Background: Studies have shown that blood platelets contain tumour-specific mRNA profiles tumour-educated platelets (TEPs). Here, we aim to train a TEP-based breast cancer detection classifier. Methods: Platelet mRNA was sequenced from 266 women with stage I–IV breast cancer and 212 female controls from 6 hospitals. A particle swarm optimised support vector machine (PSO-SVM) and an elastic net-based classifier (EN) were trained on 71% of the study population. Classifier performance was evaluated in the remainder (29%) of the population, followed by validation in an independent set (37 cases and 36 controls). Potential confounding was assessed in post hoc analyses. Results: Both classifiers reached an area under the curve (AUC) of 0.85 upon internal validation. Reproducibility in the independent validation set was poor with an AUC of 0.55 and 0.54 for the PSO-SVM and EN classifier, respectively. Post hoc analyses indicated that 19% of the variance in gene expression was associated with hospital. Genes related to platelet activity were differentially expressed between hospitals. Conclusions: We could not validate two TEP-based breast cancer classifiers in an independent validation cohort. The TEP protocol is sensitive to within-protocol variation and revision might be necessary before TEPs can be reconsidered for breast cancer detection.

PURPOSE: Patients with postpartum breast cancer diagnosed after cessation of breastfeeding (postweaning, PP-BCPW) have a particularly poor prognosis compared with patients diagnosed during lactation (PP-BCDL), or to pregnant (Pr-BC) and nulliparous (NP-BC) patients, regardless of standard prognostic characteristics. Animal studies point to a role of the involution process in stimulation of tumor growth in the mammary gland. However, in women, the molecular mechanisms that underlie this poor prognosis of patients with PP-BCPW remain vastly underexplored, due to of lack of adequate patient numbers and outcome data. EXPERIMENTAL DESIGN: We explored whether distinct prognostic features, common to all breast cancer molecular subtypes, exist in postpartum tumor tissue. Using detailed breastfeeding data, we delineated the postweaning period in PP-BC as a surrogate for mammary gland involution and performed whole transcriptome sequencing, immunohistochemical, and (multiplex) immunofluorescent analyses on tumor tissue of patients with PP-BCPW, PP-BCDL, Pr-BC, and NP-BC. RESULTS: We found that patients with PP-BCPW having a low expression level of an immunoglobulin gene signature, but high infiltration of plasma B cells, have an increased risk for metastasis and death. Although PP-BCPW tumor tissue was also characterized by an increase in CD8+ cytotoxic T cells and reduced distance among these cell types, these parameters were not associated with differential clinical outcomes among groups. CONCLUSIONS: These data point to the importance of plasma B cells in the postweaning mammary tumor microenvironment regarding the poor prognosis of PP-BCPW patients. Future prospective and in-depth research needs to further explore the role of B-cell immunobiology in this specific group of young patients with breast cancer.

In this article the data availability statement was incorrectly given. “EGAS0000100682” should have been “EGAS00001006821”. The correct statement as follows. The datasets generated during and/or analysed during the current study are deposited at the European Genome-phenome Archive (EGA) under the accession numbers EGAS00001006821 and EGAD00001009790. The original article has been corrected.

Preclinical models have been the workhorse of cancer research, producing massive amounts of drug response data. Unfortunately, translating response biomarkers derived from these datasets to human tumors has proven to be particularly challenging. To address this challenge, we developed TRANSACT, a computational framework that builds a consensus space to capture biological processes common to preclinical models and human tumors and exploits this space to construct drug response predictors that robustly transfer from preclinical models to human tumors. TRANSACT performs favorably compared to four competing approaches, including two deep learning approaches, on a set of 23 drug prediction challenges on The Cancer Genome Atlas and 226 metastatic tumors from the Hartwig Medical Foundation. We demonstrate that response predictions deliver a robust performance for a number of therapies of high clinical importance: platinum-based chemotherapies, gemcitabine, and paclitaxel. In contrast to other approaches, we demonstrate the interpretability of the TRANSACT predictors by correctly identifying known biomarkers of targeted therapies, and we propose potential mechanisms that mediate the resistance to two chemotherapeutic agents.