br Unfortunately we were unable
Unfortunately, we were unable to compare the performance of mPS with that of other representative tools including MammaPrint and Oncotype Dx 21–gene RS, given that these are commercial products and the formulas for their calculation have not been disclosed. However, we demonstrated the superiority of mPS relative to PAM50 classifica-tion (Fig. 3, a and b), which is routinely applied in hospitals, as well as to two recently proposed prognostic indicators [19,22] with their own data sets (Fig. 3, a and c–e). Our method is likely to outperform previous scores because mPS stratifies patients at the same clinical stage (Fig. 4, d–f) as well as those with M3814 (nedisertib) receptor–negative subtypes of breast cancer (Fig. 4a), unlike existing methods.
There are numerous protocols for preservation of tumour samples, RNA extraction, and analysis of expression status, which hindered us
from establishing one universal cutoff for each of the 23 genes in the present study. We aimed to build a “platform-independent” score that can be calculated from data obtained by any method once the necessary protocols and distribution patterns obtained with these protocols are established. Comparison of these protocols and development of a robust and precise method to examine the expression levels of the 23 genes, followed by the performance of pilot studies to test the distribution pat-terns, are remaining challenges that must be addressed before mPS can be applied in the clinical setting.
Other limitations of our study include the fact that all analyses were performed in a retrospective manner. Although the total number of pa-tients analyzed (n = 11,893), including the ongoing cohort GSE96058 , is among the largest of those previously examined, prospective studies will be needed to validate our findings.
The best-characterized gene among the 23 prognosis-related genes identified in the present study is FOXM1. A PubMed search for “FOXM1 breast cancer” identified ~180 papers. The FOXM1 protein functions as a transcriptional activator. It is phosphorylated in M phase of the cell cycle and up-regulates the expression of several proliferation-related genes including those for cyclin B1 and Skp2, the latter of which plays an essential role in cell cycle progression by medi-ating the ubiquitin-dependent degradation of the cyclin-dependent ki-nase inhibitors p21, p27, and p57 . The prognostic value of FOXM1 for solid tumours as identified by meta-analysis is also documented in a recent review . In contrast, most of the 23 prognosis-related genes (GARS, UTP23, HMGB3, ATP5F1B, CYB561, EZR, CIRBP, PTGER3, LAMA3, OARD1, ANKRD29, MITD1, and LAMB3) have not been studied in relation to breast cancer, given that PubMed searches for “GENE
Please cite this article as: H. Shimizu and K.I. Nakayama, A 23 gene–based molecular prognostic score precisely predicts overall survival of breast cancer pati..., EBioMedicine, https://doi.org/10.1016/j.ebiom.2019.07.046
breast cancer” identified fewer than 10 publications for each gene, with there being no published papers at all for five of these genes (UTP23, CYB561, OARD1, ANKRD29I, and MITD1). Both basic and clinical studies will be necessary for further elucidation of the fundamental mecha-nisms responsible for the effects of the 23 genes on which mPS is based and for the development of novel drugs to prolong OS of breast cancer patients.
We expect that application of mPS will not only facilitate selection of therapeutic strategies on the basis of the precise prediction of personal prognosis, but also contribute to further understanding of the basic biol-ogy of breast cancer and thereby inform the development of new ther-apeutic approaches.
Supplementary data to this article can be found online at https://doi.
This work was supported by a KAKENHI grant from the Ministry of Education, Culture, Sports, Science, and Technology of Japan to H.S. (19K20403) and K.I.N (18H05215). The funding agency played no role in this study.