Correlation-based approach for identification of endocrine interactions

The Ssec ranking score and protein-specific pathway enrichment commands in R utilize functions from the package WGCNA

In addition to tissue-specific expression arrays, this pipeline also imports information from Uniprot to filter for secreted proteins. This list has been uploaded for ease of pipeline execution, however is subjected to updates. Revised lists can be retrieved from the following sources:

UniProt <www.uniprot.org/> deposited annotations for Organism:mouse "Mus musculus (Mouse) [10090]" and subcellular localization:secreted :"Secreted [SL-0243]"

Citation: UniProt: the universal protein knowledgebase Nucleic Acids Res. 45: D158-D169 (2017)
These are provided as: Secreted_proteins_Uniprot

Filtering for tissue-specific expression was performed by manual inspection using BioGPS on the following mouse arrays http://biogps.org/dataset/BDS_00009/: GeneAtlas GNF1M, gcrma
While this can be easily automated, we felt more confident with inspection due to the small number of samples per tissue and consequent high level of variation among expression

GEO code for arrays: GSE1133

Citation: McClurg P, Janes J, Wu C, Delano DL, Walker JR, Batalov S, Takahashi JS, Shimomura K, Kohsaka A, Bass J, Wiltshire T, Su AI (2007) Genomewide association analysis in diverse inbred mice: power and population structure. Genetics 176(1):675-83.

Pretreatment of data is described below: Most mouse expression arrays were performed on a Affymetrix HT_MG-430A, unless otherwise indicated in the GEO accession

GEO Accession for adipose and liver arrays: GSE64770

Raw data deposited in GEO is RMA-normalized microarray data. From these, each numeric value represents a probe detected in each HMDP animal. Given that some of probes correspond to the same gene, we chose to start with a single expression value for each unique gene (averaged amongst probes). Also, these studies contain multiple mice per strain. Given that for each study, mice were fed the same diet and age-matched, we aggregated expression values to one average value per mouse strain. These averages are discussed in greater detail below:

Our pipeline begins with gene expression arrays for liver and adipose tissue, where each gene is represented as an averaged value across probes and strains used in the study. These aggregate matrices are also provided in this repository. The arrays consisted of ~22,400 probes which were aggregated to averages for each gene (12,242). The expression values for each mouse were also averaged to reflect a single value per gene per strain (106). Therefore, each liver and adipose tissue expression matrix consists of 12,242 genes among 106 unique HMDP strains.

To account for different arrays containing differing number of genes, a portion of the Ssec pipeline script divides the Ssec score to the total length of the target tissue genes detected. This step is meant to prevent inflation of these scores by a larger number of genes detected in each target tissue, thus providing an adequate comparison for values between cross-tissue circuits.

1. Construct cross-tissue correlation and pvalue matrices
2. Compute rowsum -ln(pvalue) for ranking interactions and filter for factors proteins as secreted
3. Condition correlation matrix on a by-gene basis for pathway enrichment

The script used to generate QQ-plots from the Ssec scores (step 2 above) was designed and provided by Simon Koplev and listed as: QQ-Plot_R-script.R

Any questions/comments, please contact [email protected]

Additional GEO Accessions for datasets used in the study are listed below:

chow aorta: GSE38120
chow heart: GSE77263
chow liver: GSE16780
chow adipose: GSE42890
high fat/high sucrose hypothalamus: GSE79551