Dataset Information

Predicting breast cancer response to neoadjuvant chemotherapy based on tumor vascular features in needle biopsies.

ABSTRACT: In clinical breast cancer intervention, selection of the optimal treatment protocol based on predictive biomarkers remains an elusive goal. Here, we present a modeling tool to predict the likelihood of breast cancer response to neoadjuvant chemotherapy using patient specific tumor vasculature biomarkers. A semi-automated analysis was implemented and performed on 3990 histological images from 48 patients, with 10-208 images analyzed for each patient. We applied a histology-based model to resected primary breast cancer tumors (n = 30), and then evaluated a cohort of patients (n = 18) undergoing neoadjuvant chemotherapy, collecting pre- and post-treatment pathology specimens and MRI data. We found that core biopsy samples can be used with acceptable accuracy (r = 0.76) to determine histological parameters representative of the whole tissue region. Analysis of model histology parameters obtained from tumor vasculature measurements, specifically diffusion distance divided by radius of drug source (L/rb) and blood volume fraction (BVF), provides a statistically significant separation of patients obtaining a pathologic complete response (pCR) from those that do not (Student's t-test; P < 0.05). With this model, it is feasible to evaluate primary breast tumor vasculature biomarkers in a patient specific manner, thereby allowing a precision approach to breast cancer treatment.

SUBMITTER: Brocato TA

PROVIDER: S-EPMC6538356 | biostudies-literature | 2019 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Predicting breast cancer response to neoadjuvant chemotherapy based on tumor vascular features in needle biopsies.

Brocato Terisse A TA Brown-Glaberman Ursa U Wang Zhihui Z Selwyn Reed G RG Wilson Colin M CM Wyckoff Edward F EF Lomo Lesley C LC Saline Jennifer L JL Hooda-Nehra Anupama A Pasqualini Renata R Arap Wadih W Brinker C Jeffrey CJ Cristini Vittorio V

JCI insight 20190305

In clinical breast cancer intervention, selection of the optimal treatment protocol based on predictive biomarkers remains an elusive goal. Here, we present a modeling tool to predict the likelihood of breast cancer response to neoadjuvant chemotherapy using patient specific tumor vasculature biomarkers. A semi-automated analysis was implemented and performed on 3990 histological images from 48 patients, with 10-208 images analyzed for each patient. We applied a histology-based model to resected ...[more]

PMID: 30835256

Similar Datasets

Project description:Background: The preoperative identification of neoadjuvant chemotherapy (NAC) treatment responsiveness in breast cancer (BC) patients is advantageous for tailoring treatment regimens. There is a relative scarcity in the current research exploring NAC treatment responsive biomarkers using bulk sequencing data obtained from fine-needle aspiration (FNA). Materials and Methods: Limma was employed for the selection of differentially expressed genes. Additionally, WGCNA, machine learning, and Genetic Perturbation Similarity Analysis (GPSA) were utilized to identify key genes associated with NAC treatment response. ConsensusClusterPlus was employed for unsupervised clustering. Rt-qPCR and WB were conducted to assess gene expression and protein levels in clinical tissues and cell lines. The Seahorse XF96 Extracellular Flux Analyzer was utilized to evaluate Extracellular Acidification Rate (ECAR) and Oxygen Consumption Rate (OCR). The "pRRophetic" package was used for drug sensitivity prediction, while CB-Dock2 was applied for molecular docking and optimal pose presentation. Spatial transcriptomic analysis was based on the CROST database. Results: Eleven biomarkers were identified associated with NAC treatment response in BC patients, with FOXA1 identified as a pivotal hub gene among them. The expression levels of FOXA1 showed a significant positive correlation with genomic stability and a marked negative correlation with the homologous recombination deficiency (HRD) score. Downregulation of the FOXA1 gene resulted in reduced glycolysis in MCF-7 cells.Additionally, FOXA1 were found to serve as a biomarker for both NAC and PARP inhibitor treatment sensitivity in BC patients. Spatial transcriptomic analysis indicates significantly elevated infiltration of T follicular helper (T-FH) cells and mast cells surrounding tumors exhibiting high FOXA1 expression. Conclusion: In summary, our study involved the analysis of diverse sequencing datasets derived from various FNA samples to identify biomarkers sensitive to NAC, thereby offering novel insights into resources for future personalized clinical treatment strategies.

Project description:BackgroundPatients with pathologic complete response (pCR) to neoadjuvant chemotherapy for invasive breast cancer (BC) have better outcomes, potentially warranting less extensive surgical and systemic treatments. Early prediction of treatment response could aid in adapting therapies.MethodsOn-treatment biopsies from 297 patients with invasive BC in three randomized, prospective neoadjuvant trials were assessed (GeparQuattro, GeparQuinto, GeparSixto). BC quantity, tumor-infiltrating lymphocytes (TILs), and the proliferation marker Ki-67 were compared to pre-treatment samples. The study investigated the correlation between residual cancer, changes in Ki-67 and TILs, and their impact on pathologic complete response (pCR) and disease-free survival (DFS).ResultsAmong the 297 samples, 138 (46%) were hormone receptor-positive (HR+)/human epidermal growth factor 2-negative (HER2-), 87 (29%) were triple-negative (TNBC), and 72 (24%) were HER2+. Invasive tumor cells were found in 70% of on-treatment biopsies, with varying rates across subtypes (HR+/HER2-: 84%, TNBC: 62%, HER2+: 51%; p < 0.001). Patients with residual tumor on-treatment had an 8% pCR rate post-treatment (HR+/HER2-: 3%, TNBC: 19%, HER2+: 11%), while those without any invasive tumor had a 50% pCR rate (HR+/HER2-: 27%; TNBC: 48%, HER2+: 66%). Sensitivity for predicting residual disease was 0.81, with positive and negative predictive values of 0.92 and 0.50, respectively. Increasing TILs from baseline to on-treatment biopsy (if residual tumor was present) were linked to higher pCR likelihood in the overall cohort (OR 1.034, 95% CI 1.013-1.056 per % increase; p = 0.001) and with a longer DFS in TNBC (HR 0.980, 95% CI 0.963-0.997 per % increase; p = 0.026). Persisting or increased Ki-67 was associated with with lower pCR probability in the overall cohort (OR 0.957, 95% CI 0.928-0.986; p = 0.004) and shorter DFS in TNBC (HR 1.023, 95% CI 1.001-1.047; p = 0.04).ConclusionOn-treatment biopsies can predict patients unlikely to achieve pCR post-therapy. This could facilitate therapy adjustments for TNBC or HER2 + BC. They also might offer insights into therapy resistance mechanisms. Future research should explore whether standardized or expanded sampling enhances the accuracy of on-treatment biopsy procedures. Trial registration GeparQuattro (EudraCT 2005-001546-17), GeparQuinto (EudraCT 2006-005834-19) and GeparSixto (EudraCT 2011-000553-23).

Project description:BackgroundInvasive breast cancer patients are increasingly being treated with neoadjuvant chemotherapy; however, only a fraction of the patients respond to it completely. To prevent overtreatment, there is an urgent need for biomarkers to predict treatment response before administering the therapy.MethodsIn this retrospective study, we developed hypothesis-driven interpretable biomarkers based on deep learning, to predict the pathological complete response (pCR, i.e., the absence of tumor cells in the surgical resection specimens) to neoadjuvant chemotherapy solely using digital pathology H&E images of pre-treatment breast biopsies. Our approach consists of two steps: First, we use deep learning to characterize aspects of the tumor micro-environment by detecting mitoses and segmenting tissue into several morphology compartments including tumor, lymphocytes and stroma. Second, we derive computational biomarkers from the segmentation and detection output to encode slide-level relationships of components of the tumor microenvironment, such as tumor and mitoses, stroma, and tumor infiltrating lymphocytes (TILs).ResultsWe developed and evaluated our method on slides from n = 721 patients from three European medical centers with triple-negative and Luminal B breast cancers and performed external independent validation on n = 126 patients from a public dataset. We report the predictive value of the investigated biomarkers for predicting pCR with areas under the receiver operating characteristic curve between 0.66 and 0.88 across the tested cohorts.ConclusionThe proposed computational biomarkers predict pCR, but will require more evaluation and finetuning for clinical application. Our results further corroborate the potential role of deep learning to automate TILs quantification, and their predictive value in breast cancer neoadjuvant treatment planning, along with automated mitoses quantification. We made our method publicly available to extract segmentation-based biomarkers for research purposes.

Project description:Neoadjuvant chemotherapy (NACT) outcomes vary according to breast cancer (BC) subtype. Since pathologic complete response is one of the most important target endpoints of NACT, further investigation of NACT outcomes in BC is crucial. Thus, identifying sensitive and specific predictors of treatment response for each phenotype would enable early detection of chemoresistance and residual disease, decreasing exposures to ineffective therapies and enhancing overall survival rates. We used liquid chromatography-high-resolution mass spectrometry (LC-HRMS)-based untargeted metabolomics to detect molecular changes in plasma of three different BC subtypes following the same NACT regimen, with the aim of searching for potential predictors of response. The metabolomics data set was analyzed by combining univariate and multivariate statistical strategies. By using ANOVA-simultaneous component analysis (ASCA), we were able to determine the prognostic value of potential biomarker candidates of response to NACT in the triple-negative (TN) subtype. Higher concentrations of docosahexaenoic acid and secondary bile acids were found at basal and presurgery samples, respectively, in the responders group. In addition, the glycohyocholic and glycodeoxycholic acids were able to classify TN patients according to response to treatment and overall survival with an area under the curve model > 0.77. In relation to luminal B (LB) and HER2+ subjects, it should be noted that significant differences were related to time and individual factors. Specifically, tryptophan was identified to be decreased over time in HER2+ patients, whereas LysoPE (22:6) appeared to be increased, but could not be associated with response to NACT. Therefore, the combination of untargeted-based metabolomics along with longitudinal statistical approaches may represent a very useful tool for the improvement of treatment and in administering a more personalized BC follow-up in the clinical practice.

Dataset Information

Predicting breast cancer response to neoadjuvant chemotherapy based on tumor vascular features in needle biopsies.

Publications

Predicting breast cancer response to neoadjuvant chemotherapy based on tumor vascular features in needle biopsies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets