The use of different statistical tools in the identification of perturbation-responsive transcription factors in yeast

Abstract

S. cerevisiae transcription data, obtained under environmental perturbations, such as 16 different macronutrient (carbon, nitrogen, phosphorus and sulfur) limitation regimes in both aerobic and anaerobic conditions, and under genetic perturbations such as the deletions of MIG1 and both MIG1 and MIG2 genes, were integrated to transcriptional regulatory network in order to find the so-called key transcriptional factors (key TFs), meanining to transcriptional factors around which significant changes occur as a perturbation responsive behaviour. Key TFs were identified by integrating the processed transcriptional regulatory network, which consists of 8494 regulatory interactions between 144 transcriptional factors and 3399 target genes, with transcription data. Two probability methods, t-test and EDGE program were used to analyze the transcriptome data. The comparison of the key TFs identified by using two different statistical tools revealed that the application of these two different tools to the same triplicate data set can identify the same set of key TF responsive to genetic perturbations, and to carbon limitation between anaerobic and aerobic conditions. EDGE can therefore replace t-test in the application of the reporter TF algorithm. Dynamic non-replicate S. cerevisiae transcription data, consisting of expression levels obtained at different time points after the glucose pulse was given at the first steady state, resulting in totally 14 time point measurements until the second steady state, was analyzed in order to identify key TFs via cumulative Z-scores calculated using p-values by EDGE. Interacting key TF pairs were identified and their ranking was followed at the time points, and it was observed that interacting key TFs show highly similar changes in ranking order.