Search

Endometrial cancer (EC) incidence and mortality continues to rise. Molecular profiling of EC promises improvement of risk assessment and treatment selection. However, we still lack robust and accurate models to predict those at risk of failing treatment. The objective of this pilot study is to create models with clinical and genomic data that will discriminate patients with EC at risk of disease recurrence. We performed a pilot, retrospective, case−control study evaluating patients with EC, endometrioid type: 7 with recurrence of disease (cases), and 55 without (controls). RNA was extracted from frozen specimens and sequenced (RNAseq). Genomic features from RNAseq included transcriptome expression, genomic, and structural variation. Feature selection for variable reduction was performed with univariate ANOVA with cross-validation. Selected variables, informative for EC recurrence, were introduced in multivariate lasso regression models. Validation of models was performed in machine-learning platforms (ML) and independent datasets (TCGA). The best performing prediction models (out of >170) contained the same lncRNA features (AUC of 0.9, and 95% CI: 0.75, 1.0). Models were validated with excellent performance in ML platforms and good performance in an independent dataset. Prediction models of EC recurrence containing lncRNA features have better performance than models with clinical data alone.

The preoperative diagnosis of pelvic masses has been elusive to date. Methods for characterization such as CA-125 have had limited specificity. We hypothesize that genomic variation can be used to create prediction models which accurately distinguish high grade serous ovarian cancer (HGSC) from benign tissue., Methods: In this retrospective, pilot study, we extracted DNA and RNA from HGSC specimens and from benign fallopian tubes. Then, we performed whole exome sequencing and RNA sequencing, and identified single nucleotide variants (SNV), copy number variants (CNV) and structural variants (SV). We used these variants to create prediction models to distinguish cancer from benign tissue. The models were then validated in independent datasets and with a machine learning platform., Results: The prediction model with SNV had an AUC of 1.00 (95% CI 1.00-1.00). The models with CNV and SV had AUC of 0.87 and 0.73, respectively. Validated models also had excellent performances., Conclusions: Genomic variation of HGSC can be used to create prediction models which accurately discriminate cancer from benign tissue. Further refining of these models (early-stage samples, other tumor types) has the potential to lead to detection of ovarian cancer in blood with cell free DNA, even in early stage.

The utility of comprehensive surgical staging in patients with low risk disease has been questioned. Thus, a reliable means of determining risk would be quite useful. The aim of our study was to create the best performing prediction model to classify endometrioid endometrial cancer (EEC) patients into low or high risk using a combination of molecular and clinical-pathological variables. We then validated these models with publicly available datasets. Analyses between low and high risk EEC were performed using clinical and pathological data, gene and miRNA expression data, gene copy number variation and somatic mutation data. Variables were selected to be included in the prediction model of risk using cross-validation analysis; prediction models were then constructed using these variables. Model performance was assessed by area under the curve (AUC). Prediction models were validated using appropriate datasets in The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. A prediction model with only clinical variables performed at 88%. Integrating clinical and molecular data improved prediction performance up to 97%. The best prediction models included clinical, miRNA expression and/or somatic mutation data, and stratified pre-operative risk in EEC patients. Integrating molecular and clinical data improved the performance of prediction models to over 95%, resulting in potentially useful clinical tests.

In the era of large genetic and genomic datasets, it has become crucially important to validate results of individual studies using data from publicly available sources, such as The Cancer Genome Atlas (TCGA). However, how generalizable are results from either an independent or a large public dataset to the remainder of the population? The study presented here aims to answer that question. Utilizing next generation sequencing data from endometrial and ovarian cancer patients from both the University of Iowa and TCGA, genomic admixture of each population was analyzed using STRUCTURE and ADMIXTURE software. In our independent data set, one subpopulation was identified, whereas in TCGA 4⁻6 subpopulations were identified. Data presented here demonstrate how different the genetic substructures of the TCGA and University of Iowa populations are. Validation of genomic studies between two different population samples must be aware of, account for and be corrected for background genetic substructure.

Nearly one-third of patients with high-grade serous ovarian cancer (HGSC) do not respond to initial treatment with platinum-based therapy. Genomic and clinical characterization of these patients may lead to potential alternative therapies. Here, the objective is to classify non-responders into subsets using clinical and molecular features. Using patients from The Cancer Genome Atlas (TCGA) dataset with platinum-resistant or platinum-refractory HGSC, we performed a genome-wide unsupervised cluster analysis that integrated clinical data, gene copy number variations, gene somatic mutations, and DNA promoter methylation. Pathway enrichment analysis was performed for each cluster to identify the targetable processes. Following the unsupervised cluster analysis, three distinct clusters of non-responders emerged. Cluster 1 had overrepresentation of the stage IV disease and suboptimal debulking, under-expression of miRNAs and mRNAs, hypomethylated DNA, "loss of function" TP53 mutations, and the overexpression of genes in the PDGFR pathway. Cluster 2 had low miRNA expression, generalized hypermethylation, MUC17 mutations, and significant activation of the HIF-1 signaling pathway. Cluster 3 had more optimally cytoreduced stage III patients, overexpression of miRNAs, mixed methylation patterns, and "gain of function" TP53 mutations. However, the survival for all clusters was similar. Integration of genomic and clinical data from patients that do not respond to chemotherapy has identified different subgroups or clusters. Pathway analysis further identified the potential alternative therapeutic targets for each cluster.

Ovarian cancer is the most lethal gynecological malignancy, with an alarmingly poor prognosis attributed to late detection and chemoresistance. Initially, most tumors respond to chemotherapy but eventually relapse due to the development of drug resistance. Currently, there are no biological markers that can be used to predict patient response to chemotherapy. However, it is clear that mutations in the tumor suppressor gene TP53, which occur in 96% of serous ovarian tumors, alter the core molecular pathways involved in drug response. One subtype of TP53 mutations, widely termed gain-of-function (GOF) mutations, surprisingly converts this protein from a tumor suppressor to an oncogene. We term the resulting change an oncomorphism. In this review, we discuss particular TP53 mutations, including known oncomorphic properties of the resulting mutant p53 proteins. For example, several different oncomorphic mutations have been reported, but each mutation acts in a distinct manner and has a different effect on tumor progression and chemoresistance. An understanding of the pathological pathways altered by each mutation is necessary in order to design appropriate drug interventions for patients suffering from this deadly disease.