Global gene expression analysis of breast cancer cell line responses to Topsentinol L Trisulfate or DMSO as the vehicle control

We generated a signature for topsentinol L trisulfate in breast cancer cells to predict drug sensitivity in other available datasets. To process the mRNA sequencing data, we used the TCGA mRNA-seq Pipeline. RNA sequencing reads for the treated and control samples were aligned using MapSplice v12_07, quantified using RSEM, and gene counts were normalized using upper quantile normalization. This was the same methodology used to normalize the PANCAN12 TCGA dataset, which we obtained from TCGA fully processed for use in this analysis. To generate a TLT sensitivity signature, we used the DESeq2 package (version 1.4.5) in the Bioconductor framework (version 2.14.0, version 3.1.0 of R) to identify genes that were significantly deregulated (adjusted p < 0.05) between the treated and control samples. One hundred forty-six genes were found to be significantly deregulated, out of which only 131 were found in the TCGA dataset. To use DEseq2, the reads had to be re-mapped using the Rsubread Bioconductor package. We used this package to map the reads to version hg19 of the human genome and to summarize the data to gene-level values. We predicted TLT sensitivity for the PANCAN12 TCGA dataset using the Bayesian binary regression algorithm version 2.0 (BinReg2.0) used as a MATLAB plug-in. We used default parameters, except that our signature used 131 genes and one metagene. The probability output from the binary regression model was subtracted from one so that probabilities closer to one indicated a higher probability of sensitivity to the drug as previously described [29]. Prior to making the predictions, the data were log2 transformed and DWD normalized to reduce biases that can result from differences in batch processing and platforms. Eight TLT-sensitive BL-CL cell lines were treated separately with TLT (at EC75) or DMSO for 6 hours, after which total RNA was extracted using the RNeasy Mini Kit