Creation and validation of models to predict response to primary treatment in serous ovarian cancer

Abstract

Nearly a third of patients with high-grade serous ovarian cancer (HGSC) do not respond to initial therapy and have an overall poor prognosis. However, there are no validated tools that accurately predict which patients will not respond. Our objective is to create and validate accurate models of prediction for treatment response in HGSC. This is a retrospective case–control study that integrates comprehensive clinical and genomic data from 88 patients with HGSC from a single institution. Responders were those patients with a progression-free survival of at least 6 months after treatment. Only patients with complete clinical information and frozen specimen at surgery were included. Gene, miRNA, exon, and long non-coding RNA (lncRNA) expression, gene copy number, genomic variation, and fusion-gene determination were extracted from RNA-sequencing data. DNA methylation analysis was performed. Initial selection of informative variables was performed with univariate ANOVA with cross-validation. Significant variables (p < 0.05) were included in multivariate lasso regression prediction models. Initial models included only one variable. Variables were then combined to create complex models. Model performance was measured with area under the curve (AUC). Validation of all models was performed using TCGA HGSC database. By integrating clinical and genomic variables, we achieved prediction performances of over 95% in AUC. Most performances in the validation set did not differ from the training set. Models with DNA methylation or lncRNA underperformed in the validation set. Integrating comprehensive clinical and genomic data from patients with HGSC results in accurate and robust prediction models of treatment response.

Integrated Clinical and Genomic Models to Predict Optimal Cytoreduction in High-Grade Serous Ovarian Cancer

“…RNA was then isolated from these tumor specimens. RNA extraction, processing, and sequencing have been described previously [ 28 , 29 ]. In brief, total cellular RNA was extracted from primary tumor tissue using the mirVana (Thermo Fisher, Waltham, MA, USA) RNA purification kit.…”

“…Most genomic data were used as continuous variables, except SNV and fusion genes, which were used as dichotomous variables. To select those variables most informative for the prediction of cytoreduction, we used univariate analysis with ANOVA ( p < 0.05), as implemented by the caret R package and detailed previously [ 28 ]. Significant predictive variables were then used in a multivariate lasso regression prediction model (statistical learning).…”

Artificial intelligence-based image analysis can predict outcome in high-grade serous carcinoma via histology alone

“…Our results provide evidence that a weakly supervised convolutional neural network can discriminate extremes in patient outcome with high specificity. This finding is of both practical and conceptual importance, as there are currently no validated tissue-based prognostic or predictive markers for primary platinum-based treatment in use for HGSC 30 . Surprisingly, and in contrast to many other tumor types, pathologist evaluation of HGSC tumor tissue provides almost no predictive or prognostic information beyond the diagnosis itself.…”

A nuclear polymorphism at the 8q24 region is associated with improved survival time and chemo‑response in high‑grade serous ovarian cancer

“…Gene expression of the biological variables listed above was determined from previous RNA sequencing experiments using the same patients (GEO accession number GSE156699) ( 18 , 19 ).…”