Minimising immunohistochemical false negative ER classification using a complementary 23 gene expression signature of ER status

Identification of a gene expression signature associated with breast cancer survival and risk that improves clinical genomic platforms

Motivation

Modern genomic technologies allow us to perform genome-wide analysis to find gene markers associated with the risk and survival in cancer patients. Accurate risk prediction and patient stratification based on robust gene signatures is a key path forward in personalized treatment and precision medicine. Several authors have proposed the identification of gene signatures to assign risk in patients with breast cancer (BRCA), and some of these signatures have been implemented within commercial platforms in the clinic, such as Oncotype and Prosigna. However, these platforms are black boxes in which the influence of selected genes as survival markers is unclear and where the risk scores provided cannot be clearly related to the standard clinicopathological tumor markers obtained by immunohistochemistry (IHC), which guide clinical and therapeutic decisions in breast cancer.

Results

Here, we present a framework to discover a robust list of gene expression markers associated with survival that can be biologically interpreted in terms of the three main biomolecular factors (IHC clinical markers: ER, PR and HER2) that define clinical outcome in BRCA. To test and ensure the reproducibility of the results, we compiled and analyzed two independent datasets with a large number of tumor samples (1024 and 879) that include full genome-wide expression profiles and survival data. Using these two cohorts, we obtained a robust subset of gene survival markers that correlate well with the major IHC clinical markers used in breast cancer. The geneset of survival markers that we identify (which includes 34 genes) significantly improves the risk prediction provided by the genesets included in the commercial platforms: Oncotype (16 genes) and Prosigna (50 genes, i.e. PAM50). Furthermore, some of the genes identified have recently been proposed in the literature as new prognostic markers and may deserve more attention in current clinical trials to improve breast cancer risk prediction.

Availability and implementation

All data integrated and analyzed in this research will be available on GitHub (https://github.com/jdelasrivas-lab/breastcancersurvsign), including the R scripts and protocols used for the analyses.

Supplementary information

Supplementary data are available at Bioinformatics Advances online.

Decision Theory versus Conventional Statistics for Personalized Therapy of Breast Cancer

Estrogen and progesterone receptors being present or not represents one of the most important biomarkers for therapy selection in breast cancer patients. Conventional measurement by immunohistochemistry (IHC) involves errors, and numerous attempts have been made to increase precision by additional information from gene expression. This raises the question of how to fuse information, in particular, if there is disagreement. It is the primary domain of Dempster–Shafer decision theory (DST) to deal with contradicting evidence on the same item (here: receptor status), obtained through different techniques. DST is widely used in technical settings, such as self-driving cars and aviation, and is also promising to deliver significant advantages in medicine. Using data from breast cancer patients already presented in previous work, we focus on comparing DST with classical statistics in this work, to pave the way for its application in medicine. First, we explain how DST not only considers probabilities (a single number per sample), but also incorporates uncertainty in a concept of ‘evidence’ (two numbers per sample). This allows for very powerful displays of patient data in so-called ternary plots, a novel and crucial advantage for medical interpretation. Results are obtained according to conventional statistics (ODDS) and, in parallel, according to DST. Agreement and differences are evaluated, and the particular merits of DST discussed. The presented application demonstrates how decision theory introduces new levels of confidence in diagnoses derived from medical data.

An Efficient Algorithm for the Detection of Outliers in Mislabeled Omics Data

High dimensionality and noise have made it difficult to detect related biomarkers in omics data. Through previous study, penalized maximum trimmed likelihood estimation is effective in identifying mislabeled samples in high-dimensional data with mislabeled error. However, the algorithm commonly used in these studies is the concentration step (C-step), and the C-step algorithm that is applied to robust penalized regression does not ensure that the criterion function is gradually optimized iteratively, because the regularized parameters change during the iteration. This makes the C-step algorithm runs very slowly, especially when dealing with high-dimensional omics data. The AR-Cstep (C-step combined with an acceptance-rejection scheme) algorithm is proposed. In simulation experiments, the AR-Cstep algorithm converged faster (the average computation time was only 2% of that of the C-step algorithm) and was more accurate in terms of variable selection and outlier identification than the C-step algorithm. The two algorithms were further compared on triple negative breast cancer (TNBC) RNA-seq data. AR-Cstep can solve the problem of the C-step not converging and ensures that the iterative process is in the direction that improves criterion function. As an improvement of the C-step algorithm, the AR-Cstep algorithm can be extended to other robust models with regularized parameters.

Decision theory for precision therapy of breast cancer

Correctly estimating the hormone receptor status for estrogen (ER) and progesterone (PGR) is crucial for precision therapy of breast cancer. It is known that conventional diagnostics (immunohistochemistry, IHC) yields a significant rate of wrongly diagnosed receptor status. Here we demonstrate how Dempster Shafer decision Theory (DST) enhances diagnostic precision by adding information from gene expression. We downloaded data of 3753 breast cancer patients from Gene Expression Omnibus. Information from IHC and gene expression was fused according to DST, and the clinical criterion for receptor positivity was re-modelled along DST. Receptor status predicted according to DST was compared with conventional assessment via IHC and gene-expression, and deviations were flagged as questionable. The survival of questionable cases turned out significantly worse (Kaplan Meier p < 1%) than for patients with receptor status confirmed by DST, indicating a substantial enhancement of diagnostic precision via DST. This study is not only relevant for precision medicine but also paves the way for introducing decision theory into OMICS data science.

Estrogen receptor variants in ER-positive basal-type breast cancers responding to therapy like ER-negative breast cancers

Immunohistochemically ER-positive HER2-negative (ER+HER2−) breast cancers are classified clinically as Luminal-type. We showed previously that molecular subtyping using the 80-gene signature (80-GS) reclassified a subset of ER+HER2− tumors to molecular Basal-type. We report here that molecular reclassification is associated with expression of dominant-negative ER variants and evaluate response to neoadjuvant therapy and outcome in the prospective neoadjuvant NBRST study (NCT01479101). The 80-GS reclassified 91 of 694 (13.1%) immunohistochemically Luminal-type tumors to molecular Basal-type. Importantly, all 91 discordant tumors were classified as high-risk, whereas only 66.9% of ER+/Luminal-type tumors were classified at high-risk for disease recurrence (i.e., Luminal B) (P < 0.001). ER variant mRNA (ER∆3, ER∆7, and ERα-36) analysis performed on 84 ER+/Basal tumors and 48 ER+/Luminal B control tumors revealed that total ER mRNA was significantly lower in ER+/Basal tumors. The relative expression of ER∆7/total ER was significantly higher in ER+/Basal tumors compared to ER+/Luminal B tumors (P < 0.001). ER+/Basal patients had similar pathological complete response (pCR) rates following neoadjuvant chemotherapy as ER−/Basal patients (34.3 vs. 37.6%), and much higher than ER+/Luminal A or B patients (2.3 and 5.8%, respectively). Furthermore, 3-year distant metastasis-free interval (DMFI) for ER+/Basal patients was 65.8%, significantly lower than 96.3 and 88.9% for ER+/Luminal A and B patients, respectively, (log-rank P < 0.001). Significantly lower total ER mRNA and increased relative ER∆7 dominant-negative variant expression provides a rationale why ER+/Basal breast cancers are molecularly ER-negative. Identification of this substantial subset of patients is clinically relevant because of the higher pCR rate to neoadjuvant chemotherapy and correlation with clinical outcome.

pAKT pathway activation is associated with PIK3CA mutations and good prognosis in luminal breast cancer in contrast to p-mTOR pathway activation

Numerous studies have focused on the PI3K/AKT/mTOR pathway in estrogen receptor positive (ER) breast cancer (BC), as a linear signal transduction pathway and reported its association with worse clinical outcomes. We developed gene signatures that reflect the level of expression of phosphorylated-Serine473-AKT (pAKT) and phosphorylated-Serine2448-mTOR (p-mTOR) separately, capturing their corresponding level of pathway activation. Our analysis revealed that the pAKT pathway activation was associated with luminal A BC while the p-mTOR pathway activation was more associated with luminal B BC (Kruskal-Wallis test p < 10^-10). pAKT pathway activation was significantly associated with better outcomes (multivariable HR, 0.79; 95%CI, 0.74-0.85; p = 2.5 × 10^-10) and PIK3CA mutations (p = 0.0001) whereas p-mTOR pathway activation showed worse outcomes (multivariable HR,1.1; 95%CI, 1.1-1.2; p = 9.9 × 10^-4) and associated with p53 mutations (p = 0.04). in conclusion, our data show that pAKT and p-mTOR pathway activation have differing impact on prognosis and suggest that they are not linearly connected in luminal breast cancers.

Co-expressed genes enhance precision of receptor status identification in breast cancer patients

PURPOSE

Therapeutic decisions in breast cancer patients crucially depend on the status of estrogen receptor, progesterone receptor and HER2, obtained by immunohistochemistry (IHC). These are known to be inaccurate sometimes, and we demonstrate how to use gene-expression to increase precision of receptor status.

METHODS

We downloaded data from 3241 breast cancer patients out of 36 clinical studies. For each receptor, we modelled the mRNA expression of the receptor gene and a co-gene by logistic regression. For each patient, predictions from logistic regression were merged with information from IHC on a probabilistic basis to arrive at a fused prediction result.

RESULTS

We introduce Sankey diagrams to visualize the step by step increase of precision as information is added from gene expression: IHC-estimates are qualified as 'confirmed', 'rejected' or 'corrected'. Additionally, we introduce the category 'inconclusive' to spot those patients in need for additional assessments so as to increase diagnostic precision and safety.

CONCLUSIONS

We demonstrate a sound mathematical basis for the fusion of information, even if partly contradictive. The concept is extendable to more than three sources of information, as particularly important for OMICS data. The overall number of undecidable cases is reduced as well as those assessed falsely. We outline how decision rules may be extended to also weigh consequences, being different in severity for false-positive and false-negative assessments, respectively. The possible benefit is demonstrated by comparing the disease free survival between patients whose IHC could be confirmed versus those for which it was corrected.

Receptor-based Surrogate Subtypes and Discrepancies with Breast Cancer Intrinsic Subtypes: Implications for Image Biomarker Development

Purpose To determine the concordance and accuracy of imaging surrogates of immunohistochemical (IHC) markers and the molecular classification of breast cancer. Materials and Methods A total of 3050 patients from 17 public breast cancer data sets containing IHC marker receptor status (estrogen receptor/progesterone receptor/human epidermal growth factor receptor 2 [HER2]) and their molecular classification (basal-like, HER2-enriched, luminal A or B) were analyzed. Diagnostic accuracy and concordance as measured with the κ statistic were calculated between the IHC and molecular classifications. Simulations were performed to assess the relationship between accuracy of imaging-based IHC markers to predict molecular classification. A simulation was performed to examine effects of misclassification of molecular type on patient survival. Results Accuracies of intrinsic subtypes based on IHC subtype were 71.7% (luminal A), 53.7% (luminal B), 64.8% (HER2-enriched), and 81.7% (basal-like). The κ agreement was fair (κ = 0.36) for luminal A and HER2-enriched subtypes, good (κ = 0.65) for the basal-like subtype, and poor (κ = 0.09) for the luminal B subtypes. Introduction of image misclassification by simulation lowered image-true subtype accuracies and κ values. Simulation analysis showed that misclassification caused survival differences between luminal A and basal-like subtypes to decrease. Conclusion There is poor concordance between triple-receptor status and intrinsic molecular subtype in breast cancer, arguing against their use in the design of prognostic genomic-based image biomarkers. © RSNA, 2018 Online supplemental material is available for this article.

A novel non-parametric method for uncertainty evaluation of correlation-based molecular signatures: its application on PAM50 algorithm

Motivation

The PAM50 classifier is used to assign patients to the highest correlated breast cancer subtype irrespectively of the obtained value. Nonetheless, all subtype correlations are required to build the risk of recurrence (ROR) score, currently used in therapeutic decisions. Present subtype uncertainty estimations are not accurate, seldom considered or require a population-based approach for this context.

Results

Here we present a novel single-subject non-parametric uncertainty estimation based on PAM50's gene label permutations. Simulations results ( n  = 5228) showed that only 61% subjects can be reliably 'Assigned' to the PAM50 subtype, whereas 33% should be 'Not Assigned' (NA), leaving the rest to tight 'Ambiguous' correlations between subtypes. The NA subjects exclusion from the analysis improved survival subtype curves discrimination yielding a higher proportion of low and high ROR values. Conversely, all NA subjects showed similar survival behaviour regardless of the original PAM50 assignment. We propose to incorporate our PAM50 uncertainty estimation to support therapeutic decisions.

Availability and Implementation

Source code can be found in 'pbcmc' R package at Bioconductor.

Contacts

cristobalfresno@gmail.com or efernandez@bdmg.com.ar.

Supplementary information

Supplementary data are available at Bioinformatics online.

Estrogen receptor α (ERα) status evaluation using RNAscope in situ hybridization: a reliable and complementary method for IHC in breast cancer tissues

Estrogen receptor α (ERα) plays a significant role in the development of breast cancer and has been used clinically as an endocrine therapeutic target. Currently, clinical laboratories use immunohistochemistry (IHC) to determine the ERα status of patients in order to distinguish those who would benefit from endocrine therapy. This method is highly subjective, requires a large amount of tumor tissue, and may generate false-negative results. To improve the detection of ERα, we used a new RNA in situ hybridization technique (RNAscope) and compared its use with IHC in 72 breast cancer tissues (47 ERα positive and 25 ERα negative). Then we evaluated ERα mRNA by RT-qPCR with RNAscope. An unobvious difference was found between reverse-transcription quantitative polymerase chain reaction (RT-qPCR) and IHC, but a positive correlation was found between RNAscope and IHC. In addition, breast cancer is a highly heterogeneous cancer, and RNAscope could easily reveal the heterogeneity in breast cancer. Moreover, we found that some ERα IHC-based negative and RNAscope-based positive test results were detected as positive after testing with IHC again. Our findings suggest that RNAscope may be a complementary method for improving the detection of patient ERα status and has potential clinical utility.

Multimodal Assessment of Estrogen Receptor mRNA Profiles to Quantify Estrogen Pathway Activity in Breast Tumors

BACKGROUND

Molecular markers have transformed our understanding of the heterogeneity of breast cancer and have allowed the identification of genomic profiles of estrogen receptor (ER)-α signaling. However, our understanding of the transcriptional profiles of ER signaling remains inadequate. Therefore, we sought to identify the genomic indicators of ER pathway activity that could supplement traditional immunohistochemical (IHC) assessments of ER status to better understand ER signaling in the breast tumors of individual patients.

MATERIALS AND METHODS

We reduced ESR1 (gene encoding the ER-α protein) mRNA levels using small interfering RNA in ER⁺ MCF7 breast cancer cells and assayed for transcriptional changes using Affymetrix HG U133 Plus 2.0 arrays. We also compared 1034 ER⁺ and ER^- breast tumors from publicly available microarray data. The principal components of ER activity generated from these analyses and from other published estrogen signatures were compared with ESR1 expression, ER-α IHC, and patient survival.

RESULTS

Genes differentially expressed in both analyses were associated with ER-α IHC and ESR1 mRNA expression. They were also significantly enriched for estrogen-driven molecular pathways associated with ESR1, cyclin D1 (CCND1), MYC (v-myc avian myelocytomatosis viral oncogene homolog), and NFKB (nuclear factor kappa B). Despite their differing constituent genes, the principal components generated from these new analyses and from previously published ER-associated gene lists were all associated with each other and with the survival of patients with breast cancer treated with endocrine therapies.

CONCLUSION

A biomarker of ER-α pathway activity, generated using ESR1-responsive mRNAs in MCF7 cells, when used alongside ER-α IHC and ESR1 mRNA expression, could provide a method for further stratification of patients and add insight into ER pathway activity in these patients.

Elafin drives poor outcome in high-grade serous ovarian cancers and basal-like breast tumors

High-grade serous ovarian carcinoma (HGSOC) and basal-like breast cancer (BLBC) share many features including TP53 mutations, genomic instability and poor prognosis. We recently reported that Elafin is overexpressed by HGSOC and is associated with poor overall survival. Here, we confirm that Elafin overexpression is associated with shorter survival in 1000 HGSOC patients. Elafin confers a proliferative advantage to tumor cells through the activation of the MAP kinase pathway. This mitogenic effect can be neutralized by RNA interference, specific antibodies and a MEK inhibitor. Elafin expression in patient-derived samples was also associated with chemoresistance and strongly correlates with bcl-xL expression. We extended these findings into the examination of 1100 primary breast tumors and six breast cancer cell lines. We observed that Elafin is overexpressed and secreted specifically by BLBC tumors and cell lines, leading to a similar mitogenic effect through activation of the MAP kinase pathway. Here too, Elafin overexpression is associated with poor overall survival, suggesting that it may serve as a biomarker and therapeutic target in this setting.

Transfer of clinically relevant gene expression signatures in breast cancer: from Affymetrix microarray to Illumina RNA-Sequencing technology

Background

Microarrays have revolutionized breast cancer (BC) research by enabling studies of gene expression on a transcriptome-wide scale. Recently, RNA-Sequencing (RNA-Seq) has emerged as an alternative for precise readouts of the transcriptome. To date, no study has compared the ability of the two technologies to quantify clinically relevant individual genes and microarray-derived gene expression signatures (GES) in a set of BC samples encompassing the known molecular BC’s subtypes. To accomplish this, the RNA from 57 BCs representing the four main molecular subtypes (triple negative, HER2 positive, luminal A, luminal B), was profiled with Affymetrix HG-U133 Plus 2.0 chips and sequenced using the Illumina HiSeq 2000 platform. The correlations of three clinically relevant BC genes, six molecular subtype classifiers, and a selection of 21 GES were evaluated.

Results

16,097 genes common to the two platforms were retained for downstream analysis. Gene-wise comparison of microarray and RNA-Seq data revealed that 52% had a Spearman’s correlation coefficient greater than 0.7 with highly correlated genes displaying significantly higher expression levels. We found excellent correlation between microarray and RNA-Seq for the estrogen receptor (ER; r_s = 0.973; 95% CI: 0.971-0.975), progesterone receptor (PgR; r_s = 0.95; 0.947-0.954), and human epidermal growth factor receptor 2 (HER2; r_s = 0.918; 0.912-0.923), while a few discordances between ER and PgR quantified by immunohistochemistry and RNA-Seq/microarray were observed. All the subtype classifiers evaluated agreed well (Cohen’s kappa coefficients >0.8) and all the proliferation-based GES showed excellent Spearman correlations between microarray and RNA-Seq (all r_s >0.965). Immune-, stroma- and pathway-based GES showed a lower correlation relative to prognostic signatures (all r_s >0.6).

Conclusions

To our knowledge, this is the first study to report a systematic comparison of RNA-Seq to microarray for the evaluation of single genes and GES clinically relevant to BC. According to our results, the vast majority of single gene biomarkers and well-established GES can be reliably evaluated using the RNA-Seq technology.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-1008) contains supplementary material, which is available to authorized users.

Microarray-based RNA profiling of breast cancer: batch effect removal improves cross-platform consistency

Microarray is a powerful technique used extensively for gene expression analysis. Different technologies are available, but lack of standardization makes it challenging to compare and integrate data. Furthermore, batch-related biases within datasets are common but often not tackled. We have analyzed the same 234 breast cancers on two different microarray platforms. One dataset contained known batch-effects associated with the fabrication procedure used. The aim was to assess the significance of correcting for systematic batch-effects when integrating data from different platforms. We here demonstrate the importance of detecting batch-effects and how tools, such as ComBat, can be used to successfully overcome such systematic variations in order to unmask essential biological signals. Batch adjustment was found to be particularly valuable in the detection of more delicate differences in gene expression. Furthermore, our results show that prober adjustment is essential for integration of gene expression data obtained from multiple sources. We show that high-variance genes are highly reproducibly expressed across platforms making them particularly well suited as biomarkers and for building gene signatures, exemplified by prediction of estrogen-receptor status and molecular subtypes. In conclusion, the study emphasizes the importance of utilizing proper batch adjustment methods when integrating data across different batches and platforms.

PIK3CA mutations, phosphatase and tensin homolog, human epidermal growth factor receptor 2, and insulin-like growth factor 1 receptor and adjuvant tamoxifen resistance in postmenopausal breast cancer patients

Introduction

Inhibitors of the phosphatidylinositol-3-kinase/protein kinase B/mammalian target of rapamycin (PI3K/AKT/mTOR) pathway can overcome endocrine resistance in estrogen receptor (ER) α-positive breast cancer, but companion diagnostics indicating PI3K/AKT/mTOR activation and consequently endocrine resistance are lacking. PIK3CA mutations frequently occur in ERα-positive breast cancer and result in PI3K/AKT/mTOR activation in vitro. Nevertheless, the prognostic and treatment-predictive value of these mutations in ERα-positive breast cancer is contradictive. We tested the clinical validity of PIK3CA mutations and other canonic pathway drivers to predict intrinsic resistance to adjuvant tamoxifen. In addition, we tested the association between these drivers and downstream activated proteins.

Methods

Primary tumors from 563 ERα-positive postmenopausal patients, randomized between adjuvant tamoxifen (1 to 3 years) versus observation were recollected. PIK3CA hotspot mutations in exon 9 and exon 20 were assessed with Sequenom Mass Spectometry. Immunohistochemistry was performed for human epidermal growth factor receptor 2 (HER2), phosphatase and tensin homolog (PTEN), and insulin-like growth factor 1 receptor (IGF-1R). We tested the association between these molecular alterations and downstream activated proteins (like phospho-protein kinase B (p-AKT), phospho-mammalian target of rapamycin (p-mTOR), p-ERK1/2, and p-p70S6K). Recurrence-free interval improvement with tamoxifen versus control was assessed according to the presence or absence of canonic pathway drivers, by using Cox proportional hazard models, including a test for interaction.

Results

PIK3CA mutations (both exon 9 and exon 20) were associated with low tumor grade. An enrichment of PIK3CA exon 20 mutations was observed in progesterone receptor- positive tumors. PIK3CA exon 20 mutations were not associated with downstream-activated proteins. No significant interaction between PIK3CA mutations or any of the other canonic pathway drivers and tamoxifen-treatment benefit was found.

Conclusion

PIK3CA mutations do not have clinical validity to predict intrinsic resistance to adjuvant tamoxifen and may therefore be unsuitable as companion diagnostic for PI3K/AKT/mTOR inhibitors in ERα- positive, postmenopausal, early breast cancer patients.

Epigenetics of estrogen receptor-negative primary breast cancer

Increasingly, breast cancer is being recognized as a heterogeneous disease comprised of molecularly and phenotypically distinct intrinsic tumor subtypes with different clinical outcomes. This biological heterogeneity has significant implications, particularly as it relates to expression profiling of estrogen receptor (ER) status, as classifying breast cancers based on hormone receptor expression impacts not only prognosis but also treatment options and long-term outcomes. Epigenetics has emerged as a promising field for the assessment of hormone receptor status. Epigenetic aberrations have been shown to regulate ER and offer reversible targets for development of new therapies. This review covers ER-negative breast tumor epigenetic aberrations and summarizes the major epigenetic mechanisms governing ER expression and how it impacts treatment of ER-negative breast cancer.

Molecular biology in breast cancer: should molecular classifiers be assessed by conventional tools or by gene expression arrays?

Breast cancer is a complex disease, with heterogeneous presentations and clinical courses. Standard clinico-pathological parameters, relying on single gene or protein characterization determined with sometimes poorly-reproducible technologies, have shown limitations in the classification of the disease and in the prediction of individual patient outcomes and responses to therapy. Gene-expression profiling has revealed great potential to accurately classify breast cancer and define patient prognosis and prediction to anti-cancer therapy. Nevertheless, the performance of molecular classifiers remains sub-optimal, and both technical and conceptual improvements are needed. It is likely that determining the ideal strategy for tailoring treatment of breast cancer will require a more systematic, structured and multi-dimensional approach than in the past. Besides implementing cutting-edge technologies to detect genetic and epigenetic cancer alterations, the future of breast cancer research will in all probability rely on the innovative and multilevel integration of molecular profiles with clinical parameters of the disease and patient-related factors.

TNBCtype: A Subtyping Tool for Triple-Negative Breast Cancer

Motivation:

Triple-negative breast cancer (TNBC) is a heterogeneous breast cancer group, and identification of molecular subtypes is essential for understanding the biological characteristics and clinical behaviors of TNBC as well as for developing personalized treatments. Based on 3,247 gene expression profiles from 21 breast cancer data sets, we discovered six TNBC subtypes from 587 TNBC samples with unique gene expression patterns and ontologies. Cell line models representing each of the TNBC subtypes also displayed different sensitivities to targeted therapeutic agents. Classification of TNBC into subtypes will advance further genomic research and clinical applications.

Result:

We developed a web-based subtyping tool TNBCtype for candidate TNBC samples using our gene expression meta data and classification methods. Given a gene expression data matrix, this tool will display for each candidate sample the predicted subtype, the corresponding correlation coefficient, and the permutation P-value. We offer a user-friendly web interface to predict the subtypes for new TNBC samples that may facilitate diagnostics, biomarker selection, drug discovery, and the more tailored treatment of breast cancer.

PAM50 assay and the three-gene model for identifying the major and clinically relevant molecular subtypes of breast cancer

It has recently been proposed that a three-gene model (SCMGENE) that measures ESR1, ERBB2, and AURKA identifies the major breast cancer intrinsic subtypes and provides robust discrimination for clinical use in a manner very similar to a 50-gene subtype predictor (PAM50). However, the clinical relevance of both predictors was not fully explored, which is needed given that a ~30 % discordance rate between these two predictors was observed. Using the same datasets and subtype calls provided by Haibe-Kains and colleagues, we compared the SCMGENE assignments and the research-based PAM50 assignments in terms of their ability to (1) predict patient outcome, (2) predict pathological complete response (pCR) after anthracycline/taxane-based chemotherapy, and (3) capture the main biological diversity displayed by all genes from a microarray. In terms of survival predictions, both assays provided independent prognostic information from each other and beyond the data provided by standard clinical–pathological variables; however, the amount of prognostic information was found to be significantly greater with the PAM50 assay than the SCMGENE assay. In terms of chemotherapy response, the PAM50 assay was the only assay to provide independent predictive information of pCR in multivariate models. Finally, compared to the SCMGENE predictor, the PAM50 assay explained a significantly greater amount of gene expression diversity as captured by the two main principal components of the breast cancer microarray data. Our results show that classification of the major and clinically relevant molecular subtypes of breast cancer are best captured using larger gene panels.

Estrogen Receptor (ER) mRNA and ER-Related Gene Expression in Breast Cancers That Are 1% to 10% ER-Positive by Immunohistochemistry

Purpose

We examined borderline estrogen receptor (ER) –positive cancers, defined as having 1% to 10% positivity by immunohistochemistry (IHC), to determine whether they show the same global gene-expression pattern and high ESR1 mRNA expression as ER-positive cancers or if they are more similar to ER-negative cancers.

Patients and Methods

ER status was determined by IHC in 465 primary breast cancers and with the Affymetrix U133A gene chip. We compared expressions of ESR1 mRNA and a 106 probe set ER-associated gene signature score between ER-negative (n = 183), 1% to 9% (n = 25), 10% (n = 6), and more than 10% (n = 251) ER-positive cancers. We also assessed the molecular class by using the PAM50 classifier and plotted survival by ER status.

Results

Among the 1% to 9%, 10%, and more than 10% ER IHC–positive patients, 24%, 67%, and 92% were also positive by ESR1 mRNA expression. The average ESR1 expression was significantly higher in the ≥ 10% ER-positive cohorts compared with the 1% to 9% or ER-negative cohort. The average ER gene signature scores were similar for the ER-negative and 1% to 9% IHC-positive patients and were significantly lower than in ≥ 10% ER-positive patients. Among the 1% to 9% ER-positive patients, 8% were luminal B and 48% were basal-like; among the 10% ER-positive patients, 50% were luminal. The overall survival rate of 1% to 9% ER-positive patients with cancer was between those of patients in the ≥ 10% ER-positive and ER-negative groups.

Conclusion

A minority of the 1% to 9% IHC ER–positive tumors show molecular features similar to those of ER-positive, potentially endocrine-sensitive tumors, whereas most show ER-negative, basal-like molecular characteristics. The safest clinical approach may be to use both adjuvant endocrine therapy and chemotherapy in this rare subset of patients.

A three-gene model to robustly identify breast cancer molecular subtypes

BACKGROUND

Single sample predictors (SSPs) and Subtype classification models (SCMs) are gene expression-based classifiers used to identify the four primary molecular subtypes of breast cancer (basal-like, HER2-enriched, luminal A, and luminal B). SSPs use hierarchical clustering, followed by nearest centroid classification, based on large sets of tumor-intrinsic genes. SCMs use a mixture of Gaussian distributions based on sets of genes with expression specifically correlated with three key breast cancer genes (estrogen receptor [ER], HER2, and aurora kinase A [AURKA]). The aim of this study was to compare the robustness, classification concordance, and prognostic value of these classifiers with those of a simplified three-gene SCM in a large compendium of microarray datasets.

METHODS

Thirty-six publicly available breast cancer datasets (n = 5715) were subjected to molecular subtyping using five published classifiers (three SSPs and two SCMs) and SCMGENE, the new three-gene (ER, HER2, and AURKA) SCM. We used the prediction strength statistic to estimate robustness of the classification models, defined as the capacity of a classifier to assign the same tumors to the same subtypes independently of the dataset used to fit it. We used Cohen κ and Cramer V coefficients to assess concordance between the subtype classifiers and association with clinical variables, respectively. We used Kaplan-Meier survival curves and cross-validated partial likelihood to compare prognostic value of the resulting classifications. All statistical tests were two-sided.

RESULTS

SCMs were statistically significantly more robust than SSPs, with SCMGENE being the most robust because of its simplicity. SCMGENE was statistically significantly concordant with published SCMs (κ = 0.65-0.70) and SSPs (κ = 0.34-0.59), statistically significantly associated with ER (V = 0.64), HER2 (V = 0.52) status, and histological grade (V = 0.55), and yielded similar strong prognostic value.

CONCLUSION

Our results suggest that adequate classification of the major and clinically relevant molecular subtypes of breast cancer can be robustly achieved with quantitative measurements of three key genes.

RecurrenceOnline: an online analysis tool to determine breast cancer recurrence and hormone receptor status using microarray data

In the last decades, several gene expression-based predictors of clinical behavior were developed for breast cancer. A common feature of these is the use of multiple genes to predict hormone receptor status and the probability of tumor recurrence, survival or response to chemotherapy. We developed an online analysis tool to compute ER and HER2 status, Oncotype DX 21-gene recurrence score and an independent recurrence risk classification using gene expression data obtained by interrogation of Affymetrix microarray profiles. We implemented rigorous quality control algorithms to promptly exclude any biases related to sample processing, hybridization and scanning. After uploading the raw microarray data, the system performs the complete evaluation automatically and provides a report summarizing the results. The system is accessible online at http://www.recurrenceonline.com . We validated the system using data from 2,472 publicly available microarrays. The validation of the prediction of the 21-gene recurrence score was significant in lymph node negative patients (Cox-Mantel, P = 5.6E-16, HR = 0.4, CI = 0.32-0.5). A correct classification was obtained for 88.5% of ER- and 90.5% of ER + tumors (n = 1,894). The prediction of recurrence risk in all patients by using the mean of the independent six strongest genes (P < 1E-16, HR = 2.9, CI = 2.5-3.3), of the four strongest genes in lymph node negative ER positive patients (P < 1E-16, HR = 2.8, CI = 2.2-3.5) and of the three genes in lymph node positive patients (P = 3.2E-9, HR = 2.5, CI = 1.8-3.4) was highly significant. In summary, we integrated available knowledge in one platform to validate currently used predictors and to provide a global tool for the online determination of different prognostic parameters simultaneously using genome-wide microarrays.