In detail, the three methods SCAD-SVM, RF-Boruta, and PAM were used [ [24], [25] and [26]]. Only those target proteins selected by all three classification algorithms in a particular bootstrap data set entered the final biomarker E7080 clinical trial ranking which reflects the selection frequency of certain biomarker proteins. Although bootfs was developed for RPPA derived protein expression data, we anticipate that this approach will also be useful for the other two-group classification tasks. Therefore, we made this method available
as an open source package. Proteins part of our biomarker signature plays a role in diverse biological processes. NDKA, for example, catalyzes the transphosphorylation of γ-phosphates from deoxynucleoside triphosphates to deoxynucleoside diphosphates to supply cells with nucleotides other than ATP [33]. Besides cell proliferation, NDKA is involved in cell differentiation, chromosomal stability, and signal transduction [[34], [35], [36] and [37]]. AZD2281 Although NDKA had initially been described as NM23-H1 by Steeg et al. in 1988 as a gene being downregulated in murine melanoma cell lines with high metastatic potential [38], contradicting results have since then been reported for this gene in other tumor entities. For example, high levels of NDKA expression were linked with aggressive types of prostate cancer and neuroblastoma [[39] and [40]]. Our results suggest that NDKA is a valuable marker also for the identification of
high risk luminal breast cancer. In detail, NDKA was found highly expressed in histologic G3 tumors as identified by RPPA and confirmed by Western blot. In addition, protein and
mRNA expression of NDKA was highly Unoprostone correlated. Using a large, publically available gene expression dataset [2], positive correlation between high NDKA expression levels and the group of luminal B tumors was confirmed. Along with several other ribosomal proteins, RPS6 is part of the ribosomal 40S subunit controlling protein synthesis rate and cell size during cell division and differentiation [41]. RPPA-based tumor profiling identified RPS6 as being highly expressed in histologic G3 tumor samples. However, RPS6 protein expression was not correlated with transcript levels for RPS6 in line with a previous report [16] indicating a regulation of RPS6 at the posttranscriptional level. In contrast to Ki-67, NDKA, and RPS6, caveolin-1 was strongly expressed in histologic G1 tumor samples and a positive correlation between protein and mRNA levels was observed. The differential expression of caveolin-1 in luminal A and luminal B tumors was also seen in the Curtis data set [2]. Caveolin-1 is the main component of caveolae, a subset of lipid rafts which, for example, serve as molecular hubs modulating the activity of signaling pathways. In the context of breast cancer, loss of caveolin-1 in cancer-associated fibroblasts results in an activated tumor microenvironment and has been linked to poor clinical outcome [[42], [43] and [44]].