Plasma-induced transcription and WGCNA. (A) Identification of gene input list. Affymetrix U133 plus 2.0 gene chip has probe sets for interrogation of >54 000 transcripts. We filtered differentially expressed transcripts to create the gene input list for WGCNA. This included using an intersection of 2 filtering methodologies. For the first method, we retained genes that were in the top 25% of transcripts meeting stated thresholds in both single-site and multisite analyses. We retained probe sets where at least 25% of samples possessed log2 intensity ≥4 RFU. These data are represented by the upper portion of the Venn diagram. For the second method, we retained genes that were in the top 5000 transcripts with the highest MAD in single-site and multisite analyses. These data are represented by the lower portion of the Venn diagram. We leveraged 2 independent data sets (single site, multisite) to identify the maximum number of the most variant transcripts that were present in the combined data sets. This approach increases the likelihood that the observed variance is based in biological as opposed to technical variance. The intersection of these 2 analyses was used as the WGCNA input list (n = 3028 probe sets). (B) Determination of soft-thresholding power. Power of 6 was chosen based on signed scale independence and mean connectivity. (C) Dendrogram representing coexpression modules. Vertical lines represent genes. Lower values indicate greater similarity of transcript expression profiles across samples. Coexpression modules are indicated by color. (D) Correlations between gene modules and number of acute care visits for pain in the prior 3 years in 6 different centers using WGCNA. Columns represent different centers, number of participants from each center is displayed. Coexpression modules are represented by rows and labeled by color. The number of probe sets assigned to each module are indicated in the far-left column labeled gene (n). The P value shown within each box is between the eigengene (defined as the first principal component of a principal component analysis on the gene expression of all genes from a module) and number of acute care visits for pain in the previous 3 years for each center; all red shaded boxes represent correlation coefficient ≥|0.3| and P <.05. FDR-adjusted meta–P value is also displayed that was calculated for those modules that had significant correlations with clinical data from ≥50% of the centers (ie, ≥3 centers). N/A, not applicable; Q3, third quartile; RFU, relative fluorescence unit.
