA۱۳۷۶, TTHA۰۷۱۸ and TTHC۰۱۰) as predicted by SVM-RFE, some of which were involved in DNA repair mechanisms and in response to abiotic stresses. Our study indicates that SVM-RFE could be utilized as a suitable machine learning method to predict key responsive genes involved in abiotic stresses.

کلیدواژه ها:

Thermus thermophilus ، machine leaning ، meta-analysis ، Gene expression ، SVM-RFE