Purpose:

Gene expression–based molecular subtypes of high-grade serous tubo-ovarian cancer (HGSOC), demonstrated across multiple studies, may provide improved stratification for molecularly targeted trials. However, evaluation of clinical utility has been hindered by nonstandardized methods, which are not applicable in a clinical setting. We sought to generate a clinical grade minimal gene set assay for classification of individual tumor specimens into HGSOC subtypes and confirm previously reported subtype-associated features.

Experimental Design:

Adopting two independent approaches, we derived and internally validated algorithms for subtype prediction using published gene expression data from 1,650 tumors. We applied resulting models to NanoString data on 3,829 HGSOCs from the Ovarian Tumor Tissue Analysis consortium. We further developed, confirmed, and validated a reduced, minimal gene set predictor, with methods suitable for a single-patient setting.

Results:

Gene expression data were used to derive the predictor of high-grade serous ovarian carcinoma molecular subtype (PrOTYPE) assay. We established a de facto standard as a consensus of two parallel approaches. PrOTYPE subtypes are significantly associated with age, stage, residual disease, tumor-infiltrating lymphocytes, and outcome. The locked-down clinical grade PrOTYPE test includes a model with 55 genes that predicted gene expression subtype with >95% accuracy that was maintained in all analytic and biological validations.

Conclusions:

We validated the PrOTYPE assay following the Institute of Medicine guidelines for the development of omics-based tests. This fully defined and locked-down clinical grade assay will enable trial design with molecular subtype stratification and allow for objective assessment of the predictive value of HGSOC molecular subtypes in precision medicine applications.

See related commentary by McMullen et al., p. 5271

Translational Relevance

Outcomes for women diagnosed with high-grade serous tubo-ovarian carcinoma (HGSOC) have limited improvements over the last few decades. While novel targeted therapeutic strategies are maturing, their widespread adoption is often dependent on biomarkers that can guide management and identify women who are more likely to benefit from new compounds. For HGSOC, several previously described, near-identical gene expression–based subclassification schemes have had little impact on practice or clinical trial design. The most prominent drawback to their implementation is that they have not been designed in a clinically applicable way. Without a de facto standard, any potential clinical utility of HGSOC gene expression subtypes cannot be determined. Here, we developed and validated a standardized and reproducible HGSOC gene expression subtype classifier that will enable prospective assessment of the clinical utility of HGSOC gene expression subtypes. The predictor of high-grade serous ovarian carcinoma molecular subtype (PrOTYPE) represents an Institute of Medicine guidelines–compliant, fully defined, and validated assay that can be used with formalin-fixed, paraffin-embedded tissues, making it practical for clinical uptake. Our report confirms the biological relevance of gene expression subtypes in HGSOC and will facilitate the incorporation of subtype classification into ongoing and future clinical trials.

Anatomy and histopathology have been the foundations of cancer classification for more than a century, but both are now complemented by objective assessment of underlying molecular features of the disease (1–8). The development of microarray-based gene expression profiling of high-grade serous tubo-ovarian carcinoma (HGSOC; refs. 9, 10) raised expectations for rapid advances in classification, prognostication, and prediction in this most common histotype (∼70%) of ovarian carcinoma, the deadliest gynecologic malignancy (11, 12).

Previous studies identified four phenotypically distinct expression-based HGSOC molecular subtypes (9, 10, 13–18). These subtypes have been repeatedly reproduced, with broad similarities in composite pathologic characteristics. The C1/mesenchymal (C1.MES) subtype is characterized by a desmoplastic stroma, high expression of extracellular matrix components, and poor outcomes compared with other HGSOCs; which is consistent with other solid tumors with highly desmoplastic stroma (19–23). The C2/immunoreactive (C2.IMM) subtype is dominated by intratumoral CD3+/CD8+ cellular infiltration, inflammatory cytokine expression, and generally more favorable outcomes. The C4/differentiated (C4.DIF) subtype is characterized by high expression of CA125/MUC16, a subset of immunomodulatory cytokines, modest lymphocyte infiltration, and clinical outcome indistinguishable from C2.IMM (10, 15, 17, 24). Finally, the C5/proliferative (C5.PRO) are depleted for both stromal and immune elements, overexpress oncofetal and stem cell–associated genes (24), and have unfavorable outcomes (13–15, 17, 18).

Unlike modern histotype classification of ovarian carcinoma (12, 25), no agreed-on gold standard exists to define expression-based HGSOC molecular subtypes. Both analytic methods and data used for subtype assignment are fragmented, differing in algorithms and specific genes used, and each defining its own brand of subtype. No methods discussed to date provide a workflow with compatibility for fixed/archival tissues that are the mainstay of modern pathology laboratories. Thus, the potential of gene expression subtype information to guide patient management remains unrealized (12, 26).

Our motivation for this project was driven by limitations of previous attempts, which contributed to low uptake of HGSOC subtyping in translational research and clinical trials. To optimize clinical uptake, a classification scheme needs to be cost-effective, compatible with available clinical specimens (ie, formalin-fixed, paraffin-embedded; FFPE), and be technically reproducible on single-patient samples. Prior methods have relied on normalization and unsupervised clustering of array-based data, requiring a cohort of samples (9, 10, 13–15, 17, 18, 24, 27). With few exceptions (18), prior studies defaulted to a single method or single dataset to train models. Finally, no prior approach reviewed histotype based on the current diagnostic standards for HGSOC, which has significantly altered over the last decades, and may have contributed to significant contamination of historic datasets with non-HGSOC specimens (28–30).

Using newly curated, previously published array data, and clinically annotated HGSOC specimens from the Ovarian Tumor Tissue Analysis (OTTA) consortium, we propose and validate (26) a predictor of high-grade serous ovarian carcinoma molecular subtype (PrOTYPE) that recapitulates previously derived gene expression–based molecular subtypes using a minimal set of genes (Fig. 1). To ensure clinical applicability, we adopted the NanoString platform, a highly automated processing method with tolerance to degraded RNA, typical of fixed tissue that are the mainstay of modern hospital pathology laboratories. Similar multigene predictors using NanoString are already in the clinic (31–34), and methods to enable single-sample analytic approaches are well established (35), tailored to the patient-at-a-time delivery of care that is a necessity for precision medicine. The PrOTYPE assay will enable the evaluation of the clinical utility of HGSOC gene expression molecular subtypes, such as response to targeted therapies that are already emerging with a potential need for subtype information (36).

Figure 1.

A schematic representation of the process we followed to obtain a final, clinical grade classifier for gene expression molecular subtypes of HGSOC. Please note that the schematics above are for orientation only and are not intended to be interpreted. In the first panel, we outline how de facto subtype labels were assigned to NanoString data, starting with two parallel approaches to build models from array data (A), and applying the resulting final models onto the NanoString dataset (B), where the consensus of the two methods became the de facto gold standard with 79% (n = 3,030) of our total NanoString cohort having agreement, that is, a consensus label (CL). C, The framework used to derive a minimal gene set classifier using the CL NanoString data after removing samples that overlapped both the NanoString and array datasets (overlap n = 76). D, A synopsis of the biological and clinical correlates that were investigated to confirm the biological validity of gene expression–based subtypes compared with previous work.

Figure 1.

A schematic representation of the process we followed to obtain a final, clinical grade classifier for gene expression molecular subtypes of HGSOC. Please note that the schematics above are for orientation only and are not intended to be interpreted. In the first panel, we outline how de facto subtype labels were assigned to NanoString data, starting with two parallel approaches to build models from array data (A), and applying the resulting final models onto the NanoString dataset (B), where the consensus of the two methods became the de facto gold standard with 79% (n = 3,030) of our total NanoString cohort having agreement, that is, a consensus label (CL). C, The framework used to derive a minimal gene set classifier using the CL NanoString data after removing samples that overlapped both the NanoString and array datasets (overlap n = 76). D, A synopsis of the biological and clinical correlates that were investigated to confirm the biological validity of gene expression–based subtypes compared with previous work.

Close modal

OTTA consortium NanoString study

We retrospectively analyzed FFPE tumors and clinical data from 20 OTTA consortium studies with available clinical, pathologic, demographic features, and survival outcomes (Supplementary Data SA.1–SA.3). Inclusion criteria (including approval through institution-specific research ethics boards), individual study settings, dates of accrual, and follow-up are described in Supplementary Table SA1. Studies were asked to contribute adnexal-sourced specimens, although others were accepted when anatomic sites was defined. Expert gynecologic pathologists reviewed samples from hematoxylin and eosin–stained sections, confirmed HGSOC diagnosis (29), and marked specimens for removal of noninvolved organ tissues but retained infiltrating stroma.

NanoString gene selection and data processing

A NanoString CodeSet included 513 genes (plus five housekeeping genes), relevant for gene expression subtyping and selected prior to beginning the analysis. We included top-ranking differentially expressed, subtype-specific genes based on prior reports (9, 10); previous supervised learning of subtype classification (37); and manual review of literature to identify genes in commonly cited molecular pathways associated with subtype (9, 10, 13, 15, 24). Additional genes were selected from a meta-analysis for their prognostic value and other specific hypothesis (38). To ensure representation from across the transcriptome, we tagged and included additional genes from 99% correlated gene expression clusters derived from previous reports, if clusters did not already have representation (9, 10, 39, 40).

We extracted RNA and ran NanoString assays at three sites (in Vancouver, Los Angeles, and Melbourne), as described previously (35). We included three regularly assayed RNA reference specimens (Pool1, Pool2, and Pool3) to monitor technical bias, allow for comparison of NanoString CodeSet synthesized in different lots, and integrate a single-patient data normalization strategy (35). Additional description is in Supplementary Data SA.4–SA.7; data can be found in NCBI Gene Expression Omnibus (GEO) accession GSE135820.

Subtype labels assignment to NanoString data

There is currently no definitive standard for gene expression–based subtypes, therefore, we derived a de facto standard through application of two parallel approaches, led by independent teams (Fig. 1A and B). One approach, denoted all array, aggregated gene expression datasets to take advantage of broad sample representation and increased statistical power. The other, denoted The Cancer Genome Atlas (TCGA), was conservative with respect to potential loss of signal associated with post hoc batch correction and used the largest, optimally batch-corrected dataset (9). See also Supplementary Data SB.

All array

One team curated data to retain only HGSOC specimens from historic datasets (30), and datasets with greater than 40 remaining unique HGSOC. This reduced 49 potential studies (n = 3,437) to 1,650 unique HGSOC from 14 studies (refs. 9, 10, 41–51; Supplementary Table SB1). Individual samples where data were also available from NanoString assays were excluded (Supplementary Fig. SB1). The team combined and batch corrected 11 of 14 array studies (training 1), and used an ensemble of nine clustering algorithms (52) to reestablish previously recognized subtypes. They next restricted the data to preselected NanoString genes also present in all array platforms (454/513 possible NanoString genes), and trained and evaluated nine supervised learning algorithms using a bootstrap approach (53). The top five algorithms were retained and validated on the remaining three (3/14) array studies (confirmation1) with a final selection based on how well the predicted subtypes correlated with previously published signatures (13, 24). The tree-based ensemble classification algorithm (AdaBoost) was selected.

TCGA

Another team curated the TCGA data using the same criteria described above and using data and TCGA-published subtype labels (9), retaining 434 unique HGSOC (Supplementary Fig. SB1). They next trained and evaluated five different supervised learning algorithms, as above, using NanoString gene–restricted data (438/513 genes), using fivefold cross-validation, selecting random forest. This approach was validated externally on originally published dataset and labels from Tothill and colleagues (10).

Minimal gene set classifier

We used the above two approaches to label 3,829 NanoString samples and retained only samples with concordant labels, denoted the consensus labels (CLs). We discarded previous models and started anew to rederive a minimal gene set classifier using NanoString data. Samples were randomly partitioned from the dataset into three independent groups on a per-study basis: a training set (eight studies), a confirmation set (five studies), and a validation set (four studies). A fourth partition/second validation (three studies) comprised of clinical trial cohorts, and was set aside to validate any modifications to the predictive model after confirmation (refs. 26, 54; Fig. 1C; Supplementary Fig. SA1). See also Supplementary Data SC.

We adopted a leave-one-study-out cross-validation approach and assessed performance of three algorithms (LASSO, random forest, and AdaBoost) in recovering the CL. We removed one study at a time and bootstrapped the remaining seven (500 repetitions) to train a full model that uses all the genes to predict subtype. For each bootstrap sample, we ranked the genes on the basis of the aggregated Gini coefficients, for random forest and AdaBoost (55), or the proportion of nonzero coefficients for Lasso. We then ranked genes overall on the proportion of times they were included in the top 100, across bootstrap iterations. This was repeated for each study.

For n increasing from four to 100 in increments of five, we used the top n overall-ranked genes to predict the left-out study, comparing the predicted label with the CL. We selected the top algorithm on the basis of accuracy, consistency, and stability in predictions across studies. We refined gene selection within the confirmation set by considering a smaller range of gene numbers (40–78) and repeating the previous step with one gene increments to define a minimal number of genes needed to sustain performance, and we validated it in two additional datasets.

Biological associations

We confirmed associations of predicted labels with clinical and pathologic features including age, stage, residual disease, cellularity, necrosis, BRCA1/2 germline status, race/ethnicity, and CD8+ tumor-infiltrating lymphocytes (TIL; Supplementary Data SA.3). We used one-way ANOVA to compare continuous variables and the χ2 test for categorical variables. We evaluated univariate survival using Kaplan–Meier survival curves and the log-rank test. In multivariate models, we used the Cox proportional hazard and computed P values using an omnibus likelihood ratio test. All statistical tests were two-sided. We applied pairwise deletion (available-case analysis) on missing data, as applicable.

Subtyping the NanoString data

Parallel array–based approaches resulted in two final models: the all array (AdaBoost) and TCGA (random forest) models (Supplementary Data SB). Each of these algorithms were used to generate per-subtype probabilities and predictive entropy (56) on the 3,829 HGSOC samples run on the NanoString platform. The label of the subtype with the highest probability was taken as the final label from each model. The observed concordance between the two models was high (accuracy 79%; kappa 0.72) and discordance was seen mostly between C1.MES/C2.IMM and C2.IMM/C4.DIF subtypes (Fig. 2A). Discordant samples were enriched for lower signal-to-noise ratio in NanoString data, consistent with lower quality RNA (ratio < 1,000 in 7.5% vs 5%; P = 0.0130; Supplementary Data SB.4). No other technical variables showed differences between concordant and nonconcordant labels. In concordant samples (CL), the predictive entropy was significantly lower (P < 0.0001; Fig. 2B). In a set of 67 cases, repeated on both array and NanoString (and excluded from training), the CL reproduced originally published labels with 94% accuracy (kappa 0.92; refs. 9, 10). Concordant samples (n = 3,030) were considered the de facto standard and subsequently used for training a minimal gene set classifier.

Figure 2.

Evaluation metrics of consensus in subtype assignment between the all array and TCGA models. A, Confusion matrix comparing the agreement between the TCGA and the all array approaches. In bold, we present the results where there is agreement and highlighted in red are the most sizeable disagreements. We also present sensitivity, specificity, and F-score for each subtype. B, Predictive entropy computed from per-class probabilities generated by each of the TCGA and the all array model. When entropy approaches 0, it is indicative that the probability used to assign a sample to class is close to 1, while a high entropy (approaching 2) indicates that assignment to any class has a roughly equal probability. Overall, samples where consensus was not reached, had higher entropy in both models (P < 0.0001; Mann–Whitney U test).

Figure 2.

Evaluation metrics of consensus in subtype assignment between the all array and TCGA models. A, Confusion matrix comparing the agreement between the TCGA and the all array approaches. In bold, we present the results where there is agreement and highlighted in red are the most sizeable disagreements. We also present sensitivity, specificity, and F-score for each subtype. B, Predictive entropy computed from per-class probabilities generated by each of the TCGA and the all array model. When entropy approaches 0, it is indicative that the probability used to assign a sample to class is close to 1, while a high entropy (approaching 2) indicates that assignment to any class has a roughly equal probability. Overall, samples where consensus was not reached, had higher entropy in both models (P < 0.0001; Mann–Whitney U test).

Close modal

Development of a NanoString minimal gene classifier

Using a leave-one-study-out cross-validation design, random forest and LASSO outperformed AdaBoost (Fig. 3A) in the training set (n = 1,135). Despite requiring more genes overall, we chose the random forest model based on stability in gene selection across studies and a less variable overall accuracy with increasing numbers of genes (Fig. 3B). Accuracy of random forest in the confirmation set (n = 817) ranged from 95% to 97% and achieved marginal gains after 55 genes. The locked-down assay, named PrOTYPE, is represented by a final 55-gene model with specified NanoString probe set and controls, specific computational procedures, and requirements for specimen input from primary tubo-ovarian, treatment-naïve HGSOC samples, as outlined in Fig. 4 (see also Supplementary Table SC7; Supplementary Data SE). Computational methods to normalize and generate predictions are available as a web application and R-script (https://ovcare.shinyapps.io/PrOType/).

Figure 3.

Model selection metrics for a minimal gene classifier. A, The aggregate accuracy (left) and F1-score (right; for all samples in all studies) obtained by increasing numbers of genes and using the top n genes from each frequency list computed above, where n varied from four to 100 in increments of five. Note that the top n genes from each study were not necessarily the same. B, Boxplots of the prediction accuracy by study using the LASSO and the random forest algorithm (top). Each point in the boxplot corresponds to the individual study prediction (when left-out). Heatmap depicting the important rank of the top 50 ranking genes obtained from each data partition in the leave-one-study out scheme (bottom).

Figure 3.

Model selection metrics for a minimal gene classifier. A, The aggregate accuracy (left) and F1-score (right; for all samples in all studies) obtained by increasing numbers of genes and using the top n genes from each frequency list computed above, where n varied from four to 100 in increments of five. Note that the top n genes from each study were not necessarily the same. B, Boxplots of the prediction accuracy by study using the LASSO and the random forest algorithm (top). Each point in the boxplot corresponds to the individual study prediction (when left-out). Heatmap depicting the important rank of the top 50 ranking genes obtained from each data partition in the leave-one-study out scheme (bottom).

Close modal
Figure 4.

Locked-down PrOTYPE. The schematic illustrates the four critical components of the clinical grade PrOTYPE assay (Supplementary Data SE) consisting of: (1) 500 ng of total RNA from primary chemo-naïve HGSOC and (2) 100 ng (each) of validated reference specimens. Each of these assayed individually by mixing specimen RNA with a (3) custom NanoString CodeSet (Supplementary Table SC7) containing 55 prediction model gene probes and five control gene probes (*55+5 NanoString assay). RNA is hybridized with CodeSet, processed on a NanoString nCounter Prep-Station, and imaged at maximum fields of view on a NanoString nCounter Digital Analyzer. Resulting raw data are then normalized and HGSOC molecular subtypes are predicted with our PrOTYPE computational algorithms using either a web-based tool or R-script. This process will return (4) a prediction probability for the assayed specimen, for each subtype, and a single predictive entropy value. The latter can be used to estimate the certainty of prediction where 0 entropy corresponds to a near perfect prediction or “pure” subtype, while 2 entropy corresponds to near equal chance of assignment to any subtype.

Figure 4.

Locked-down PrOTYPE. The schematic illustrates the four critical components of the clinical grade PrOTYPE assay (Supplementary Data SE) consisting of: (1) 500 ng of total RNA from primary chemo-naïve HGSOC and (2) 100 ng (each) of validated reference specimens. Each of these assayed individually by mixing specimen RNA with a (3) custom NanoString CodeSet (Supplementary Table SC7) containing 55 prediction model gene probes and five control gene probes (*55+5 NanoString assay). RNA is hybridized with CodeSet, processed on a NanoString nCounter Prep-Station, and imaged at maximum fields of view on a NanoString nCounter Digital Analyzer. Resulting raw data are then normalized and HGSOC molecular subtypes are predicted with our PrOTYPE computational algorithms using either a web-based tool or R-script. This process will return (4) a prediction probability for the assayed specimen, for each subtype, and a single predictive entropy value. The latter can be used to estimate the certainty of prediction where 0 entropy corresponds to a near perfect prediction or “pure” subtype, while 2 entropy corresponds to near equal chance of assignment to any subtype.

Close modal

PrOTYPE genes (55 genes plus five housekeeping genes; Supplementary Table SC7) included representation from pathways previously reported as enriched in HGSOC subtypes (Supplementary Fig. SC9), including components of extracellular matrix (COL11A1, COL1A2, and FBN1), immune cell markers (CD3D, CD3E, and CD8A), surface receptors and kinases (CSF1R, CD2, and AXL), cytokines and cell morphology (CXCL9, CXCL11, and CCL5), and angiogenesis genes (PDGFRB, FGF1, and TCF7L1). The per-subtype pattern of expression of PrOTYPE genes was near-identical between the NanoString data and the array data, used in establishing the CL standard (Supplementary Fig. SC10–SC12).

PrOTYPE was validated in two independent NanoString dataset partitions (n = 719 and 283, respectively; Supplementary Fig. SB1). Partitions showed 95% and 94% accuracy and kappa = 0.94 and 0.92, respectively, relative to the CL (Supplementary Tables SB10 and SB11).

In a set of 103 samples reassayed in a newly synthesized NanoString CodeSet, containing only the 55 PrOTYPE genes and controls, PrOTYPE predictions achieved 97% accuracy (95% confidence interval, 92%–99%) and kappa 0.96 (0.91–1) in recovering the CL. We observed similar results in 100 samples that we replicated in another newly synthesized CodeSet that included PrOTYPE genes as well as others (Supplementary Tables SD2–SD4). Of the 80 samples that overlapped all three CodeSets (original, PrOTYPE genes only, and PrOTYPE genes plus others), Fleiss' kappa was 0.95, indicating excellent repeatability (P < 0.0001). This confirmed the analytic validity of the PrOTYPE assay, our reference-based normalization, and single-sample processing strategy.

Confirmation of subtype signatures with clinicopathologic associations

Patients were diagnosed between 1982 and 2014, with no differences in the distribution of subtypes related to year of diagnosis (Supplementary Fig. SD3). Omental-sourced specimens were enriched for C1.MES (72%) compared with adnexal specimens (25%), and the overall distribution of subtypes was significantly different (P < 0.0001; Table 1A; Supplementary Table SD7). We also noted a similar C1.MES enrichment at other anatomic sites, including the peritoneum (46%) and upper gynecologic tract (50%). In tumors, where anatomic site was presumed adnexal, but not specifically annotated (n = 1,647), subtypes showed a distribution similar to those known to be adnexal (P = 0.089).

Table 1A.

The distribution of HGSOC molecular subtype within different anatomic specimen collection sites.

Sampling site
AdnexalPresumed adnexalOmentumLower genital trackUpper genital trackPeritoneal
C1.MES 429 (24.7%) 394 (23.9%) 256 (72.1%) 0 (0.0%) 8 (50.0%) 32 (45.7%) 
C2.IMM 447 (25.7%) 389 (23.6%) 69 (19.4%) 1 (100.0%) 2 (12.5%) 16 (22.9%) 
C4.DIF 550 (31.6%) 574 (34.9%) 23 (6.5%) 0 (0.0%) 2 (12.5%) 12 (17.1%) 
C5.PRO 314 (18.0%) 290 (17.6%) 7 (2.0%) 0 (0.0%) 4 (25.0%) 10 (14.3%) 
Total 1,740 (45.4%) 1,647 (43.0%) 355 (9.3%) 1 (0.0%) 16 (0.4%) 70 (1.8%) 
Sampling site
AdnexalPresumed adnexalOmentumLower genital trackUpper genital trackPeritoneal
C1.MES 429 (24.7%) 394 (23.9%) 256 (72.1%) 0 (0.0%) 8 (50.0%) 32 (45.7%) 
C2.IMM 447 (25.7%) 389 (23.6%) 69 (19.4%) 1 (100.0%) 2 (12.5%) 16 (22.9%) 
C4.DIF 550 (31.6%) 574 (34.9%) 23 (6.5%) 0 (0.0%) 2 (12.5%) 12 (17.1%) 
C5.PRO 314 (18.0%) 290 (17.6%) 7 (2.0%) 0 (0.0%) 4 (25.0%) 10 (14.3%) 
Total 1,740 (45.4%) 1,647 (43.0%) 355 (9.3%) 1 (0.0%) 16 (0.4%) 70 (1.8%) 

Note: Percentages are column wise, except for totals where they are computed row wise. P values are computed using one-way ANOVA for numerical parameters, and χ2 test for categorical ones.

In 53 patients, where paired adnexal and omentum samples were available, we observed poor agreement (kappa 0.06) in classification from the two sites. For 30 of 39 (77%) adnexal samples that were assigned non-C1.MES subtypes, their corresponding omental sample was C1.MES (Supplementary Table SD5). For all 14 adnexal specimens that were C1.MES, their omental classification was also C1.MES. As previously reported, subtype designation varies between metastatic sites within a patient, therefore, we interpreted this to be a characteristic of tumors within their specific microenvironment rather than a weakness in the classification (37, 57). Heterogeneity in subtype assignment per-patient would confound clinicopathologic associations; therefore, we present associations to subtype of adnexal-sourced specimen as this was the most commonly acquired specimen type (known, n = 1,740; or presumed, n = 1,647; Table 1B; Supplementary Data SD contains results also excluding presumed adnexal samples).

Table 1B.

Clinical and pathologic parameters across HGSOC molecular subtypes from adnexal and presumed adnexal specimens.

C1.MESC2.IMMC4.DIFC5.PROTotalP
N (%) 823 (24.3%) 836 (24.7%) 1,124 (33.2%) 604 (17.8%) 3,387  
Age at diagnosis 
 Mean (SD) 61 (10.4) 60.4 (10.6) 57.8 (10.2) 62.7 (10.3) 60.1 (10.5) <0.0001 
 Median (IQR) 62 (54–68) 60 (53–68) 58 (50–65) 63 (56–70) 60 (53–67)  
 Missing 13 15 22 15 65  
Stage 
 Low 52 (6.5%) 164 (20.7%) 281 (26.4%) 95 (16.4%) 592 (18.3%) <0.0001 
 High 746 (93.5%) 627 (79.3%) 783 (73.6%) 484 (83.6%) 2,640 (81.7%)  
 Missing 25 45 60 25 155  
Residual disease 
 None 100 (27.5%) 143 (39.7%) 210 (43.4%) 116 (41.4%) 569 (38.2%) <0.0001 
 Any 264 (72.5%) 217 (60.3%) 274 (56.6%) 164 (58.6%) 919 (61.8%)  
 Missing 459 476 640 324 1,899  
Cellularity 
 0–20 15 (1.9%) 10 (1.3%) 1 (0.1%) 6 (1.1%) 32 (1.0%) <0.0001 
 21–40 85 (10.7%) 33 (4.1%) 17 (1.6%) 17 (3.0%) 152 (4.8%)  
 41–60 187 (23.6%) 129 (16.2%) 63 (6.1%) 49 (8.6%) 428 (13.4%)  
 61–80 312 (39.3%) 329 (41.3%) 410 (39.7%) 202 (35.6%) 1,253 (39.2%)  
 81–100 195 (24.6%) 296 (37.1%) 543 (52.5%) 294 (51.8%) 1,328 (41.6%)  
 Missing 29 39 90 36 194  
Necrosis 
 None 216 (31.7%) 178 (26.1%) 261 (30.5%) 134 (28.8%) 789 (29.4%) 0.001 
 ≤20% 432 (63.4%) 431 (63.2%) 542 (63.3%) 294 (63.1%) 1,699 (63.3%)  
 >20% 33 (4.8%) 73 (10.7%) 53 (6.2%) 38 (8.2%) 197 (7.3%)  
 Missing 142 154 268 138 702  
BRCA1/BRCA2 
 Wild-type 153 (79.7%) 134 (79.3%) 201 (74.2%) 111 (84.1%) 599 (78.4%) 0.1518 
BRCA1 26 (13.5%) 24 (14.2%) 47 (17.3%) 9 (6.8%) 106 (13.9%)  
BRCA2 13 (6.8%) 11 (6.5%) 23 (8.5%) 12 (9.1%) 59 (7.7%)  
 Missing 631 667 853 472 2,623  
Race 
 White 515 (85.8%) 493 (81.4%) 669 (81.3%) 354 (86.1%) 2031 (83.2%) 0.0523 
 Hispanic 82 (13.7%) 106 (17.5%) 150 (18.2%) 56 (13.6%) 394 (16.1%)  
 Other 3 (0.5%) 7 (1.2%) 4 (0.5%) 1 (0.2%) 15 (0.6%)  
 Missing 223 230 301 193 947  
CD8 
 None 86 (21.2%) 19 (4.8%) 69 (11.6%) 132 (45.2%) 306 (18.1%) <0.0001 
 Low 52 (12.8%) 19 (4.8%) 99 (16.6%) 65 (22.3%) 235 (13.9%)  
 Med 210 (51.9%) 187 (47.1%) 297 (49.7%) 88 (30.1%) 782 (46.2%)  
 High 57 (14.1%) 172 (43.3%) 132 (22.1%) 7 (2.4%) 368 (21.8%)  
 Missing 418 439 527 312 1,696  
Anatomic site 
 Adnexal 429 (52.1%) 447 (53.5%) 550 (48.9%) 314 (52.0%) 1,740 (51.4%) 0.2188 
 Presumed adnexal 394 (47.9%) 389 (46.5%) 574 (51.1%) 290 (48.0%) 1,647 (48.6%)  
C1.MESC2.IMMC4.DIFC5.PROTotalP
N (%) 823 (24.3%) 836 (24.7%) 1,124 (33.2%) 604 (17.8%) 3,387  
Age at diagnosis 
 Mean (SD) 61 (10.4) 60.4 (10.6) 57.8 (10.2) 62.7 (10.3) 60.1 (10.5) <0.0001 
 Median (IQR) 62 (54–68) 60 (53–68) 58 (50–65) 63 (56–70) 60 (53–67)  
 Missing 13 15 22 15 65  
Stage 
 Low 52 (6.5%) 164 (20.7%) 281 (26.4%) 95 (16.4%) 592 (18.3%) <0.0001 
 High 746 (93.5%) 627 (79.3%) 783 (73.6%) 484 (83.6%) 2,640 (81.7%)  
 Missing 25 45 60 25 155  
Residual disease 
 None 100 (27.5%) 143 (39.7%) 210 (43.4%) 116 (41.4%) 569 (38.2%) <0.0001 
 Any 264 (72.5%) 217 (60.3%) 274 (56.6%) 164 (58.6%) 919 (61.8%)  
 Missing 459 476 640 324 1,899  
Cellularity 
 0–20 15 (1.9%) 10 (1.3%) 1 (0.1%) 6 (1.1%) 32 (1.0%) <0.0001 
 21–40 85 (10.7%) 33 (4.1%) 17 (1.6%) 17 (3.0%) 152 (4.8%)  
 41–60 187 (23.6%) 129 (16.2%) 63 (6.1%) 49 (8.6%) 428 (13.4%)  
 61–80 312 (39.3%) 329 (41.3%) 410 (39.7%) 202 (35.6%) 1,253 (39.2%)  
 81–100 195 (24.6%) 296 (37.1%) 543 (52.5%) 294 (51.8%) 1,328 (41.6%)  
 Missing 29 39 90 36 194  
Necrosis 
 None 216 (31.7%) 178 (26.1%) 261 (30.5%) 134 (28.8%) 789 (29.4%) 0.001 
 ≤20% 432 (63.4%) 431 (63.2%) 542 (63.3%) 294 (63.1%) 1,699 (63.3%)  
 >20% 33 (4.8%) 73 (10.7%) 53 (6.2%) 38 (8.2%) 197 (7.3%)  
 Missing 142 154 268 138 702  
BRCA1/BRCA2 
 Wild-type 153 (79.7%) 134 (79.3%) 201 (74.2%) 111 (84.1%) 599 (78.4%) 0.1518 
BRCA1 26 (13.5%) 24 (14.2%) 47 (17.3%) 9 (6.8%) 106 (13.9%)  
BRCA2 13 (6.8%) 11 (6.5%) 23 (8.5%) 12 (9.1%) 59 (7.7%)  
 Missing 631 667 853 472 2,623  
Race 
 White 515 (85.8%) 493 (81.4%) 669 (81.3%) 354 (86.1%) 2031 (83.2%) 0.0523 
 Hispanic 82 (13.7%) 106 (17.5%) 150 (18.2%) 56 (13.6%) 394 (16.1%)  
 Other 3 (0.5%) 7 (1.2%) 4 (0.5%) 1 (0.2%) 15 (0.6%)  
 Missing 223 230 301 193 947  
CD8 
 None 86 (21.2%) 19 (4.8%) 69 (11.6%) 132 (45.2%) 306 (18.1%) <0.0001 
 Low 52 (12.8%) 19 (4.8%) 99 (16.6%) 65 (22.3%) 235 (13.9%)  
 Med 210 (51.9%) 187 (47.1%) 297 (49.7%) 88 (30.1%) 782 (46.2%)  
 High 57 (14.1%) 172 (43.3%) 132 (22.1%) 7 (2.4%) 368 (21.8%)  
 Missing 418 439 527 312 1,696  
Anatomic site 
 Adnexal 429 (52.1%) 447 (53.5%) 550 (48.9%) 314 (52.0%) 1,740 (51.4%) 0.2188 
 Presumed adnexal 394 (47.9%) 389 (46.5%) 574 (51.1%) 290 (48.0%) 1,647 (48.6%)  

Note: Percentages are column wise, except for totals where they are computed row wise. P values are computed using one-way ANOVA for numerical parameters, and χ2 test for categorical ones.

The median age at diagnosis was lowest among C4.DIF (58 years) and highest among C5.PRO (63 years), P < 0.0001. Stage was significantly associated with subtype (P < 0.0001; Table 1B), with 94% of C1.MES at high-stage and only 74% of C4.DIF. Residual disease was significantly associated with subtype, with C1.MES tumors being the most enriched. Similarly, both tumor cellularity and necrosis were associated with subtype. Lowest cellularity was in the C1.MES and highest necrosis was seen in C2.IMM. BRCA1/BRCA2 pathogenic germline mutation status was not associated with subtype. We found CD8+ TIL levels, derived from prior work (58), highest in C2.IMM: 43% with high TIL and only 10% with absent/low CD8+ TIL. C5.PRO had the lowest CD8+ TIL, with 68% having absent/low CD8+ TIL. C4.DIF had the second highest level of CD8+ TIL at 22%.

Median follow-up time was 8.1 years for overall survival (OS) and 6.5 years for progression-free survival (PFS; reverse Kaplan–Meier), and were slightly longer for C2.IMM and C4.DIFF. Significant difference in survival was observed between subtypes for both OS and PFS (log-rank P < 0.0001; Fig. 5A), as reported previously (9, 10, 13–18, 27). C2.IMM and C4.DIF had the best survival outcome and C1.MES had the poorest outcome. In multivariate analyses, we adjusted for risk factors known to be associated with survival: age at diagnosis, stage, residual disease, and germline deleterious BRCA1/2 status. Molecular subtypes were prognostic when adjusting for age and stage in both OS and PFS (Fig. 5B). With the addition of CD8+ TIL, there was a change in the HR corresponding to subtype for both OS and PFS, but subtypes remained independently prognostic for OS only. With the addition of residual disease and/or BRCA1/2 to the model, molecular subtypes lost independent prognostic value in both OS and PFS.

Figure 5.

Univariate and multivariate survival analysis with PrOTYPE subtypes. A, Kaplan–Meir survival curves for OS and PFS by molecular subtype. C2.IMM and C4.DIF had the best survival in both OS and PFS in univariate analyses, while C5.MES had the worst survival. While C2.IMM and C4.DIF had inseparable outcomes, other clinical features were distinct between these groups (see also Table 1B). B, Multivariate survival analysis results from Cox proportional hazard models adjusting for different known prognostic risk factors. The top table provides OS results, while the bottom portion provides PFS results. Each column in the table represents an independent model that adjusts for different risk factors. To assess the significance of a factor, we used the omnibus likelihood ratio test evaluating the likelihood with and without that factor in the model. As such, the resulting P values are associated with the entire factor and not a specific level of that factor; this is indicated by a vertical bar to clarify and the asterisks (*) indicate that the omnibus likelihood ratio test P value was below 0.05 for the entire marked variable. The score test was used to compute confidence intervals (CIs), therefore, these may not always match the P value results.

Figure 5.

Univariate and multivariate survival analysis with PrOTYPE subtypes. A, Kaplan–Meir survival curves for OS and PFS by molecular subtype. C2.IMM and C4.DIF had the best survival in both OS and PFS in univariate analyses, while C5.MES had the worst survival. While C2.IMM and C4.DIF had inseparable outcomes, other clinical features were distinct between these groups (see also Table 1B). B, Multivariate survival analysis results from Cox proportional hazard models adjusting for different known prognostic risk factors. The top table provides OS results, while the bottom portion provides PFS results. Each column in the table represents an independent model that adjusts for different risk factors. To assess the significance of a factor, we used the omnibus likelihood ratio test evaluating the likelihood with and without that factor in the model. As such, the resulting P values are associated with the entire factor and not a specific level of that factor; this is indicated by a vertical bar to clarify and the asterisks (*) indicate that the omnibus likelihood ratio test P value was below 0.05 for the entire marked variable. The score test was used to compute confidence intervals (CIs), therefore, these may not always match the P value results.

Close modal

Any potential of gene expression subtype information to guide patient management cannot be realized without a de facto standard and validated assay that can be applied in a single-patient setting using pathology-standard fixed tissues, such as would be encountered in the clinic. Here, we have defined a de facto standard for HGSOC gene expression molecular subtypes using the consensus from two independent models derived from 1,650 bona fide HGSOC samples with array data. Using these samples, we designed and validated PrOTYPE, a robust and pragmatic 55-gene classifier based on the NanoString gene expression platform. We evaluated the analytic validity of PrOTYPE by testing it in newly synthesized CodeSets. Finally, we confirmed reported associations between subtype and clinicopathologic parameters.

We have addressed limitations of prior work, including designing PrOTYPE with an established single-sample normalization and batch correction approach (35). PrOTYPE is built on the NanoString platform, known to be tolerant to different analytes and well suited for FFPE tissues (31, 32, 59, 60). This particular feature is critical to implementation in modern pathology laboratories and may also enable retrospective reexamination of archival specimen collections and clinical trials. Our model is not derived from a single dataset, but instead uses two approaches to integrate information from 14 array studies and a consortium collection of >3,000 tumors. Every sample included has been curated to ensure inclusion of a pure population of HGSOC, using either central review by expert gyne-pathologists (NanoString cohort) or a proven mechanism to minimize non-HGSOCs from historic datasets (array data cohorts; ref. 30). Using the intersection of parallel approaches as a de facto standard, we provide a first example of an HGSOC gene expression subtype classifier derived using the step-wise best practice recommended by the Institute of Medicine (26). The PrOTYPE assay is, therefore, at the so-called “bright line,” bringing gene expression molecular subtypes to the stage at which evaluation for clinical utility and use may begin.

Similar to NanoString's Prosigna assay for breast cancer (31, 32), we used a reference-based strategy for single-sample classification and batch effect correction (35). In our development phase, one limitation is that the chosen references are finite resources and will not be sufficient for long-term, widespread distribution. Less restricted reference source material will need to be chosen and integrated into the PrOTYPE assay to ensure sustainability. PrOTYPE is designed exclusively for gene expression HGSOC molecular subtyping, application on other histotypes is uninterpretable. Furthermore, the relationship between subtype and effects of neoadjuvant chemotherapy, a common practice for modern management of HGSOC, are unclear. Mitigating this could be solved by using pretreatment biopsies, however, diagnostic biopsies currently favor omentum for ease of access to the tumor mass, and our data suggest the omental microenvironment strongly biases toward a C1.MES prediction. Thus, the clinical utility of PrOTYPE may relate to consistency of phenotypes predicted from multiple anatomic sites within a patient and remains to be tested.

Our dataset enables validation of biological characteristics that smaller datasets have been unable to address. Consistent with prominent desmoplastic stroma reported from metastatic disease (61–63), we noted a systematic shift of all subtypes to a C1.MES phenotype at extra-adnexal sites. In addition, few cases of C1.MES were clear of visible macroscopic residual disease, suggesting a potential application for PrOTYPE may be predicting cytoreductability. Application of PrOTYPE to biopsied specimens may provide valuable information prior to surgery and allow investigation of whether C1.MES tumors are a logical choice for neoadjuvant- or other presurgical-targeted therapies. However, given the limitations of our retrospective cohort, with potential heterogeneity in surgical practice, a well-designed prospective study is warranted to test this hypothesis.

In multivariate models, we observed waning prognostic value for molecular subtypes in the context of known age, stage, CD8+ TIL infiltration, residual disease, and germline deleterious BRCA1/2 status, albeit with reduced sample size. Previous studies have suggested there may be an overall enrichment of BRCA1 disruptions (including methylation, somatic, and germline events together) within C2.IMM (64), however, data on somatic events affecting BRCA1/2, and other measures of homologous repair deficiency, are currently unavailable in our dataset. Nonetheless, subtype appears to capture some information for critical prognostic variables. However, for a disease with a generally poor prognosis, prediction may be more important.

In keeping with previous observations, only a modest proportion of cases reflect a “pure” phenotype signature (13, 18). We suggest that thresholds for subtype prediction, and implied utility, should be determined empirically, these may be specific to a given intervention. While few clinical trials have invested in HGSOC gene expression subtyping, at least one points to differential benefits of bevacizumab across subtypes (36). Potential benefits to C2.IMM are currently being tested using PrOTYPE in a trial of pembrolizumab in recurrent disease (NCT03732950). Likewise, there is an ongoing investigation in targeting both the reactive stromal features of C1.MES, in the BEACON trial (NCT03363867; combined bevacizumab, atezolizumab and cobimetinib), and the stem-like features of C5.PRO, in a phase II study of vinorelbine (NCT03188159). It remains to be seen whether stringent or lax subtype thresholding is important to patient selection for these interventions. Other umbrella multicenter pragmatic studies such as INOVATe (Individualized Ovarian Cancer Treatment Through Integration of Genomic Pathology into Multidisciplinary Care) are incorporating PrOTYPE in their evaluations of guided treatment modalities (65).

While only small improvements in HGSOC outcomes have been achieved in the past decades, an increasing number of therapeutic options are emerging with a growing need to identify response groups to targeted therapies such as angiogenesis inhibitors (36, 66), immune modulators (67–69), and PARP inhibitors (70–72). In the context of these new therapeutics, PrOTYPE will enable objective testing of the clinical utility of intrinsic HGSOC gene expression subtypes, a threshold that has previously been elusive. Similar to molecular profiling tools that are already emerging for other cancers (31, 32, 73–75), the clinical grade PrOTYPE assay is ready for integration into clinical trials, as well as research applications.

AOCS Group reports grants from AstraZeneca Pty Ltd outside the submitted work. S.N. Banerjee reports grants and personal fees from AstraZeneca and Tesaro, personal fees from GlaxoSmithKline, Clovis, Amgen, Seattle Genetics, Genmab, Roche, MSD, and Merck, and other from NuCana (travel grant to congress) outside the submitted work. Y.-E. Chiew reports grants from National Health and Medical Research Council (enabling grants ID 310670 and 628903) and Cancer Institute New South Wales (grants ID 12/RIG/1-17 and 15/RIG/1-16) during the conduct of the study. S. Fereday reports grants from AstraZeneca Pty Ltd outside the submitted work. R.T. Fortner reports grants from the German Federal Ministry of Education and Research, Programme of Clinical Biomedical Research (01 GB 9401), and other intramural funding from the German Cancer Research Center (DKFZ) during the conduct of the study. J. García-Donas reports grants and personal fees from Pfizer, Roche, Astellas, Bristol-Myers, and Merck and personal fees from Janssen outside the submitted work. C.S. Greene reports grants from NIH during the conduct of the study and grants from Pfizer (to develop software for connectivity search over heterogeneous networks of biomedical entities) outside the submitted work. J. Hendley reports grants from AstraZeneca Pty Ltd outside the submitted work. S.H. Kaufmann reports grants from NCI (during the conduct of the study), Eli Lilly, Takeda, and Cyteir (outside the submitted work), and is listed as a co-inventor on a patent involving prediction of ovarian cancer response to PARP inhibitors and platinum compounds owned by Mayo Clinic (outside the submitted work). C.J. Kennedy reports grants from National Health and Medical Research Council (enabling grants ID 310670 and 628903) and Cancer Institute New South Wales (grants ID 12/RIG/1-17 and 15/RIG/1-16) during the conduct of the study. G. Macintyre reports he is CTO and co-founder of a startup Tailor Bio Ltd which develops genomic biomarkers for treatment response in high-grade serous ovarian cancer. S. Mahner reports grants and personal fees from AbbVie, Astra Zeneca, Clovis, Eisai, GlaxoSmithKline, Medac, MSD, Novartis, Olympus, PharmaMar, Pfizer, Roche, Sensor Kinesis, Teva, and Tesaro outside the submitted work. I.A. McNeish reports personal fees from Clovis Oncology (advisory board), grants and personal fees from AstraZeneca (institutional grant support and advisory board), personal fees from Tesaro (advisory board), Carrick Therapeutics (advisory Board), and Scancell (advisory board) outside the submitted work. N. Nevins reports grants from National Health and Medical Research Council (enabling grants ID 310670 and 628903) and Cancer Institute New South Wales (grants ID 12/RIG-1-17 and 15/RIG/1-16) during the conduct of the study. C.L. Pearce reports grants from NCI and DoD during the conduct of the study. M.C. Pike reports grants from Department of Defense Ovarian Cancer Research Program and NCI during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

A. Talhouk: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, visualization, methodology, writing-original draft, project administration, writing-review and editing. J. George: Conceptualization, data curation, formal analysis, funding acquisition, validation, visualization, methodology, writing-review and editing. C. Wang: Resources, data curation, investigation. T. Budden: Data curation, formal analysis, validation, investigation, visualization, methodology. T.Z. Tan: Resources, data curation, formal analysis, validation, visualization, methodology, writing-review and editing. D.S. Chiu: Resources, data curation, software, formal analysis, validation, visualization. S. Kommoss: Conceptualization, resources, data curation, writing-review and editing. H.S. Leong: Resources, data curation, investigation. S. Chen: Resources, data curation, investigation. M.P. Intermaggio: Resources, data curation, investigation, project administration. B. Gilks: Resources, investigation, methodology. T.M. Nazeran: Data curation, investigation, methodology. M. Volchek: Data curation, investigation, methodology. W. Elatre: Data curation, investigation, methodology. R.C. Bentley: Data curation, investigation, methodology. J. Senz: Resources, data curation, supervision, investigation, methodology. A. Lum: Resources, data curation, supervision, investigation, methodology. V. Chow: Data curation, investigation, methodology. H. Sudderuddin: Data curation, investigation, methodology. R. Mackenzie: Data curation, investigation, methodology. S.C.Y. Leung: Resources, data curation, validation. G. Liu: Software, visualization, methodology. D. Johnson: Software, formal analysis, visualization, methodology. B. Chen: Software, visualization, methodology. AOCS Group: Resources, data curation, project administration. J. Alsop: Resources, data curation. S.N. Banerjee: Resources, data curation. S. Behrens: Resources, data curation. C. Bodelon: Resources, data curation. A.H. Brand: Resources, data curation, writing-review and editing. L. Brinton: Resources, data curation, writing-review and editing. M.E. Carney: Resources, data curation. Y.-E. Chiew: Resources, data curation, investigation. K.L. Cushing-Haugen: Resources, data curation, formal analysis, methodology, writing-review and editing. C. Cybulski: Resources, data curation. D. Ennis: Resources, data curation. S. Fereday: Resources, data curation, project administration, writing-review and editing. R.T. Fortner: Resources, data curation, writing-review and editing. J. García-Donas: Resources, data curation. A. Gentry-Maharaj: Resources, data curation. R. Glasspool: Resources, data curation. T. Goranova: Resources, data curation. C.S. Greene: Conceptualization, supervision, funding acquisition, validation, methodology, writing-review and editing. P. Haluska: Resources, data curation. H.R. Harris: Resources, data curation, writing-review and editing. J. Hendley: Resources, data curation. B.Y. Hernandez: Resources, data curation. E. Herpel: Resources, data curation. M. Jimenez-Linan: Resources, data curation. C. Karpinskyj: Resources, data curation. S.H. Kaufmann: Resources, data curation. G.L. Keeney: Resources, data curation. C.J. Kennedy: Resources, data curation, investigation, writing-review and editing. M. Köbel: Resources, data curation, investigation, writing-review and editing. J.M. Koziak: Resources, data curation. M.C. Larson: Resources, data curation. J. Lester: Resources, data curation. L.-A. Lewsley: Resources, data curation. J. Lissowska: Resources, data curation. J. Lubiński: Resources, data curation. H. Luk: Resources, data curation. G. Macintyre: Resources, data curation. S. Mahner: Resources, data curation. I.A. McNeish: Resources, data curation, writing-review and editing. J. Menkiszak: Resources, data curation. N. Nevins: Resources, data curation. A. Osorio: Resources, data curation. O. Oszurek: Resources, data curation. J. Palacios: Resources, data curation. S. Hinsley: Resources, data curation. C.L. Pearce: Resources, data curation, funding acquisition, writing-review and editing. M.C. Pike: Resources, data curation, funding acquisition. A.M. Piskorz: Resources, data curation, investigation. I. Ray-Coquard: Resources, data curation. V. Rhenius: Resources, data curation. C. Rodriguez-Antona: Resources, data curation. R. Sharma: Data curation, investigation. M.E. Sherman: Resources, data curation, investigation. D. De Silva: Resources, data curation. N. Singh: Resources, data curation, investigation. P. Sinn: Resources, data curation. D. Slamon: Resources, data curation. H. Song: Resources, data curation. H. Steed: Resources, data curation. E.A. Stronach: Resources, data curation. P.J. Thompson: Resources, data curation. A. Tołoczko: Resources, data curation. B. Trabert: Resources, data curation, writing-review and editing. N. Traficante: Resources, data curation, project administration. C.-C. Tseng: Resources, data curation. M. Widschwendter: Resources, data curation, writing-review and editing. L.R. Wilkens: Resources, data curation. S.J. Winham: Resources, data curation. B. Winterhoff: Resources, data curation, methodology. A. Beeghly-Fadiel: Resources, data curation. J. Benitez: Resources, data curation. A. Berchuck: Resources, data curation, funding acquisition, writing-review and editing. J.D. Brenton: Conceptualization, resources, data curation, funding acquisition, writing-review and editing. R. Brown: Resources, data curation. J. Chang-Claude: Resources, data curation. G. Chenevix-Trench: Resources, data curation, funding acquisition, writing-review and editing. A. deFazio: Conceptualization, resources, data curation, funding acquisition, writing-review and editing. P.A. Fasching: Resources, data curation, writing-review and editing. M.J. García: Resources, data curation. S.A. Gayther: resources, data curation. M.T. Goodman: Resources, data curation. J. Gronwald: Resources, data curation. M.J. Henderson: Resources, data curation. B.Y. Karlan: Resources, data curation, writing-review and editing. L.E. Kelemen: Resources, data curation, writing-review and editing. U. Menon: Resources, data curation. S. Orsulic: Resources, data curation, funding acquisition. P.D.P. Pharoah: Conceptualization, resources, data curation, funding acquisition, writing-review and editing. N. Wentzensen: Resources, data curation, funding acquisition, writing-review and editing. A.H. Wu: Resources, data curation, funding acquisition, writing-review and editing. J.M. Schildkraut: Resources, data curation, funding acquisition, writing-review and editing. M.A. Rossing: Conceptualization, resources, data curation, supervision, funding acquisition. G.E. Konecny: Resources, data curation, funding acquisition, writing-review and editing. D.G. Huntsman: Resources, funding acquisition. R.Y.-J. Huang: Resources, data curation, supervision, methodology, writing-review and editing. E.L. Goode: Resources, data curation, supervision, funding acquisition, writing-review and editing. S.J. Ramus: Conceptualization, resources, data curation, supervision, funding acquisition, project administration, writing-review and editing. J.A. Doherty: Conceptualization, resources, data curation, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing. D.D. Bowtell: Conceptualization, resources, supervision, funding acquisition, writing-review and editing. M.S. Anglesio: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing.

We thank all the study participants who contributed to this study and all the researchers, clinicians, and technical and administrative staff who have made this work possible. This project received technical and data management support from OVCARE through the Cheryl Brown Ovarian Cancer Outcomes Unit and the Genetic Pathology Evaluation Centre. The AOV study recognizes the valuable contributions from Mie Konno, Shuhong Liu, Michelle Darago, Faye Chambers, and the staff at the Tom Baker Cancer Centre Translational Laboratories. The Australian Ovarian Cancer Study (AOCS) gratefully acknowledges additional support from Ovarian Cancer Australia and the Peter MacCallum Foundation. The AOCS also acknowledges the cooperation of the participating institutions in Australia and acknowledges the contribution of the study nurses, research assistants, and all clinical and scientific collaborators to the study. The complete AOCS Group can be found at www.aocstudy.org. Furthermore, the authors thank Olivier Tredan and Pierre Heudel as investigators on the TRIO14 study and Sandrine Berge-Montamat as assistant for clinical research. Direct funding for this project was provided by the NIH (R01-CA172404, to principal investigator: S.J. Ramus; and R01-CA168758, to principal investigators: J.A. Doherty and M.A. Rossing), the Canadian Institutes for Health Research (Proof-of-Principle I program, to principal investigators: D.G. Huntsman and M.S. Anglesio), and the United States Department of Defense Ovarian Cancer Research Program (OC110433, to principal investigator: D.D. Bowtell). A. Talhouk was funded through a Michael Smith Foundation for Health Research Scholar Award. M.S. Anglesio was funded through a Michael Smith Foundation for Health Research Scholar Award and the Janet D. Cottrelle Foundation Scholars program managed by the BC Cancer Foundation. J. George was partially supported by the NIH/NCI award number P30CA034196. In addition, other coauthor, biobanks, patient-recruitment studies, and programs received funding that has indirectly supported this work. C. Wang was a Career Enhancement Awardee of the Mayo Clinic SPORE in Ovarian Cancer (P50 CA136393). D.G. Huntsman received support from the Dr. Chew Wei Memorial Professorship in Gynecologic Oncology and the Canada Research Chairs program (Research Chair in Molecular and Genomic Pathology). M. Widschwendter received funding from the European Union's Horizon 2020 European Research Council Programme, H2020 BRCA-ERC under grant agreement no. 742432, as well as the charity, The Eve Appeal (https://eveappeal.org.uk/), and support of the National Institute for Health Research and the University College London Hospitals Biomedical Research Centre. M.J. Henderson received funding from Cancer Australia (1067110). G.E. Konecny was supported by the Miriam and Sheldon Adelson Medical Research Foundation. B.Y. Karlan was funded by the American Cancer Society Early Detection Professorship (SIOP-06-258-01-COUN) and the National Center for Advancing Translational Sciences, grant UL1TR000124. H.R. Harris was supported by the NIH/NCI award number K22 CA193860. OVCARE (including the VAN study) received support through the BC Cancer Foundation and The VGH+UBC Hospital Foundation (to authors A. Talhouk, B. Gilks, D.G. Huntsman, and M.S. Anglesio). The AOV study was supported by the Canadian Institutes of Health Research (MOP-86727). The Gynaecological Oncology Biobank at Westmead, a member of the Australasian Biospecimen Network-Oncology group, was funded by the National Health and Medical Research Council Enabling Grants ID 310670 and ID 628903 and the Cancer Institute NSW Grants ID 12/RIG/1-17 and 15/RIG/1-16. The AOCS Group was supported by the U.S. Army Medical Research and Materiel Command under DAMD17-01-1-0729, The Cancer Council Victoria, Queensland Cancer Fund, The Cancer Council New South Wales, The Cancer Council South Australia, The Cancer Council Tasmania and The Cancer Foundation of Western Australia (Multi-State Applications 191, 211, and 182), and the National Health and Medical Research Council of Australia (NHMRC; ID199600; ID400413; and ID400281). BriTROC-1 was funded by Ovarian Cancer Action (to I.A. McNeish and J.D. Brenton, grant number 006) and supported by Cancer Research UK (grant numbers A15973, A15601, A18072, A17197, A19274, and A19694) and the National Institute for Health Research Cambridge and Imperial Biomedical Research Centres. Samples from the Mayo Clinic were collected and provided with support from the National Institutes of Health/National Cancer Institute P50 CA136393 (to E.L. Goode, G.L. Keeney, S.H. Kaufmann, and M.E. Sherman). Samples from the German Ovarian Cancer Study were collected and provided with support from the German Federal Ministry of Education and Research, Programme of Clinical Biomedical Research (01 GB 9401), and the German Cancer Research Center (DKFZ).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
The Cancer Genome Atlas Network
. 
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
2.
McAlpine
J
,
Leon-Castillo
A
,
Bosse
T
. 
The rise of a novel classification system for endometrial carcinoma; integration of molecular subclasses
.
J Pathol
2018
;
244
:
538
49
.
3.
Northcott
PA
,
Shih
DJ
,
Remke
M
,
Cho
YJ
,
Kool
M
,
Hawkins
C
, et al
Rapid, reliable, and reproducible molecular sub-grouping of clinical medulloblastoma samples
.
Acta Neuropathol
2012
;
123
:
615
26
.
4.
Clot
G
,
Jares
P
,
Gine
E
,
Navarro
A
,
Royo
C
,
Pinyol
M
, et al
A gene signature that distinguishes conventional and leukemic nonnodal mantle cell lymphoma helps predict outcome
.
Blood
2018
;
132
:
413
22
.
5.
Mottok
A
,
Wright
G
,
Rosenwald
A
,
Ott
G
,
Ramsower
C
,
Campo
E
, et al
Molecular classification of primary mediastinal large B-cell lymphoma using routinely available tissue specimens
.
Blood
2018
;
132
:
2401
5
.
6.
McCart Reed
AE
,
Kalita-De Croft
P
,
Kutasovic
JR
,
Saunus
JM
,
Lakhani
SR
. 
Recent advances in breast cancer research impacting clinical diagnostic practice
.
J Pathol
2019
;
247
:
552
62
.
7.
Shah
SP
,
Kobel
M
,
Senz
J
,
Morin
RD
,
Clarke
BA
,
Wiegand
KC
, et al
Mutation of FOXL2 in granulosa-cell tumors of the ovary
.
N Engl J Med
2009
;
360
:
2719
29
.
8.
Talhouk
A
,
McConechy
MK
,
Leung
S
,
Li-Chang
HH
,
Kwon
JS
,
Melnyk
N
, et al
A clinically applicable molecular-based classification for endometrial cancers
.
Br J Cancer
2015
;
113
:
299
310
.
9.
Cancer Genome Atlas Research Network
. 
Integrated genomic analyses of ovarian carcinoma
.
Nature
2011
;
474
:
609
15
.
10.
Tothill
RW
,
Tinker
AV
,
George
J
,
Brown
R
,
Fox
SB
,
Lade
S
, et al
Novel molecular subtypes of serous and endometrioid ovarian cancer linked to clinical outcome
.
Clin Cancer Res
2008
;
14
:
5198
208
.
11.
Torre
LA
,
Trabert
B
,
DeSantis
CE
,
Miller
KD
,
Samimi
G
,
Runowicz
CD
, et al
Ovarian cancer statistics, 2018
.
CA Cancer J Clin
2018
;
68
:
284
96
.
12.
Committee on the State of the Science in Ovarian Cancer Research, Board on Health Care Services, Institute of Medicine, National Academies of Sciences Engineering and Medicine
,
editors
.
Ovarian cancers: evolving paradigms in research and care
. 1 ed.
Washington, D.C.
:
The National Academies Press
; 
2016
.
13.
Verhaak
RG
,
Tamayo
P
,
Yang
JY
,
Hubbard
D
,
Zhang
H
,
Creighton
CJ
, et al
Prognostically relevant gene signatures of high-grade serous ovarian carcinoma
.
J Clin Invest
2013
;
123
:
517
25
.
14.
Konecny
GE
,
Wang
C
,
Hamidi
H
,
Winterhoff
B
,
Kalli
KR
,
Dering
J
, et al
Prognostic and therapeutic relevance of molecular subtypes in high-grade serous ovarian cancer
.
J Natl Cancer Inst
2014
;
106
:
dju249
.
15.
Tan
TZ
,
Miow
QH
,
Huang
RY
,
Wong
MK
,
Ye
J
,
Lau
JA
, et al
Functional genomics identifies five distinct molecular subtypes with clinical relevance and pathways for growth control in epithelial ovarian cancer
.
EMBO Mol Med
2013
;
5
:
983
98
.
16.
Tan
TZ
,
Yang
H
,
Ye
J
,
Low
J
,
Choolani
M
,
Tan
DS
, et al
CSIOVDB: a microarray gene expression database of epithelial ovarian cancer subtype
.
Oncotarget
2015
;
6
:
43843
52
.
17.
Way
GP
,
Rudd
J
,
Wang
C
,
Hamidi
H
,
Fridley
BL
,
Konecny
GE
, et al
Comprehensive cross-population analysis of high-grade serous ovarian cancer supports no more than three subtypes
.
G3
2016
;
6
:
4097
103
.
18.
Chen
GM
,
Kannan
L
,
Geistlinger
L
,
Kofia
V
,
Safikhani
Z
,
Gendoo
DMA
, et al
Consensus on molecular subtypes of high-grade serous ovarian carcinoma
.
Clin Cancer Res
2018
;
24
:
5037
47
.
19.
Gruosso
T
,
Gigoux
M
,
Manem
VSK
,
Bertos
N
,
Zuo
D
,
Perlitch
I
, et al
Spatially distinct tumor immune microenvironments stratify triple-negative breast cancers
.
J Clin Invest
2019
;
129
:
1785
800
.
20.
Ueno
H
,
Kanemitsu
Y
,
Sekine
S
,
Ishiguro
M
,
Ito
E
,
Hashiguchi
Y
, et al
Desmoplastic pattern at the tumor front defines poor-prognosis subtypes of colorectal cancer
.
Am J Surg Pathol
2017
;
41
:
1506
12
.
21.
Roman-Perez
E
,
Casbas-Hernandez
P
,
Pirone
JR
,
Rein
J
,
Carey
LA
,
Lubet
RA
, et al
Gene expression in extratumoral microenvironment predicts clinical outcome in breast cancer patients
.
Breast Cancer Res
2012
;
14
:
R51
.
22.
Busuttil
RA
,
George
J
,
Tothill
RW
,
Ioculano
K
,
Kowalczyk
A
,
Mitchell
C
, et al
A signature predicting poor prognosis in gastric and ovarian cancer represents a coordinated macrophage and stromal response
.
Clin Cancer Res
2014
;
20
:
2761
72
.
23.
Cohen
N
,
Shani
O
,
Raz
Y
,
Sharon
Y
,
Hoffman
D
,
Abramovitz
L
, et al
Fibroblasts drive an immunosuppressive and growth-promoting microenvironment in breast cancer via secretion of Chitinase 3-like 1
.
Oncogene
2017
;
36
:
4457
68
.
24.
Helland
A
,
Anglesio
MS
,
George
J
,
Cowin
PA
,
Johnstone
CN
,
House
CM
, et al
Deregulation of MYCN, LIN28B and LET7 in a molecular subtype of aggressive high-grade serous ovarian cancers
.
PLoS One
2011
;
6
:
e18064
.
25.
Kobel
M
,
Rahimi
K
,
Rambau
PF
,
Naugler
C
,
Le Page
C
,
Meunier
L
, et al
An immunohistochemical algorithm for ovarian carcinoma typing
.
Int J Gynecol Pathol
2016
;
35
:
430
41
.
26.
Institute of Medicine
;
Micheel
CM
,
Nass
SJ
,
Omenn
GS
,
editors
.
Evolution of translational omics lessons learned and the path forward
.
Washington, DC
:
The National Academies Press
; 
2012
.
27.
Wang
C
,
Armasu
SM
,
Kalli
KR
,
Maurer
MJ
,
Heinzen
EP
,
Keeney
GL
, et al
Pooled clustering of high-grade serous ovarian cancer gene expression leads to novel consensus subtypes associated with survival and surgical outcomes
.
Clin Cancer Res
2017
;
23
:
4077
85
.
28.
Kommoss
S
,
Gilks
CB
,
du Bois
A
,
Kommoss
F
. 
Ovarian carcinoma diagnosis: the clinical impact of 15 years of change
.
Br J Cancer
2016
;
115
:
993
9
.
29.
Kurman
RJ
,
Carcangiu
ML
,
Herrington
CS
,
Young
RH
,
editors.
WHO classification of tumours of female reproductive organs
. 4 ed.
Lyon, France
:
IARC-WHO Press
; 
2014
.
30.
Peres
LC
,
Cushing-Haugen
KL
,
Anglesio
M
,
Wicklund
K
,
Bentley
R
,
Berchuck
A
, et al
Histotype classification of ovarian carcinoma: a comparison of approaches
.
Gynecol Oncol
2018
;
151
:
53
60
.
31.
Nielsen
T
,
Wallden
B
,
Schaper
C
,
Ferree
S
,
Liu
S
,
Gao
D
, et al
Analytical validation of the PAM50-based Prosigna breast cancer prognostic gene signature assay and nCounter analysis system using formalin-fixed paraffin-embedded breast tumor specimens
.
BMC Cancer
2014
;
14
:
177
.
32.
Wallden
B
,
Storhoff
J
,
Nielsen
T
,
Dowidar
N
,
Schaper
C
,
Ferree
S
, et al
Development and verification of the PAM50-based Prosigna breast cancer gene signature assay
.
BMC Med Genomics
2015
;
8
:
54
.
33.
Scott
DW
,
Chan
FC
,
Hong
F
,
Rogic
S
,
Tan
KL
,
Meissner
B
, et al
Gene expression-based model using formalin-fixed paraffin-embedded biopsies predicts overall survival in advanced-stage classical Hodgkin lymphoma
.
J Clin Oncol
2013
;
31
:
692
700
.
34.
Staiger
AM
,
Ziepert
M
,
Horn
H
,
Scott
DW
,
Barth
TFE
,
Bernd
HW
, et al
Clinical Impact of the cell-of-origin classification and the MYC/BCL2 dual expresser status in diffuse large B-cell lymphoma treated within prospective clinical trials of the German High-Grade Non-Hodgkin's Lymphoma Study Group
.
J Clin Oncol
2017
;
35
:
2515
26
.
35.
Talhouk
A
,
Kommoss
S
,
Mackenzie
R
,
Cheung
M
,
Leung
S
,
Chiu
DS
, et al
Single-patient molecular testing with NanoString nCounter data using a reference-based strategy for batch effect correction
.
PLoS One
2016
;
11
:
e0153844
.
36.
Kommoss
S
,
Winterhoff
B
,
Oberg
AL
,
Konecny
GE
,
Wang
C
,
Riska
SM
, et al
Bevacizumab may differentially improve ovarian cancer outcome in patients with proliferative and mesenchymal molecular subtypes
.
Clin Cancer Res
2017
;
23
:
3794
801
.
37.
Leong
HS
,
Galletta
L
,
Etemadmoghadam
D
,
George
J
,
Australian Ovarian Cancer Study Group
,
Kobel
M
, et al
Efficient molecular subtype classification of high-grade serous ovarian cancer
.
J Pathol
2015
;
236
:
272
7
.
38.
Millstein
J
,
Budden
T
,
Goode
EL
,
Anglesio
MS
,
Talhouk
A
,
Intermaggio
MP
, et al
Prognostic gene expression signature for high-grade serous ovarian cancer
.
Ann Oncol
2020
.
doi: 10.1016/j.annonc.2020.05.019
.
39.
Yoshihara
K
,
Tsunoda
T
,
Shigemizu
D
,
Fujiwara
H
,
Hatae
M
,
Fujiwara
H
, et al
High-risk ovarian cancer based on 126-gene expression signature is uniquely characterized by downregulation of antigen presentation pathway
.
Clin Cancer Res
2012
;
18
:
1374
85
.
40.
Rudd
J
,
Zelaya
RA
,
Demidenko
E
,
Goode
EL
,
Greene
CS
,
Doherty
JA
. 
Leveraging global gene expression patterns to predict expression of unmeasured genes
.
BMC Genomics
2015
;
16
:
1065
.
41.
Denkert
C
,
Budczies
J
,
Darb-Esfahani
S
,
Gyorffy
B
,
Sehouli
J
,
Konsgen
D
, et al
A prognostic gene expression index in ovarian cancer - validation across different independent data sets
.
J Pathol
2009
;
218
:
273
80
.
42.
Mateescu
B
,
Batista
L
,
Cardon
M
,
Gruosso
T
,
de Feraudy
Y
,
Mariani
O
, et al
miR-141 and miR-200a act on ovarian tumorigenesis by controlling oxidative stress response
.
Nat Med
2011
;
17
:
1627
35
.
43.
Zsiros
E
,
Duttagupta
P
,
Dangaj
D
,
Li
H
,
Frank
R
,
Garrabrant
T
, et al
The ovarian cancer chemokine landscape is conducive to homing of vaccine-primed and CD3/CD28-costimulated T cells prepared for adoptive therapy
.
Clin Cancer Res
2015
;
21
:
2840
50
.
44.
Meyniel
JP
,
Cottu
PH
,
Decraene
C
,
Stern
MH
,
Couturier
J
,
Lebigot
I
, et al
A genomic and transcriptomic approach for a differential diagnosis between primary and secondary ovarian carcinomas in patients with a previous history of breast cancer
.
BMC Cancer
2010
;
10
:
222
.
45.
Mok
SC
,
Bonome
T
,
Vathipadiekal
V
,
Bell
A
,
Johnson
ME
,
Wong
KK
, et al
A gene signature predictive for outcome in advanced ovarian cancer identifies a survival factor: microfibril-associated glycoprotein 2
.
Cancer Cell
2009
;
16
:
521
32
.
46.
Ferriss
JS
,
Kim
Y
,
Duska
L
,
Birrer
M
,
Levine
DA
,
Moskaluk
C
, et al
Multi-gene expression predictors of single drug responses to adjuvant chemotherapy in ovarian carcinoma: predicting platinum resistance
.
PLoS One
2012
;
7
:
e30550
.
47.
Jochumsen
KM
,
Tan
Q
,
Hogdall
EV
,
Hogdall
C
,
Kjaer
SK
,
Blaakaer
J
, et al
Gene expression profiles as prognostic markers in women with ovarian cancer
.
Int J Gynecol Cancer
2009
;
19
:
1205
13
.
48.
Hendrix
ND
,
Wu
R
,
Kuick
R
,
Schwartz
DR
,
Fearon
ER
,
Cho
KR
. 
Fibroblast growth factor 9 has oncogenic activity and is a downstream target of Wnt signaling in ovarian endometrioid adenocarcinomas
.
Cancer Res
2006
;
66
:
1354
62
.
49.
Lisowska
KM
,
Olbryt
M
,
Dudaladava
V
,
Pamula-Pilat
J
,
Kujawa
K
,
Grzybowska
E
, et al
Gene expression analysis in ovarian cancer - faults and hints from DNA microarray study
.
Front Oncol
2014
;
4
:
6
.
50.
Bild
AH
,
Yao
G
,
Chang
JT
,
Wang
Q
,
Potti
A
,
Chasse
D
, et al
Oncogenic pathway signatures in human cancers as a guide to targeted therapies
.
Nature
2006
;
439
:
353
7
.
51.
Bonome
T
,
Levine
DA
,
Shih
J
,
Randonovich
M
,
Pise-Masison
CA
,
Bogomolniy
F
, et al
A gene signature predicting for survival in suboptimally debulked patients with ovarian cancer
.
Cancer Res
2008
;
68
:
5478
86
.
52.
Chiu
DS
,
Talhouk
A
. 
diceR: an R package for class discovery using an ensemble driven approach
.
BMC Bioinformatics
2018
;
19
:
1
4
.
53.
Breiman
L
.
Out-of-bag estimation
.
Berkeley, CA
:
University of California, Berkeley
; 
1996
.
54.
Moons
KG
,
Altman
DG
,
Reitsma
JB
,
Collins
GS
. 
New guideline for the reporting of studies developing, validating, or updating a multivariable clinical prediction model: The TRIPOD statement
.
Adv Anat Pathol
2015
;
22
:
303
5
.
55.
Nembrini
S
,
Konig
IR
,
Wright
MN
. 
The revival of the Gini importance?
Bioinformatics
2018
;
34
:
3711
8
.
56.
MacKay
DJC
.
Information theory, inference, and learning algorithms
.
Cambridge, UK
:
Cambridge University Press
; 
2003
.
57.
Zhang
AW
,
McPherson
A
,
Milne
K
,
Kroeger
DR
,
Hamilton
PT
,
Miranda
A
, et al
Interfaces of malignant and immunologic clonal dynamics in ovarian cancer
.
Cell
2018
;
173
:
1755
69
.
58.
Ovarian Tumor Tissue Analysis Consortium
,
Goode
EL
,
Block
MS
,
Kalli
KR
,
Vierkant
RA
,
Chen
W
, et al
Dose-response association of CD8+ tumor-infiltrating lymphocytes and survival time in high-grade serous ovarian cancer
.
JAMA Oncol
2017
;
3
:
e173290
.
59.
Veldman-Jones
MH
,
Brant
R
,
Rooney
C
,
Geh
C
,
Emery
H
,
Harbron
CG
, et al
Evaluating robustness and sensitivity of the NanoString technologies nCounter Platform to enable multiplexed gene expression analysis of clinical samples
.
Cancer Res
2015
;
75
:
2587
93
.
60.
Malkov
VA
,
Serikawa
KA
,
Balantac
N
,
Watters
J
,
Geiss
G
,
Mashadi-Hossein
A
, et al
Multiplexed measurements of gene signatures in different analytes using the NanoString nCounter assay system
.
BMC Res Notes
2009
;
2
:
80
.
61.
Torres
D
,
Kumar
A
,
Wallace
SK
,
Bakkum-Gamez
JN
,
Konecny
GE
,
Weaver
AL
, et al
Intraperitoneal disease dissemination patterns are associated with residual disease, extent of surgery, and molecular subtypes in advanced ovarian cancer
.
Gynecol Oncol
2017
;
147
:
503
8
.
62.
Torres
D
,
Wang
C
,
Kumar
A
,
Bakkum-Gamez
JN
,
Weaver
AL
,
McGree
ME
, et al
Factors that influence survival in high-grade serous ovarian cancer: A complex relationship between molecular subtype, disease dissemination, and operability
.
Gynecol Oncol
2018
;
150
:
227
32
.
63.
Tan
TZ
,
Heong
V
,
Ye
J
,
Lim
D
,
Low
J
,
Choolani
M
, et al
Decoding transcriptomic intra-tumour heterogeneity to guide personalised medicine in ovarian cancer
.
J Pathol
2019
;
247
:
305
19
.
64.
George
J
,
Alsop
K
,
Etemadmoghadam
D
,
Hondow
H
,
Mikeska
T
,
Dobrovic
A
, et al
Nonequivalent gene expression and copy number alterations in high-grade serous ovarian cancers with BRCA1 and BRCA2 mutations
.
Clin Cancer Res
2013
;
19
:
3474
84
.
65.
deFazio
A
,
Harnett
P
,
Friedlander
M
,
Balleine
R
,
Bowtell
D
,
Samimi
G
, et al
INOVATe
.
Available from: <
https://www.westmeadinstitute.org.au/research/featured-projects/inovate-(1)/overview
>
.
66.
Oza
AM
,
Cook
AD
,
Pfisterer
J
,
Embleton
A
,
Ledermann
JA
,
Pujade-Lauraine
E
, et al
Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial
.
Lancet Oncol
2015
;
16
:
928
36
.
67.
Pietzner
K
,
Nasser
S
,
Alavi
S
,
Darb-Esfahani
S
,
Passler
M
,
Muallem
MZ
, et al
Checkpoint-inhibition in ovarian cancer: rising star or just a dream?
J Gynecol Oncol
2018
;
29
:
e93
.
68.
Oda
K
,
Hamanishi
J
,
Matsuo
K
,
Hasegawa
K
. 
Genomics to immunotherapy of ovarian clear cell carcinoma: Unique opportunities for management
.
Gynecol Oncol
2018
;
151
:
381
9
.
69.
Webb
JR
,
Milne
K
,
Kroeger
DR
,
Nelson
BH
. 
PD-L1 expression is associated with tumor-infiltrating T cells and favorable prognosis in high-grade serous ovarian cancer
.
Gynecol Oncol
2016
;
141
:
293
302
.
70.
Lheureux
S
,
Lai
Z
,
Dougherty
BA
,
Runswick
S
,
Hodgson
DR
,
Timms
KM
, et al
Long-term responders on olaparib maintenance in high-grade serous ovarian cancer: clinical and molecular characterization
.
Clin Cancer Res
2017
;
23
:
4086
94
.
71.
da Costa
A
,
do Canto
LM
,
Larsen
SJ
,
Ribeiro
ARG
,
Stecca
CE
,
Petersen
AH
, et al
Genomic profiling in ovarian cancer retreated with platinum based chemotherapy presented homologous recombination deficiency and copy number imbalances of CCNE1 and RB1 genes
.
BMC Cancer
2019
;
19
:
422
.
72.
Mirza
MR
,
Monk
BJ
,
Herrstedt
J
,
Oza
AM
,
Mahner
S
,
Redondo
A
, et al
Niraparib maintenance therapy in platinum-sensitive, recurrent ovarian cancer
.
N Engl J Med
2016
;
375
:
2154
64
.
73.
Cook
DP
,
Vanderhyden
BC
. 
Ovarian cancer and the evolution of subtype classifications using transcriptional profiling
.
Biol Reprod
2019
;
101
:
645
68
.
74.
Scott
DW
,
Abrisqueta
P
,
Wright
GW
,
Slack
GW
,
Mottok
A
,
Villa
D
, et al
New molecular assay for the proliferation signature in mantle cell lymphoma applicable to formalin-fixed paraffin-embedded biopsies
.
J Clin Oncol
2017
;
35
:
1668
77
.
75.
Scott
DW
,
Wright
GW
,
Williams
PM
,
Lih
CJ
,
Walsh
W
,
Jaffe
ES
, et al
Determining cell-of-origin subtypes of diffuse large B-cell lymphoma using gene expression in formalin-fixed paraffin-embedded tissue
.
Blood
2014
;
123
:
1214
7
.