Genome-wide BigData analytics: Case of yeast stress signature detection

It has been generally recognized that BigData analytics presently have most significant impact on computer inference in life sciences, such as genome wide association studies (GWAS) in basic research and personalized medicine, and its importance will further increase in near future. In this work non-parametric separation of responsive yeast genes from experimental data obtained in chemostat cultivation under dilution rate and nutrient limitations with basic biogenic elements (C,N,S,P), and the specific leucine and uracil auxothropic limitations. Elastic net models are applied for the detection of the key responsive genes for each of the specific limitations. Bootstrap and perturbation methods are used to determine the most important responsive genes and corresponding quantiles applied to the complete data set for all of the nutritional and growth rate limitations. The model predicts that response of gene YOR 348 C, involved in proline metabolism, as the key signature of stress. Based on literature data, the obtained result are confirmed experimentally by the biochemistry of plants under physical and chemical stress, also by functional genomics of bakers yeast, and also its important function in human tumorogenesis is observed.