Results and Discussion Microarray gene expression profiles and validation by qRT PCR We used Agilent whole human genome microarrays to measure relative gene expression in IMR 90 human fibroblast cells exposed to 0. 5 Gy a particles and in inhibitor Volasertib their bystanders at 0. 5, 1, 2, 4, 6 and 24 hours post exposure. The data set was comprised of three treat ment conditions at six time points, with four biologi cal replicates of each condition. The data were background corrected but not normalized in order to preserve dependence across time points. We have previously reported on the analysis of data at the 4 hour time point, and took the 238 genes responding significantly in that study as the focus for the present analysis.
Forty of these genes were selected on the basis of the lowest FDR for differential expression in the microarray analysis, and were analyzed by quantitative real time reverse transcription PCR to validate microarray measurements. The heat maps in Figure 1 depict the qRT PCR data as hier archically clustered logarithmically transformed median gene expression ratios after irradiation and bystander treatment. Figure 1 also shows close concordance between ratios obtained by the microarray and qRT PCR platforms. Overall, we found that qRT PCR meth ods can give higher expression ratios as compared with microarray measurements, as reported previously. We also confirmed previously observed gene expression patterns in irradiated and bystander treated samples. One such pattern was the biphasic response of a large group of inflammatory cytokine genes, including emo kine ligand genes.
The other pattern, a response of cell cycle and DNA damage genes reaching max imum at 4 6 hours after treatment, was more pro nounced in irradiated samples. Among the subset of genes analyzed here it was evident that there was more than one group of coordinately regulated genes, leading to our interest in developing an approach to group temporal profiles of gene expression in order to provide insight into regulatory nodes that may coor dinately control gene expression. Manually curating clusters for comparison validation of clustering methods To evaluate the quality of clustering between methods, we manually curated clusters. Of 80 possible microarray profiles confirmed by qRT PCR, 67 were selected, on the basis of pattern and known pathway information, as distinct and were grouped into AV-951 seven clusters, no early peak, no change, two peaks and two dips, two peaks and two dips with a shallow second dip, two peaks and one dip with a low magnitude first peak, two peaks and one dip with a high magnitude first peak, and down at 4 hours. The graphs in Additional File 1 depict the results of manually curated clustering.