Breast Cancer Subtype Classification Research Articles

Bovine serum albumin encapsulation of near infrared fluorescent nano-probe with low nonspecificity and cytotoxicity for imaging of HER2-positive breast cancer cells

Breast cancer with HER2 overexpressing type links to malignant tumor growth and poor clinical outcome. Successful development of sensitive and selective nano-probe for identification of HER2-positive breast cancer cells is of great importance for breast cancer early diagnosis, subtype classification, and treatment planning. Herein, we report a HER2 antibody conjugated near infrared (NIR) emitted MnCuInS/ZnS qumtun dots (QDs) encapsulated bovine serum albumin (BSA) nano-probe for accurately targeted imaging of HER2-positive breast cancer cells. This NIR nano-probe shows good biocompatibility, low nonspecificity and cytotoxicity, high colloidal stability, and allows HER2-positive breast cancer cell identification with good selectivity. The practicality of this targeted NIR fluorescent nano-probe was proved by successful identifying HER2-positive breast cancer cells from HER2-negative breast cancer cells, which indicates that it can be efficiently applied in selective screening of HER2 overexpressing cancer cells, and provide a platform for the strategy design on the distinction of different breast cancer subtypes.

Abstract 5255: Breast cancer subtype classification using a multi-gene expression profile

Abstract Background: Immunohistochemistry (IHC) is the method conventionally used in clinical practice to define breast cancer (BC) subtypes. This technique is subjective and semi-quantitative. Molecular approaches like multigene panels are increasingly used in clinical practice for their reliability and accuracy. This preliminary study was performed to assess the applicability of the NanoString Breast Cancer 360™ panel, which analyze 770 transcripts of genes involved in a number of signatures useful to define the molecular subtype. We evaluated the correlation between the gene expression levels with the IHC values of conventional biomarkers on 12 BC formalin-fixed paraffin-embedded (FFPE) samples. Methods: Immunostaining was performed by using Ventana Benchmark Ultra system. Anti-ER, -PR, -Ki67 and -HER2 Ventana antibodies were used. Tumors were classified according to the most recent St. Gallen classification. RNA was isolated from FFPE tumor with AllPrep DNA/RNA FFPE Kit and quantified by Nanodrop. The Breast Cancer 360™ panel assay was performed according to manufacturer’s instructions. The normalized count for each gene analyzed using nSolver analysis software was compared with the respective IHC biomarker status by Mann-Whitney test to assess the correlation between the two methods. The mean mRNA count was compared between luminal and triple negative BC by unpaired T test. All statistical analyses were performed by Graphpad Prism software 5. Results: The 12 BC samples were classified in 3 luminal A, 3 luminal B (HER2+) and 6 triple negative tumors by IHC. The expression level of ESR1, PGR, MKI67 and ERBB2 genes showed a statistically significant correlation with the corresponding protein expression (p=0.0022, p=0.0054, p=0.0025 and p=0.0091, respectively). A significant difference in the expression levels of ERBB2 gene between HER2+ and HER2- BC was observed (mean mRNA counts 16848 vs 1307, p=0,035). Furthermore, ESR1 expression was significantly higher in ER+ than in ER- BC (mean mRNA counts 9105 vs 195, p=0,021). We also evaluated the differential expression between luminal and triple negative BC for AR, EGFR, HIF1A, MKI67, MYC, NOTCH1, RAD51 genes. The expression levels for all these genes were significantly higher in triple negative cases than luminal BC, except for AR that was higher in luminal tumors (p<0.05). Conclusions: In this study Breast Cancer 360™ panel showed high concordance with IHC data. We are aware that our preliminary results require validation in a larger case series to establish the real value of this multi-gene expression profile in BC. However, our results highlight the potential of this molecular approach to properly identify tumor subtypes, allowing a better management of BC patients. Citation Format: Sara Ravaioli, Francesca Pirini, Andrea Rocca, Maurizio Puccetti, Massimiliano Bonafè, Giovanni Martinelli, Sara Bravaccini. Breast cancer subtype classification using a multi-gene expression profile [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5255.

Machine Learning-Based Analysis of MR Multiparametric Radiomics for the Subtype Classification of Breast Cancer.

Objective: To investigate whether machine learning analysis of multiparametric MR radiomics can help classify immunohistochemical (IHC) subtypes of breast cancer.Study design: One hundred and thirty-four consecutive patients with pathologically-proven invasive ductal carcinoma were retrospectively analyzed. A total of 2,498 features were extracted from the DCE and DWI images, together with the new calculated images, including DCE images changing over six time points (DCEsequential) and DWI images changing over three b-values (DWIsequential). We proposed a novel two-stage feature selection method combining traditional statistics and machine learning-based methods. The accuracies of the 4-IHC classification and triple negative (TN) vs. non-TN cancers were assessed.Results: For the 4-IHC classification task, the best accuracy of 72.4% was achieved based on linear discriminant analysis (LDA) or subspace discrimination of assembled learning in conjunction with 20 selected features, and only small dependent emphasis of Kendall-tau-b for sequential features, based on the DWIsequential with the LDA model, yielding an accuracy of 53.7%. The linear support vector machine (SVM) and medium k-nearest neighbor using eight features yielded the highest accuracy of 91.0% for comparing TN to non-TN cancers, and the maximum variance for DWIsequential alone, together with a linear SVM model, achieved an accuracy of 83.6%.Conclusions: Whole-tumor radiomics on MR multiparametric images, DCE images changing over time points, and DWI images changing over different b-values provide a non-invasive analytical approach for breast cancer subtype classification and TN cancer identification.

Personalized medicine in breast cancer: pharmacogenomics approaches.

: Breast cancer is the fifth cause of cancer death among women worldwide and represents a global health concern due to the lack of effective therapeutic regimens that could be applied to all disease groups. Nowadays, strategies based on pharmacogenomics constitute novel approaches that minimize toxicity while maximizing drug efficacy; this being of high importance in the oncology setting. Besides, genetic profiling of malignant tumors can lead to the development of targeted therapies to be included in effective drug regimens. Advances in molecular diagnostics have revealed that breast cancer is a multifaceted disease, characterized by inter-tumoral and intra-tumoral heterogeneity and, unlike the past, molecular classifications based on the expression of individual biomarkers have led to devising novel therapeutic strategies that improve patient survival. In this review, we report and discuss the molecular classification of breast cancer subtypes, the heterogeneity resource, and the advantages and disadvantages of current drug regimens with consideration of pharmacogenomics in response and resistance to treatment.

Supervivencia en cáncer de mama por subtipo mediante inmunohistoquímica: Un estudio retrospectivo

Breast cancer subtype classification according to hormone receptors (HR) and human epidermal growth factor receptor 2 (HER2) using immunohistochemistry is the standard practice for therapeutic decision making. To design future studies information on characteristics and survival of each subtype is essential. We conducted a retrospective study to analyze clinical and pathologic features as well as survival data according to breast cancer immunohistochemistry subtype. There were 211 women with a RH(+)/HER2(-) breast cancer subtype, 53 HR(+)/HER2(+), 16 HER2(+) and 23 HR(-)/HER2(-), with a median overall survival in months of 39 (20.5-62.7), 42 (25.5-65), 42 (13.7-67.7) and 26 (11-78), respectively, for a 3.7 hazard ratio of death (95% Confidence Interval [CI]: 1.3-10.3) for the triple negative group as compared to the HR(+)/HER2(-) group (p = 0.01). HR positive subtypes by immunohistochemistry where most frequent and showed a greater overall survival compared to the triple negative subtype.

Classifying Breast Cancer Subtypes Using Multiple Kernel Learning Based on Omics Data.

It is very significant to explore the intrinsic differences in breast cancer subtypes. These intrinsic differences are closely related to clinical diagnosis and designation of treatment plans. With the accumulation of biological and medicine datasets, there are many different omics data that can be viewed in different aspects. Combining these multiple omics data can improve the accuracy of prediction. Meanwhile; there are also many different databases available for us to download different types of omics data. In this article, we use estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2) to define breast cancer subtypes and classify any two breast cancer subtypes using SMO-MKL algorithm. We collected mRNA data, methylation data and copy number variation (CNV) data from TCGA to classify breast cancer subtypes. Multiple Kernel Learning (MKL) is employed to use these omics data distinctly. The result of using three omics data with multiple kernels is better than that of using single omics data with multiple kernels. Furthermore; these significant genes and pathways discovered in the feature selection process are also analyzed. In experiments; the proposed method outperforms other state-of-the-art methods and has abundant biological interpretations.

Selection of Radiomic Features for the Classification of Triple-negative Breast Cancer Based on Radiogenomics

Subtype classification of breast cancer by analyzing the gene expression profile of cancer cells is becoming a standard procedure. Breast cancer subtype classification is more useful than the conventional method because the characteristics of subtype classification is directly connected with the treatment method. However, genetic testing is invasive, and a part of cancer cells may not represent the overall nature of the cancer. In the computer-aided diagnosis (CAD) scheme for differentiation of triple-negative breast cancer (TNBC) by estimating the genetic properties of cancer based on Radiogenomics, principal component analysis (PCA) and least absolute shrinkage and selection operator (Lasso) were used for reducing the dimension of radiomic features, and we compared usefulness of both. We collected 81 magnetic resonance (MR) images, which included 30 TNBC and 51 others, from the public database. From the MR slice images, we selected the slice containing the largest area of the cancer and manually marked the cancer region. We subsequently calculated 294 radiomic features in the cancer region, and reduced the dimension of radiomic features. Finally, linear discriminant analysis, with the dimensionally compressed 10 image features, was used for distinguishing between TNBC and others. Area under the curve (AUC) was 0.60 when we used PCA, whereas AUC was 0.70 when we used Lasso (p=0.0058). Therefore, Lasso is useful for the determination of radiomic features in Radiogenomics.

A comparative study of multiclass feature selection on RNAseq and microarray data

Gene expression profiles are widely used for identifying phenotype-specific biomarkers in clinical cancer research. By examining important genes expressed in different phenotypes, patients can be classified into different treatment groups. Microarray and RNAseq are the two leading technologies to measure gene expression data. However, due to the heterogeneity of the two platforms, their selected genes are different. In this project, we systematically compared the breast cancer subtype classification accuracies from the selected genes by four popular multiclass feature selection algorithms and discussed the strengths and weakness of selected genes across different platforms and cohorts. Our results showed that the classification of selected genes performs best within the same platform across different cohorts. It suggested that merging the dataset belonging to the same platform will increase the statistical power and improve the prediction accuracy of the selected gene for multiclass classification analysis.

An isomiR expression panel based novel breast cancer classification approach using improved mutual information

BackgroundGene expression-based profiling has been used to identify biomarkers for different breast cancer subtypes. However, this technique has many limitations. IsomiRs are isoforms of miRNAs that have critical roles in many biological processes and have been successfully used to distinguish various cancer types. Biomarker isomiRs for identifying different breast cancer subtypes has not been investigated. For the first time, we aim to show that isomiRs are better performing biomarkers and use them to explain molecular differences between breast cancer subtypes.ResultsIn this study, a novel method is proposed to identify specific isomiRs that faithfully classify breast cancer subtypes. First, as a null hypothesis method we removed the lowly expressed isomiRs from small sequencing data generated from diverse breast cancers types. Second, we developed an improved mutual information-based feature selection method to calculate the weight of each isomiR expression. The weight of isomiR measures the importance of a given isomiR in classifying breast cancer subtypes. The improved mutual information enables to apply the dataset in which the feature is continuous data and label is discrete data; whereby, the traditional mutual information cannot be applied in this dataset. Finally, the support vector machine (SVM) classifier is applied to find isomiR biomarkers for subtyping.ConclusionsHere we demonstrate that isomiRs can be used as biomarkers in the identification of different breast cancer subtypes, and in addition, they may provide new insights into the diverse molecular mechanisms of breast cancers. We have also shown that the classification of different subtypes of breast cancer based on isomiRs expression is more effective than using published gene expression profiling. The proposed method provides a better performance outcome than Fisher method and Hellinger method for discovering biomarkers to distinguish different breast cancer subtypes. This novel technique could be directly applied to identify biomarkers in other diseases.

Heterogeneous Domain Adaptation for IHC Classification of Breast Cancer Subtypes.

Increasingly, multiple parallel omics datasets are collected from biological samples. Integrating these datasets for classification is an open area of research. Additionally, whilst multiple datasets may be available for the training samples, future samples may only be measured by a single technology requiring methods which do not rely on the presence of all datasets for sample prediction. This enables us to directly compare the protein and the gene profiles. New samples with just one set of measurements (e.g., just protein) can then be mapped to this latent common space where classification is performed. Using this approach, we achieved an improvement of up to 12 percent in accuracy when classifying samples based on their protein measurements compared with baseline methods which were trained on the protein data alone. We illustrate that the additional inclusion of the gene expression or protein expression in the training process enabled the separation between the classes to become clearer.

Robustness of biomarker determination in breast cancer by RT-qPCR: impact of tumor cell content, DCIS and non-neoplastic breast tissue

BackgroundTissue heterogeneity in formalin-fixed paraffin-embedded (FFPE) breast cancer specimens may affect the accuracy of reverse transcription quantitative real-time PCR (RT-qPCR). Herein, we tested the impact of tissue heterogeneity of breast cancer specimen on the RT-qPCR-based gene expression assay MammaTyper®.MethodsMammaTyper® quantifies the mRNA expression of the four biomarkers ERBB2, ESR1, PGR, and MKI67. Based on pre-defined cut-off values, this molecular in vitro diagnostic assay permits binary marker classification and determination of breast cancer subtypes as defined by St Gallen 2013. In this study, we compared data from whole FFPE sections with data obtained in paired RNA samples after enrichment for invasive carcinoma via macro- or laser-capture micro-dissection.ResultsCompared to whole sections, removal of surrounding adipose tissue by macrodissection generated mean absolute 40-ddCq differences of 0.28–0.32 cycles for all four markers, with ≥90% concordant binary classifications. The mean raw marker Cq values in the adipose tissue were delayed by 6 to 7 cycles compared with the tumor-enriched sections, adding a trivial linear fold change of 1.0078 to 1.0156. Comparison of specimens enriched for invasive tumor with whole sections with as few as 20% tumor cell content resulted in mean absolute differences that remained on average below 0.59 Cq. The mean absolute difference between whole sections containing up to 60% ductal carcinoma in situ (DCIS) and specimens after dissection of DCIS was only 0.16–0.25 cycles, although there was a tendency for higher gene expression in DCIS. Observed variations were related to small size of samples and proximity of values to the limit of detection.ConclusionExpression of ESR1, PGR, ERBB2 and MKI67 by MammaTyper® is robust in clinical FFPE samples. Assay performance was unaffected by adipose tissue and was stable in samples with as few as 20% tumor cell content and up to 60% DCIS.

Robust identification of target genes and outliers in triple-negative breast cancer data.

Correct classification of breast cancer subtypes is of high importance as it directly affects the therapeutic options. We focus on triple-negative breast cancer which has the worst prognosis among breast cancer types. Using cutting edge methods from the field of robust statistics, we analyze Breast Invasive Carcinoma transcriptomic data publicly available from The Cancer Genome Atlas data portal. Our analysis identifies statistical outliers that may correspond to misdiagnosed patients. Furthermore, it is illustrated that classical statistical methods may fail to identify outliers due to their heavy influence, prompting the need for robust statistics. Using robust sparse logistic regression we obtain 36 relevant genes, of which ca. 60% have been previously reported as biologically relevant to triple-negative breast cancer, reinforcing the validity of the method. The remaining 14 genes identified are new potential biomarkers for triple-negative breast cancer. Out of these, JAM3, SFT2D2, and PAPSS1 were previously associated to breast tumors or other types of cancer. The relevance of these genes is confirmed by the new DetectDeviatingCells outlier detection technique. A comparison of gene networks on the selected genes showed significant differences between triple-negative breast cancer and non-triple-negative breast cancer data. The individual role of FOXA1 in triple-negative breast cancer and non-triple-negative breast cancer, and the strong FOXA1-AGR2 connection in triple-negative breast cancer stand out. The goal of our paper is to contribute to the breast cancer/triple-negative breast cancer understanding and management. At the same time it demonstrates that robust regression and outlier detection constitute key strategies to cope with high-dimensional clinical data such as omics data.

Signal translational efficiency between mRNA expression and antibody-based protein expression for breast cancer and its subtypes from cell lines to tissue

Background: Although gene transcripts and protein expression have been utilised to classify breast cancer subtypes, it is not clear whether the observed measurement of gene transcript abundance can predict its protein expression. Herein, we attempt to address gene transcript/protein associations using publically-available data on breast cancer tumour tissues and cell lines. Method: Correlation analysis between mRNAs and Reverse-phase protein arrays (RPPA) among 421 primary breast tumours and 33 breast cancer cell lines was conducted. Highly concordant proteins/genes were further analysed in different breast cancer subtypes. Results: The overall accordance of mRNA/RPPA correlation between cell lines and primary tissue is R2 = 0.71. Since most of these genes are well known drug targets, highly concordant gene/RPPA associations not only confirm that these gene transcripts can serve as biomarkers for their protein products in drug target selection, but also imply that breast cancer cell lines can serve as good models for primary breast cancer tumours.

Diagnostic Value of ER, PR, FR and HER-2-Targeted Molecular Probes for Magnetic Resonance Imaging in Patients with Breast Cancer

Background/Aims: Smart molecular probes are required in the application of Magnetic resonance imaging (MRI) for biochemical and clinical research. This study aims to investigate the diagnostic values of estrogen receptor (ER), progesterone receptor (PR), folate receptor (FR) and human epidermal growth factor receptor 2 (HER-2)-targeted molecular probes in the MRI diagnosis of breast cancer. Methods: Initially, a total of 508 female breast cancer patients were selected for breast cancer subtype classification by immunohistochemistry. Subsequently, the tumor size, lymph node metastasis, and histological grade of different breast cancer subtypes were compared. Molecular probes of Ab-ER-USPIO, Ab-PR-USPIO, Ab-FR-USPIO and Ab-HER-2-USPIO were constructed and screened. The specific binding of molecular probes to breast cancer cells was detected both in vitro and in vivo by Prussian blue staining and MRI using T1 and T2 weighted images. Finally, in vivo toxicity of Ab-HER-2-USPIO was analyzed using hematoxylin and eosin staining. Results: We identified the following subtypes of breast cancer: Luminal A (ER-positive, FR-positive, HER-2-negative), Luminal B (ER-positive, FR-positive, HER-2-positive), HER-2 overexpression (ER-negative, FR-negative, HER-2-positive), and triple-negative breast cancer (ER-negative, FR-negative, HER-2-negative). Featuring favorable in vitro biocompatibility and low in vivo toxicity, Ab-HER-2-USPIO can specifically bind to breast cancer cells BT47 and SKBR3, thus enhancing the quality of T1 weighted MRI images. Conclusion: The results indicate that HER-2-targeted MRI molecular probes may be used in the clinical diagnosis of breast cancer and facilitate the development of promising strategies for breast cancer treatments.

Androgen receptor expression in hormone-negative breast cancers and its prognostic significance

Background/aim Breast carcinoma is a common, yet heterogeneous aggressive disease affecting relatively young patients. The androgen receptor (AR) is expressed in majority of breast cancers and across the main breast cancer subtypes. The aim of this study was to evaluate AR expression in hormone-negative breast cancer subtypes. Materials and methods Sixty cases of breast cancer were involved in this study; the samples were received in the Department of Pathology of Kasr El-Aini Hospital, Cairo University, Egypt. The expression of AR and human epidermal growth factor receptor-2 receptors were studied by immunohistochemistry in 60 formalin-fixed paraffin-embedded selected hormone-negative breast cancer surgical specimens. The immunohistochemistry expression of the marker was correlated with the clinicopathological variables. Results Of the hormone-negative cases, 61.6% show positive AR expression, 89% of which are invasive duct carcinoma, 68.3% are associated with ductal carcinoma in situ, and 55% are human epidermal growth factor receptor-2-enriched subtype. A significant correlation was found between the AR expression and tumor type. There is no evident significant correlation with tumor grade, multicentricity or lymphovascular invasion. Conclusion The AR has recently emerged as a useful marker for the further refinement of breast cancer subtype classification. Antiandrogens are thought to markedly enhance treatments and to be the first targeted therapy in hormone-negative breast cancer cases.

Analysis of breast cancer subtypes by AP-ISA biclustering

BackgroundGene expression profiling has led to the definition of breast cancer molecular subtypes: Basal-like, HER2-enriched, LuminalA, LuminalB and Normal-like. Different subtypes exhibit diverse responses to treatment. In the past years, several traditional clustering algorithms have been applied to analyze gene expression profiling. However, accurate identification of breast cancer subtypes, especially within highly variable LuminalA subtype, remains a challenge. Furthermore, the relationship between DNA methylation and expression level in different breast cancer subtypes is not clear.ResultsIn this study, a modified ISA biclustering algorithm, termed AP-ISA, was proposed to identify breast cancer subtypes. Comparing with ISA, AP-ISA provides the optimized strategy to select seeds and thresholds in the circumstance that prior knowledge is absent. Experimental results on 574 breast cancer samples were evaluated using clinical ER/PR information, PAM50 subtypes and the results of five peer to peer methods. One remarkable point in the experiment is that, AP-ISA divided the expression profiles of the luminal samples into four distinct classes. Enrichment analysis and methylation analysis showed obvious distinction among the four subgroups. Tumor variability within the Luminal subtype is observed in the experiments, which could contribute to the development of novel directed therapies.ConclusionsAiming at breast cancer subtype classification, a novel biclustering algorithm AP-ISA is proposed in this paper. AP-ISA classifies breast cancer into seven subtypes and we argue that there are four subtypes in luminal samples. Comparison with other methods validates the effectiveness of AP-ISA. New genes that would be useful for targeted treatment of breast cancer were also obtained in this study.

Breast cancer subtype discrimination using standardized 4-IHC and digital image analysis.

Breast cancer is a heterogeneous disease. Surrogate classification of intrinsic subtypes of invasive carcinomas by combined immunohistochemistry for estrogen receptor (ER), progesterone receptor (PR), HER2, and Ki67 (4-IHC) has increased steadily since the 2011 St Gallen symposium, due to its rapid subtyping of tumors at a reasonable cost. An important step in improving 4-IHC reproducibility and reliability will be to provide reference values from the routine use of standardized 4-IHC followed by image analysis. The aims of the current study were (1) to analyze invasive breast carcinomas using standardized 4-IHC and quantitative image analysis and (2) to compare the results obtained in the classification of biological subtypes using current Ki67 and PR threshold values proposed by different authors to sub-classifying the luminal A-like and the luminal B-like (HER2-negative) subtypes. Five hundred twenty-one tumors were analyzed by standardized immunohistochemistry, with automatic image analysis, and HER2 FISH technique. Positivity for ER was found in 82.7% and for PR in 70.1% of cases. Using the Allred scoring system, hormone receptor results showed a bimodal distribution, particularly for ER. HER2 positivity was found in 15.7% of cases, and the mean Ki67 score was 32.3%. Using the most recently proposed surrogate definitions for the classification of luminal breast cancer subtypes, the percentages of different subtypes that we found were similar to those published with genomic platforms: 40.7% luminal A-like, 32.4% luminal B-like/HER2-negative, 9.8% luminal B-like/HER2-positive, 6.0% HER2-positive, and 11.1% triple negative. Standardized 4-IHC with automatic image analysis constitutes a low-cost method for surrogate definitions of biological subtypes of breast cancer that delivers accurate results in a day.

Assessment of Ki67 expression for breast cancer subtype classification and prognosis in the Nurses' Health Study.

Ki67 is a proliferation marker commonly assessed by immunohistochemistry in breast cancer, and it has been proposed as a clinical marker for subtype classification, prognosis, and prediction of therapeutic response. However, the clinical utility of Ki67 is limited by the lack of consensus on the optimal cut point for each application. We assessed Ki67 by immunohistochemistry using Definiens digital image analysis (DIA) in 2653 cases of incident invasive breast cancer diagnosed in the Nurses' Health Study from 1976 to 2006. Ki67 was scored as continuous percentage of positive tumor cells, and dichotomized at various cut points. Multivariable hazard ratios (HR) and 95% confidence intervals (CI) were calculated using Cox regression models for distant recurrence, breast cancer-specific mortality and overall mortality in relation to luminal subtypes defined with various Ki67 cut points, adjusting for breast cancer prognostic factors, clinico-pathologic features and treatment. DIA was highly correlated with manual scoring of Ki67 (Spearman correlation ρ=0.86). Mean Ki67 score was higher in grade-defined luminal B (12.6%), HER2-enriched (17.9%) and basal-like (20.6%) subtypes compared to luminal A (8.9%). In multivariable-adjusted models, luminal B tumors had higher breast cancer-specific mortality compared to luminal A cancer classified using various cut points for Ki67 positivity including the 14% cut point routinely reported in the literature (HR 1.38, 95% CI 1.11-1.72, p=0.004). There was no significant difference in clinical outcomes for ER- tumors according to Ki67 positivity defined at various cut points. Assessment of Ki67 in breast tumors by DIA was a robust and quantitative method. Results from this large prospective cohort study provide support for the clinical relevance of using Ki67 at the 14% cut point for luminal subtype classification and breast cancer prognosis.

Abstract 4734: miR-9-5p expression in breast cancer correlates with hormone receptor status and affects patients survival

Abstract miR-9 (hsa-miR-9-5p, miR-9-5p) is a highly conserved microRNA (miRNA) primarily expressed in the central nervous system. The three promoters of miR-9 (miR-9-1 at 1q22, miR-9-2 at 5q14.3 and miR-9-3 at 15q26.1) are embedded within CpG islands, and hypermethylation in at least one of these regions has been reported in several tumor types, thus miR-9-5p has been designated as an epigenetically regulated miRNA. However, only a partial or none correlation between methylation status of miR-9 loci and miR-9-5p expression has been demonstrated so far. Furthermore, the expression levels of miR-9-5p resulted highly variable across different carcinomas and conflicting evidences also exist about its functional role within the tumor context. In light of this, we took the effort of clarifying the role of miR-9-5p in primary breast cancer tissues and of evaluating its potential as clinical relevant prognostic biomarker. We analysed miR-9-5p expression and miR-9 promoters methylation status in 129 breast cancer cases and 12 normal breast tissues (NBTs), by qRT-PCR and MSP respectively. Overall, miR-9-5p was increased in tumors compared to NBTs (P<0.001). No significant correlation between promoter methylation and miR-9-5p expression was found either in tumors or in NBTs. Interestingly, miR-9-5p expression was inversely correlated with ER and PgR positivity (P=0.004 and P=0.003, respectively). Consistently, triple negative breast cancer (TNBC) showed the highest miR-9-5p levels and luminal subtype the lowest (P=0.04). The analysis of the TCGA Breast Cancer dataset (n=256) confirmed our results and further demonstrated that miR-9-5p was differentially expressed between the two luminal subtypes (P=0.009) and HER2-amplified compared to TNBCs (P<0.0001). In Univariable Cox regression analysis miR-9-5p expression was significantly associated with higher risk of death (P=0.023). Ingenuity Pathway Analysis exploring the putative interactions among miR-9-5p, ER and PgR upstream and downstream regulators suggested a regulatory loop by which miR-9-5p is induced by steroid hormone receptor and acts within hormone-receptor regulated pathways. Our data suggest miR-9-5p may help refine the molecular classification of breast cancer subtypes and provide additional prognostic information. Citation Format: Raffaela Barbano, Barbara Pasculli, Michelina Rendina, Andrea Fontana, Caterina Fusilli, Massimiliano Copetti, Stefamo Castellana, Vanna Maria Valori, Maria Morritti, Paolo Graziano, Luigi Ciuffreda, Michelina Coco, Francesco Picardo, Tommaso Mazza, Ella Evron, Roberto Murgo, Evaristo Maiello, Manel Esteller, Vito Michele Fazio, Paola Parrella. miR-9-5p expression in breast cancer correlates with hormone receptor status and affects patients survival [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4734. doi:10.1158/1538-7445.AM2017-4734

Abstract 4539: Novel feature selection strategies for enhanced predictive modeling and deep learning in the biosciences

Abstract Artificial Intelligence (AI) is the single most transformative technology in human history. Advancements in personalized medicine depend upon significantly furthering our current understanding of how genetic variation and somatic mutation regulate aberrant gene activity and subsequent disease biology, including the myriad of dysregulated molecular mechanisms of cancer. To this end, we have built a robust AI method to precisely assess pathogenicity for all genomic missense variants. Coupled with our advanced deepCODE feature selection strategy for constructing deep learning models, we can quantitatively integrate a priori pathway-based biological knowledge with multiple types of high-throughput omics data. This approach significantly improves performance of established classification methods via enhanced semantic interoperability for mapping between multiple biomedical ontologies and the subsequent identification of genes and molecular pathways more highly predictive of cancer type etiology. Specifically, our deepCODE AI methods greatly improve tumor subtype and drug-response classification accuracies by combining tumor DNA- and RNA-seq data. Our results indicate classification of human breast and lung cancer subtypes with 96% and greater than 99% accuracy, respectively. Our novel multinomial statistical-learning method, which was validated with 6 cancer type patient-derived tumor xenograft (PDTX) mouse models, achieved greater than 95% accuracy across 10,000 human tumors within the collection of 28 TCGA cancer types. Finally, our stepwise deepCODE drug target discovery approach systematically integrates our deep learning classification strategy with advanced Bayesian Network modeling to identify causative driver genes, representing probable novel drug targets. This stepwise machine learning approach also serves as a unique means to assess drug efficacy. Taken together, our advanced deepCODE AI tools greatly expand the means to fully understand the underlying molecular determinants of human cancer. We have developed user-friendly decision support and research tools for in-depth tumor analysis and large-scale case-control methods that afford investigators access to our deepCODE variant annotation algorithms. Citation Format: Pengwei Yang, Ryan Abo, Chang Liu, Zehua Chen, Haiguo Wu, Jike Cui, Chandri Yandava, Shannon T. Bailey, Curt Balch, Jeffery R. Gulcher, Thomas W. Chittenden. Novel feature selection strategies for enhanced predictive modeling and deep learning in the biosciences [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4539. doi:10.1158/1538-7445.AM2017-4539