Integrated analysis of gene expression and tumor nuclear image profiles associated with chemotherapy response in serous ovarian carcinoma

Prediction of Ovarian Cancer Response to Therapy Based on Deep Learning Analysis of Histopathology Images

BACKGROUND

Ovarian cancer remains the leading gynecological cause of cancer mortality. Predicting the sensitivity of ovarian cancer to chemotherapy at the time of pathological diagnosis is a goal of precision medicine research that we have addressed in this study using a novel deep-learning neural network framework to analyze the histopathological images.

METHODS

We have developed a method based on the Inception V3 deep learning algorithm that complements other methods for predicting response to standard platinum-based therapy of the disease. For the study, we used histopathological H&E images (pre-treatment) of high-grade serous carcinoma from The Cancer Genome Atlas (TCGA) Genomic Data Commons portal to train the Inception V3 convolutional neural network system to predict whether cancers had independently been labeled as sensitive or resistant to subsequent platinum-based chemotherapy. The trained model was then tested using data from patients left out of the training process. We used receiver operating characteristic (ROC) and confusion matrix analyses to evaluate model performance and Kaplan-Meier survival analysis to correlate the predicted probability of resistance with patient outcome. Finally, occlusion sensitivity analysis was piloted as a start toward correlating histopathological features with a response.

RESULTS

The study dataset consisted of 248 patients with stage 2 to 4 serous ovarian cancer. For a held-out test set of forty patients, the trained deep learning network model distinguished sensitive from resistant cancers with an area under the curve (AUC) of 0.846 ± 0.009 (SE). The probability of resistance calculated from the deep-learning network was also significantly correlated with patient survival and progression-free survival. In confusion matrix analysis, the network classifier achieved an overall predictive accuracy of 85% with a sensitivity of 73% and specificity of 90% for this cohort based on the Youden-J cut-off. Stage, grade, and patient age were not statistically significant for this cohort size. Occlusion sensitivity analysis suggested histopathological features learned by the network that may be associated with sensitivity or resistance to the chemotherapy, but multiple marker studies will be necessary to follow up on those preliminary results.

CONCLUSIONS

This type of analysis has the potential, if further developed, to improve the prediction of response to therapy of high-grade serous ovarian cancer and perhaps be useful as a factor in deciding between platinum-based and other therapies. More broadly, it may increase our understanding of the histopathological variables that predict response and may be adaptable to other cancer types and imaging modalities.

Statistical analysis of high-dimensional biomedical data: a gentle introduction to analytical goals, common approaches and challenges

BACKGROUND

In high-dimensional data (HDD) settings, the number of variables associated with each observation is very large. Prominent examples of HDD in biomedical research include omics data with a large number of variables such as many measurements across the genome, proteome, or metabolome, as well as electronic health records data that have large numbers of variables recorded for each patient. The statistical analysis of such data requires knowledge and experience, sometimes of complex methods adapted to the respective research questions.

METHODS

Advances in statistical methodology and machine learning methods offer new opportunities for innovative analyses of HDD, but at the same time require a deeper understanding of some fundamental statistical concepts. Topic group TG9 "High-dimensional data" of the STRATOS (STRengthening Analytical Thinking for Observational Studies) initiative provides guidance for the analysis of observational studies, addressing particular statistical challenges and opportunities for the analysis of studies involving HDD. In this overview, we discuss key aspects of HDD analysis to provide a gentle introduction for non-statisticians and for classically trained statisticians with little experience specific to HDD.

RESULTS

The paper is organized with respect to subtopics that are most relevant for the analysis of HDD, in particular initial data analysis, exploratory data analysis, multiple testing, and prediction. For each subtopic, main analytical goals in HDD settings are outlined. For each of these goals, basic explanations for some commonly used analysis methods are provided. Situations are identified where traditional statistical methods cannot, or should not, be used in the HDD setting, or where adequate analytic tools are still lacking. Many key references are provided.

CONCLUSIONS

This review aims to provide a solid statistical foundation for researchers, including statisticians and non-statisticians, who are new to research with HDD or simply want to better evaluate and understand the results of HDD analyses.

Deep Learning Approaches in Histopathology

The revolution of artificial intelligence and its impacts on our daily life has led to tremendous interest in the field and its related subtypes: machine learning and deep learning. Scientists and developers have designed machine learning- and deep learning-based algorithms to perform various tasks related to tumor pathologies, such as tumor detection, classification, grading with variant stages, diagnostic forecasting, recognition of pathological attributes, pathogenesis, and genomic mutations. Pathologists are interested in artificial intelligence to improve the diagnosis precision impartiality and to minimize the workload combined with the time consumed, which affects the accuracy of the decision taken. Regrettably, there are already certain obstacles to overcome connected to artificial intelligence deployments, such as the applicability and validation of algorithms and computational technologies, in addition to the ability to train pathologists and doctors to use these machines and their willingness to accept the results. This review paper provides a survey of how machine learning and deep learning methods could be implemented into health care providers' routine tasks and the obstacles and opportunities for artificial intelligence application in tumor morphology.

Polygenic and Network-Based Studies in Risk Identification and Demystification of cancer

INTRODUCTION

Diseases were initially thought to be the consequence of a single gene mutation. Advances in DNA sequencing tools and our understanding of gene behavior have revealed that complex diseases, such as cancer, are the product of genes cooperating with each other and with their environment in orchestrated communication networks. Seeing that the function of individual genes is still used to analyze cancer, the shift to using functionally interacting groups of genes as a new unit of study holds promise for demystifying cancer.

AREAS COVERED

The literature search focused on three types of cancer, namely breast, lung, and prostate, but arguments from other cancers were also included. The aim was to prove that multigene analyses can accurately predict and prognosticate cancer risk, subtype cancer for more personalized and effective treatments, and discover anti-cancer therapies. Computational intelligence is being harnessed to analyze this type of data and is proving indispensable to scientific progress.

EXPERT OPINION

In the future, comprehensive profiling of all kinds of patient data (e.g., serum molecules, environmental exposures) can be used to build universal networks that should help us elucidate the molecular mechanisms underlying diseases and provide appropriate preventive measures, ensuring lifelong health and longevity.

Emerging role of deep learning-based artificial intelligence in tumor pathology

The development of digital pathology and progression of state-of-the-art algorithms for computer vision have led to increasing interest in the use of artificial intelligence (AI), especially deep learning (DL)-based AI, in tumor pathology. The DL-based algorithms have been developed to conduct all kinds of work involved in tumor pathology, including tumor diagnosis, subtyping, grading, staging, and prognostic prediction, as well as the identification of pathological features, biomarkers and genetic changes. The applications of AI in pathology not only contribute to improve diagnostic accuracy and objectivity but also reduce the workload of pathologists and subsequently enable them to spend additional time on high-level decision-making tasks. In addition, AI is useful for pathologists to meet the requirements of precision oncology. However, there are still some challenges relating to the implementation of AI, including the issues of algorithm validation and interpretability, computing systems, the unbelieving attitude of pathologists, clinicians and patients, as well as regulators and reimbursements. Herein, we present an overview on how AI-based approaches could be integrated into the workflow of pathologists and discuss the challenges and perspectives of the implementation of AI in tumor pathology.

Clinical significance of FBXO17 gene expression in high-grade glioma

BACKGROUND

High-grade gliomas (HGGs) exhibit marked heterogeneity in clinical behavior. The purpose of this study was to identify a novel biomarker that predicts patient outcome, which is helpful in HGG patient management.

METHODS

We analyzed gene expression profiles of 833 HGG cases, representing the largest patient population ever reported. Using the data set from the Cancer Genome Atlas (TCGA) and random partitioning approach, we performed Cox proportional hazards model analysis to identify novel prognostic mRNAs in HGG. The predictive capability was further assessed via multivariate analysis and validated in 4 additional data sets. The Kaplan-Meier method was used to evaluate survival difference between dichotomic groups of patients. Correlation of gene expression and DNA methylation was evaluated via Student's t-test.

RESULTS

Patients with elevated FBXO17 expression had a significantly shorter overall survival (OS) (P = 0.0011). After adjustment by IDH1 mutation, sex, and patient age, FBXO17 gene expression was significantly associated with OS (HR = 1.29, 95% CI =1.04-1.59, P = 0.018). In addition, FBXO17 expression can significantly distinguish patients by OS not only among patients who received temozolomide chemotherapy (HR 1.35, 95% CI =1.12-1.64, P = 0.002) but also among those who did not (HR = 1.48, 95% CI =1.20-1.82, P < 0.0001). The significant association of FBXO17 gene expression with OS was further validated in four external data sets. We further found that FBXO17 endogenous expression is significantly contributable from its promoter methylation.

CONCLUSION

Epigenetically modulated FBXO17 has a potential as a stratification factor for clinical decision-making in HGG.

Predicting high-risk endometrioid carcinomas using proteins

Background

The lethality of endometrioid endometrial cancer (EEC) is primarily attributable to advanced-stage diseases. We sought to develop a biomarker model that predicts EEC surgical stage at the time of clinical diagnosis.

Results

PSES was significantly correlated with surgical stage in the TCGA cohort (P < 0.0001) and in the validation cohort (P = 0.0003). Even among grade 1 or 2 tumors, PSES was significantly higher in advanced than in early stage tumors in both the TCGA (P = 0.005) and MD Anderson Cancer Center (MDACC) (P = 0.006) cohorts. Patients with positive PSES score had significantly shorter progression-free survival than those with negative PSES in the TCGA (hazard ratio [HR], 2.033; 95% CI, 1.031 to 3.809; P = 0.04) and validation (HR, 3.306; 95% CI, 1.836 to 9.436; P = 0.0007) cohorts. The ErbB signaling pathway was most significantly enriched in the PSES proteins and downregulated in advanced stage tumors.

Methods

Using reverse-phase protein array expression profiles of 170 antibodies for 210 EEC cases from TCGA, we constructed a Protein Scoring of EEC Staging (PSES) scheme comprising 6 proteins (3 of them phosphorylated) for surgical stage prediction. We validated and evaluated its diagnostic potential in an independent cohort of 184 EEC cases obtained at MDACC using receiver operating characteristic curve analyses. Kaplan-Meier survival analysis was used to examine the association of PSES score with patient outcome, and Ingenuity pathway analysis was used to identify relevant signaling pathways. Two-sided statistical tests were used.

Conclusions

PSES may provide clinically useful prediction of high-risk tumors and offer new insights into tumor biology in EEC.

Quantitative histology analysis of the ovarian tumour microenvironment

Concerted efforts in genomic studies examining RNA transcription and DNA methylation patterns have revealed profound insights in prognostic ovarian cancer subtypes. On the other hand, abundant histology slides have been generated to date, yet their uses remain very limited and largely qualitative. Our goal is to develop automated histology analysis as an alternative subtyping technology for ovarian cancer that is cost-efficient and does not rely on DNA quality. We developed an automated system for scoring primary tumour sections of 91 late-stage ovarian cancer to identify single cells. We demonstrated high accuracy of our system based on expert pathologists' scores (cancer = 97.1%, stromal = 89.1%) as well as compared to immunohistochemistry scoring (correlation = 0.87). The percentage of stromal cells in all cells is significantly associated with poor overall survival after controlling for clinical parameters including debulking status and age (multivariate analysis p = 0.0021, HR = 2.54, CI = 1.40-4.60) and progression-free survival (multivariate analysis p = 0.022, HR = 1.75, CI = 1.09-2.82). We demonstrate how automated image analysis enables objective quantification of microenvironmental composition of ovarian tumours. Our analysis reveals a strong effect of the tumour microenvironment on ovarian cancer progression and highlights the potential of therapeutic interventions that target the stromal compartment or cancer-stroma signalling in the stroma-high, late-stage ovarian cancer subset.

Prediction of resistance to chemotherapy in ovarian cancer: a systematic review

Background

Patient response to chemotherapy for ovarian cancer is extremely heterogeneous and there are currently no tools to aid the prediction of sensitivity or resistance to chemotherapy and allow treatment stratification. Such a tool could greatly improve patient survival by identifying the most appropriate treatment on a patient-specific basis.

Methods

PubMed was searched for studies predicting response or resistance to chemotherapy using gene expression measurements of human tissue in ovarian cancer.

Results

42 studies were identified and both the data collection and modelling methods were compared. The majority of studies utilised fresh-frozen or formalin-fixed paraffin-embedded tissue. Modelling techniques varied, the most popular being Cox proportional hazards regression and hierarchical clustering which were used by 17 and 11 studies respectively. The gene signatures identified by the various studies were not consistent, with very few genes being identified by more than two studies. Patient cohorts were often noted to be heterogeneous with respect to chemotherapy treatment undergone by patients.

Conclusions

A clinically applicable gene signature capable of predicting patient response to chemotherapy has not yet been identified. Research into a predictive, as opposed to prognostic, model could be highly beneficial and aid the identification of the most suitable treatment for patients.

Electronic supplementary material

The online version of this article (doi:10.1186/s12885-015-1101-8) contains supplementary material, which is available to authorized users.

Somatic mutations favorable to patient survival are predominant in ovarian carcinomas

Somatic mutation accumulation is a major cause of abnormal cell growth. However, some mutations in cancer cells may be deleterious to the survival and proliferation of the cancer cells, thus offering a protective effect to the patients. We investigated this hypothesis via a unique analysis of the clinical and somatic mutation datasets of ovarian carcinomas published by the Cancer Genome Atlas. We defined and screened 562 macro mutation signatures (MMSs) for their associations with the overall survival of 320 ovarian cancer patients. Each MMS measures the number of mutations present on the member genes (except for TP53) covered by a specific Gene Ontology (GO) term in each tumor. We found that somatic mutations favorable to the patient survival are predominant in ovarian carcinomas compared to those indicating poor clinical outcomes. Specially, we identified 19 (3) predictive MMSs that are, usually by a nonlinear dose-dependent effect, associated with good (poor) patient survival. The false discovery rate for the 19 "positive" predictors is at the level of 0.15. The GO terms corresponding to these MMSs include "lysosomal membrane" and "response to hypoxia", each of which is relevant to the progression and therapy of cancer. Using these MMSs as features, we established a classification tree model which can effectively partition the training samples into three prognosis groups regarding the survival time. We validated this model on an independent dataset of the same disease (Log-rank p-value < 2.3 × 10(-4)) and a dataset of breast cancer (Log-rank p-value < 9.3 × 10(-3)). We compared the GO terms corresponding to these MMSs and those enriched with expression-based predictive genes. The analysis showed that the GO term pairs with large similarity are mainly pertinent to the proteins located on the cell organelles responsible for material transport and waste disposal, suggesting the crucial role of these proteins in cancer mortality.

MicroRNAs as prognostic markers in ovarian cancer

Ovarian cancer (OC) is the most lethal gynecological malignancy among women. Over 70% of women with OC are diagnosed in advanced stages and most of these cases are incurable. Although most patients respond well to primary chemotherapy, tumors become resistant to treatment. Mechanisms of chemoresistance in cancer cells may be associated with mutational events and/or alterations of gene expression through epigenetic events. Although focusing on known genes has already yielded new information, previously unknown non-coding RNAs, such as microRNAs (miRNAs), also lead insight into the biology of chemoresistance. In this review we summarize the current evidence examining the role of miRNAs as biomarkers of response and survival to therapy in OC. Beside their clinical implications, we also discuss important differences between studies that may have limited their use as clinical biomarkers and suggest new approaches.

Network-based stratification of tumor mutations

Many forms of cancer have multiple subtypes with different causes and clinical outcomes. Somatic tumor genome sequences provide a rich new source of data for uncovering these subtypes but have proven difficult to compare, as two tumors rarely share the same mutations. Here we introduce network-based stratification (NBS), a method to integrate somatic tumor genomes with gene networks. This approach allows for stratification of cancer into informative subtypes by clustering together patients with mutations in similar network regions. We demonstrate NBS in ovarian, uterine and lung cancer cohorts from The Cancer Genome Atlas. For each tissue, NBS identifies subtypes that are predictive of clinical outcomes such as patient survival, response to therapy or tumor histology. We identify network regions characteristic of each subtype and show how mutation-derived subtypes can be used to train an mRNA expression signature, which provides similar information in the absence of DNA sequence.

Whole transcriptome sequencing enables discovery and analysis of viruses in archived primary central nervous system lymphomas

Primary central nervous system lymphomas (PCNSL) have a dramatically increased prevalence among persons living with AIDS and are known to be associated with human Epstein Barr virus (EBV) infection. Previous work suggests that in some cases, co-infection with other viruses may be important for PCNSL pathogenesis. Viral transcription in tumor samples can be measured using next generation transcriptome sequencing. We demonstrate the ability of transcriptome sequencing to identify viruses, characterize viral expression, and identify viral variants by sequencing four archived AIDS-related PCNSL tissue samples and analyzing raw sequencing reads. EBV was detected in all four PCNSL samples and cytomegalovirus (CMV), JC polyomavirus (JCV), and HIV were also discovered, consistent with clinical diagnoses. CMV was found to express three long non-coding RNAs recently reported as expressed during active infection. Single nucleotide variants were observed in each of the viruses observed and three indels were found in CMV. No viruses were found in several control tumor types including 32 diffuse large B-cell lymphoma samples. This study demonstrates the ability of next generation transcriptome sequencing to accurately identify viruses, including DNA viruses, in solid human cancer tissue samples.

Clinical cancer advances 2012: annual report on progress against cancer from the american society of clinical oncology

A MESSAGE FROM ASCO'S PRESIDENT

I am delighted to present you with “Clinical Cancer Advances 2012: Annual Report on Progress Against Cancer From the American Society of Clinical Oncology.” The American Society of Clinical Oncology (ASCO) uses this opportunity each year to share the steady progress occurring in our understanding and treatment of cancer. For 2012, we offer again an inspiring perspective on clinical cancer advances over the past year, but with a cautionary note: if current threats to federal funding materialize, future progress in cancer research will be seriously undermined.

Continued progress against cancer. As you read the following pages of this report, I hope you will share my unabashed enthusiasm—and pride—in how far we have come. To appreciate what this progress has meant to the millions of people who receive a cancer diagnosis each year, consider the following: (1) two of three people in the United States live at least 5 years after a cancer diagnosis (up from roughly one of two in the 1970s); (2) the nation's cancer death rate has dropped 18% since the early 1990s, reversing decades of increases; and (3) individuals with cancer are increasingly able to live active, fulfilling lives because of better management of symptoms and treatments with fewer adverse effects.

Importance of clinical cancer trials. These dramatic trends—and the advances highlighted in this report—would have been unthinkable without the engine that drives life-saving cancer treatment: clinical cancer research. Advances in technology and in our knowledge of how patient-specific molecular characteristics of the tumor and its environment fuel the growth of cancer have brought new hope to patients. Clinical trials are the key to translating cutting-edge laboratory discoveries into treatments that extend and improve the lives of those with cancer.

But progress is only part of the story. Cancer remains a challenge, with many cancers undetected until their latest stages and others resisting most attempts at treatment. Tragically, cancer still kills more than 500,000 people in the United States every year, and its global burden is growing rapidly.

Bridges to better care. To conquer cancer, we need to build bridges to the future—bridges that will get scientific advances to the patient's bedside quicker, bridges that will enable us to share information and learn what works in real time, and bridges that will improve care for all patients around the world.

At ASCO, we recognize the unique role that oncologists must play. ASCO's “Accelerating Progress Against Cancer: Blueprint for Transforming Clinical and Translational Cancer Research,”1published last year, presents our vision and recommendations to make cancer research and patient care vastly more targeted, more efficient, and more effective. We have also launched a groundbreaking initiative, CancerLinQ, that aims to improve cancer care and speed research by drawing insights from the vast pool of data on patients in real-world settings.

Renewing a national commitment to cancer research. We are on the threshold of major advances in cancer prevention, detection, and treatment—but only if, as a nation, we remain committed to this critical endeavor.

The federally funded cancer research system is currently under threat by larger federal budget concerns. Clearly, Congress faces a complex budget environment, but now is not the time to retreat from our nation's commitment to conquering a disease that affects nearly all of us. Bold action must be taken to ensure that we can take full advantage of today's scientific and technologic opportunities.

Please join me in celebrating our nation's progress against cancer and in recommitting ourselves to supporting cancer research. Millions of lives depend on it.

Sandra M. Swain, MD

President

American Society of Clinical Oncology

Biological Interpretation of Morphological Patterns in Histopathological Whole-Slide Images

We propose a framework for studying visual morphological patterns across histopathological whole-slide images (WSIs). Image representation is an important component of computer-aided decision support systems for histopathological cancer diagnosis. Such systems extract hundreds of quantitative image features from digitized tissue biopsy slides and produce models for prediction. The performance of these models depends on the identification of informative features for selection of appropriate regions-of-interest (ROIs) from heterogeneous WSIs and for development of models. However, identification of informative features is hindered by the semantic gap between human interpretation of visual morphological patterns and quantitative image features. We address this challenge by using data mining and information visualization tools to study spatial patterns formed by features extracted from sub-sections of WSIs. Using ovarian serous cystadenocarcinoma (OvCa) WSIs provided by the cancer genome atlas (TCGA), we show that (1) individual and (2) multivariate image features correspond to biologically relevant ROIs, and (3) supervised image feature selection can map histopathology domain knowledge to quantitative image features.