Figure 1.
Critical responses to human inflammation.
Gene expression patterns selected from the LPS dataset, including early up – 182 probesets (red), middle up – 119 probesets (green), late up – 284 probesets (blue), and down – 1,118 probesets (magenta); totally 1,730 selected probesets over 3,269. Top-left is the average expression profiles of these patterns; bottom-left is the corresponding heat-map; and the rest are expression profiles of selected genes in four patterns (the horizontal axis is six time-points (0, 2, 4, 6, 9, 24 hours) and the vertical axis is the intensity of mRNA levels).
Table 1.
Data information and inflammation-relevant significant functions.
Figure 2.
Statistical significance thresholds of CRMs.
A procedure randomly picks a gene-set with N genes from the background and search for common CRMs (δ = 0.7) in that gene-set. The statistical significant p-value for each CRM is estimated and the minimum one is reported. Each point in the blue curve is a transformed value of the mean of the minimum p-values of CRMs in 100 times running the procedure for the corresponding k. Approximately, the red curve shows which thresholds should be used for the non-random cases. After N = 14 genes, only one threshold is used to ensure the significance (p-value = 0.01).
Table 2.
Critical transcription factors in human endotoxemia model.
Figure 3.
Putative temporal regulatory program in human endotoxemia plus schematic illustration of the integrated computational framework.
The clustering and selection step extracts a ‘clusterable’ subset of differentially expressed probesets and cluster it into a number of expression patterns. Subsequently, pathway enrichment is performed in each pattern and relevant significant pathways are selected based on literature information. The process of CRM searching is then applied to each gene battery which is a group of genes that belong to an expression pattern and a particular pathway. Eventually, 34 TFs are identified as human inflammation-relevant transcriptional regulators. The results show a highly dynamic perspective of regulation and interactions between genes, functions, and TF across the time.
Figure 4.
Pleiotropic effects of transcription factors across biological processes.
Venn diagram shows pair-wise transcription factor combinations that overlap between the inflammatory relevant pathways (*: not present as TFs that regulate Toll-like receptor signaling pathway in this case).
Figure 5.
Dynamic representation of transcriptional regulatory network for apoptosis signaling.
Transcription factors and target genes are shown as nodes and their putative regulatory interactions are drawn as edges.
Table 3.
Statistical significance of selected cis-regulatory modules*.
Table 4.
Critical transcription factors identified from the in vitro endotoxin study.
Figure 6.
Flowchart of the CRM searching process.
Each binding site is characterized by the TF name, position and binding strand orientation (+: forward and −: backward). Promoter sequences are converted into promoter profiles to speed up the calculation. A gene profile contains a set of promoter profiles that are corresponding to a set of alternative promoters of that gene. The background set contains 5,000 randomly selected human genes.