Construction of prostate cancer cell line (PC-3)-specific interactome

The purpose of our study was to construct prostate cancer cell-specific interactome by aggregating microarray samples available in public, and identify novel transcriptional regulators that control prostate cancer cell cycle progression. We collected 121 samples of PC-3 cell microarray data from 11 prior studies and our own study data, GSE45567, and performed pre-processing steps, such as quality control test, normalization, and batch effect adjusting. In quality control test, 10 low-quality samples were removed. The matrix data we deposited in GEO has normalized log2 signal intensity of 11,877 genes common in 110 samples, as the 110 samples came from different Affymetrix. The Gene IDs were identified by Entrez Gene IDs that overlapped across different Affymetrix platforms (File: GSE67157_Matrix_PC-3cell_Interactome.txt). In GSE45567 dataset, which is one of the 12 data sets, we selected microarray samples of vehicle-, cineol-, linalool-, and geraniol-treated PC-3 cells, and performed unsupervised clustering analysis to establish dependable clusters of a clear phenotypic transition. Then, the gene set specifically enriched or depleted in the geraniol cluster against vehicle cluster was used for gene set enrichment analysis (GSEA). We found cell cycle-related gene signatures were specifically down-regulated in the geraniol cluster. PC-3 cell-specific interactome was assembled from 110 microarray samples by Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNe, PMID: 16723010). We used Master Regulator Analysis-Fisher’s Extact Test (MRA-FET, PMID: 20531406) to infer transcription factors which are geraniol target molecules and control cell cycle gene signatures. We discovered E2F8 is as a master regulator to modulate geraniol-specific genes toward G2/M cell cycle arrest. Our findings will provide biological insights into the role of E2F8 in prostate cancer for transcriptional regulation leading to the development of cell cycle-targeting chemotherapeutic reagent. In addition, our approach serves as an example to decide the most potent anti-cancer regime in other cancer types and identify the target molecules of cancer-preventive reagents through computational analyses. A total number of 12 datasets containing 121 microarray samples were downloaded from NCBI GEO (http://ncbi.nlm.nih.gov/geo) or ArrayExpress (http://www.ebi.ac.uk/arrayexpress). The individual data sets had been used to address various types of research questions, such as the effects of various chemotherapeutic reagents. We focused on microarray data that had been obtained using the Human Genome U133A, U133 Plus 2.0, and 1.0 ST Affymetrix platforms because they constitute the majority of available microarray samples for the PC-3cell type and data can be combined across these platforms with ease. Pre-processing We applied quality-control test to each sample to remove low-quality samples, normalized only high-quality samples by Single-Channel Array Normalization (SCAN) algorithm (PMID: 22959562), and adjusted for intra- and inter-study batch effects across data sets. Unsupervised clustering analysis and validation We used 12 of monoterpene-treated microarray samples in GSE45567 and filtered out 5,000 genes of the high variance across samples. Then, we performed unsupervised clustering analysis to confirm transcriptional transition based on gene expression profile. Principal Component Analysis (PCA) and supervised classification were carried out to internally validate the clustering result and to evaluate cluster compactness. Significant gene set SAM analysis (PMID: 11309499) was used to identify differentially expressed genes out of the 5,000 genes between the vehicle and the geraniol clusters. We found out the get set specifically enriched in the geraniol cluster and applied this gene set for MRA-FET to infer geraniol-targeted transcription factors. PC-3 cell-specific interactome and Master Regulator Analysis (MRA) From the 110 microarray samples that have passed quality control test, normalization, and batch adjusting, we selected the same 5,000 genes of high variance which were used in the clustering analysis. Then, ARACNe built up regulatory interactions between the 5,000 genes and 472 human transcription factors (TFs) as hub markers through mutual information (MI) calculation. From the PC-3 cell-specific interactome, MRA-FET inferred master regulator (MR) candidates which control the geraniol-specific cell cycle signatures. The information of 121 microarray data samples are summarized in the meta-data spreadsheet. The meta-data have 121 samples with GSE accession number, array platform, original experiment type, quality control test and clustering analysis result. The matrix data txt file (Matrix_PC-3cell_interactome.txt) has normalized log2 expression values of 110 samples after excluding 11 low-quality samples. The 11 low-quality samples are indicated in the meta-data spreadsheet.