Development and validation of a reliable DNA copy-number-based machine learning algorithm (CopyClust) for breast cancer integrative cluster classification.

The Integrative Cluster subtypes (IntClusts) provide a framework for the classification of breast cancer tumors into 10 distinct groups based on copy number and gene expression, each with unique biological drivers of disease and clinical prognoses. Gene expression data is often lacking, and accurate classification of samples into IntClusts with copy number data alone is essential. Current classification methods achieve low accuracy when gene expression data are absent, warranting the development of new approaches to IntClust classification. Copy number data from 1980 breast cancer samples from METABRIC was used to train multiclass XGBoost machine learning algorithms (CopyClust). A piecewise constant fit was applied to the average copy number profile of each IntClust and unique breakpoints across the 10 profiles were identified and converted into ~ 500 genomic regions used as features for CopyClust. These models consisted of two approaches: a 10-class model with the final IntClust label predicted by a single multiclass model and a 6-class model with binary reclassification in which four pairs of IntClusts were combined for initial multiclass classification. Performance was validated on the TCGA dataset, with copy number data generated from both SNP arrays and WES platforms. CopyClust achieved 81% and 79% overall accuracy with the TCGA SNP and WES datasets, respectively, a nine-percentage point or greater improvement in overall IntClust subtype classification accuracy. CopyClust achieves a significant improvement over current methods in classification accuracy of IntClust subtypes for samples without available gene expression data and is an easily implementable algorithm for IntClust classification of breast cancer samples with copy number data.

A Machine-Learning Tool Concurrently Models Single Omics and Phenome Data for Functional Subtyping and Personalized Cancer Medicine.

One of the major challenges in defining clinically-relevant and less heterogeneous tumor subtypes is assigning biological and/or clinical interpretations to etiological (intrinsic) subtypes. Conventional clustering/subtyping approaches often fail to define such subtypes, as they involve several discrete steps. Here we demonstrate a unique machine-learning method, phenotype mapping (PhenMap), which jointly integrates single omics data with phenotypic information using three published breast cancer datasets (n = 2045). The PhenMap framework uses a modified factor analysis method that is governed by a key assumption that, features from different omics data types are correlated due to specific "hidden/mapping" variables (context-specific mapping variables (CMV)). These variables can be simultaneously modeled with phenotypic data as covariates to yield functional subtypes and their associated features (e.g., genes) and phenotypes. In one example, we demonstrate the identification and validation of six novel "functional" (discrete) subtypes with differential responses to a cyclin-dependent kinase (CDK)4/6 inhibitor and etoposide by jointly integrating transcriptome profiles with four different drug response data from 37 breast cancer cell lines. These robust subtypes are also present in patient breast tumors with different prognosis. In another example, we modeled patient gene expression profiles and clinical covariates together to identify continuous subtypes with clinical/biological implications. Overall, this genome-phenome machine-learning integration tool, PhenMap identifies functional and phenotype-integrated discrete or continuous subtypes with clinical translational potential.

Intratumoral Transcriptome Heterogeneity Is Associated With Patient Prognosis and Sidedness in Patients With Colorectal Cancer Treated With Anti-EGFR Therapy From the CO.20 Trial.

PURPOSE: Metastatic colorectal cancers (mCRCs) assigned to the transit-amplifying (TA) CRCAssigner subtype are more sensitive to anti-epidermal growth factor receptor (EGFR) therapy. We evaluated the association between the intratumoral presence of TA signature (TA-high/TA-low, dubbed as TA-ness classification) and outcomes in CRCs treated with anti-EGFR therapy. PATIENTS AND METHODS: The TA-ness classes were defined in a discovery cohort (n = 84) and independently validated in a clinical trial (CO.20; cetuximab monotherapy arm; n = 121) and other samples using an established NanoString-based gene expression assay. Progression-free survival (PFS), overall survival (OS), and disease control rate (DCR) according to TA-ness classification were assessed by univariate and multivariate analyses. RESULTS: The TA-ness was measured in 772 samples from 712 patients. Patients (treated with anti-EGFR therapy) with TA-high tumors had significantly longer PFS (discovery hazard ratio [HR], 0.40; 95% CI, 0.25 to 0.64; P < .001; validation HR, 0.65; 95% CI, 0.45 to 0.93; P = .018), longer OS (discovery HR, 0.48; 95% CI, 0.29 to 0.78; P = .003; validation HR, 0.67; 95% CI, 0.46 to 0.98; P = .04), and higher DCR (discovery odds ratio [OR]; 14.8; 95% CI, 4.30 to 59.54; P < .001; validation OR, 4.35; 95% CI, 2.00 to 9.09; P < .001). TA-ness classification and its association with anti-EGFR therapy outcomes were further confirmed using publicly available data (n = 80) from metastatic samples (PFS P < .001) and patient-derived xenografts (P = .042). In an exploratory analysis of 55 patients with RAS/BRAF wild-type and left-sided tumors, TA-high class was significantly associated with longer PFS and trend toward higher response rate (PFS HR, 0.53; 95% CI, 0.28 to 1.00; P = .049; OR, 5.88; 95% CI, 0.71 to 4.55; P = .09; response rate 33% in TA-high and 7.7% in TA-low). CONCLUSION: TA-ness classification is associated with prognosis in patients with mCRC treated with anti-EGFR therapy and may further help understanding the value of sidedness in patients with RAS/BRAF wild-type tumors.

Detection of postoperative plasma circulating tumour DNA and lack of CDX2 expression as markers of recurrence in patients with localised colon cancer.

BACKGROUND: Colon cancer (CC) is a heterogeneous disease. Novel prognostic factors beyond pathological staging are required to accurately identify patients at higher risk of relapse. Integrating these new biological factors, such as plasma circulating tumour DNA (ctDNA), CDX2 staining, inflammation-associated cytokines and transcriptomic consensus molecular subtypes (CMS) classification, into a multimodal approach may improve our accuracy in determining risk of recurrence. METHODS: One hundred and fifty patients consecutively diagnosed with localised CC were prospectively enrolled in our study. ctDNA was tracked to detect minimal residual disease by droplet digital PCR. CDX2 expression was analysed by immunostaining. Plasma levels of cytokines potentially involved in disease progression were measured using ELISAs. A 96 custom gene panel for nCounter assay was used to classify CC into colorectal cancer assigner and CMS. RESULTS: Most patients were classified into CMS4 (37%) and CMS2 (28%), followed by CMS1 (20%) and CMS3 (15%) groups. CDX2-negative tumours were enriched in CMS1 and CMS4 subtypes. In univariable analysis, prognosis was influenced by primary tumour location, stage, vascular and perineural invasion together with high interleukin-6 plasma levels at baseline, tumours belonging to CMS 1 vs CMS2 +CMS3, ctDNA presence in plasma and CDX2 loss. However, only positive ctDNA in plasma samples (HR 13.64; p=0.002) and lack of CDX2 expression (HR 23.12; p=0.001) were found to be independent prognostic factors for disease-free survival in the multivariable model. CONCLUSIONS: ctDNA detection after surgery and lack of CDX2 expression identified patients at very high risk of recurrence in localised CC.

Analytical Validation of Multiplex Biomarker Assay to Stratify Colorectal Cancer into Molecular Subtypes.

Previously, we classified colorectal cancers (CRCs) into five CRCAssigner (CRCA) subtypes with different prognoses and potential treatment responses, later consolidated into four consensus molecular subtypes (CMS). Here we demonstrate the analytical development and validation of a custom NanoString nCounter platform-based biomarker assay (NanoCRCA) to stratify CRCs into subtypes. To reduce costs, we switched from the standard nCounter protocol to a custom modified protocol. The assay included a reduced 38-gene panel that was selected using an in-house machine-learning pipeline. We applied NanoCRCA to 413 samples from 355 CRC patients. From the fresh frozen samples (n = 237), a subset had matched microarray/RNAseq profiles (n = 47) or formalin-fixed paraffin-embedded (FFPE) samples (n = 58). We also analyzed a further 118 FFPE samples. We compared the assay results with the CMS classifier, different platforms (microarrays/RNAseq) and gene-set classifiers (38 and the original 786 genes). The standard and modified protocols showed high correlation (> 0.88) for gene expression. Technical replicates were highly correlated (> 0.96). NanoCRCA classified fresh frozen and FFPE samples into all five CRCA subtypes with consistent classification of selected matched fresh frozen/FFPE samples. We demonstrate high and significant subtype concordance across protocols (100%), gene sets (95%), platforms (87%) and with CMS subtypes (75%) when evaluated across multiple datasets. Overall, our NanoCRCA assay with further validation may facilitate prospective validation of CRC subtypes in clinical trials and beyond.

Benefit from anti-EGFRs in RAS and BRAF wild-type metastatic transverse colon cancer: a clinical and molecular proof of concept study.

OBJECTIVE: Primary tumour location is regarded as a reliable surrogate of colorectal cancer biology. Sensitivity to anti-EGFRs (Epidermal Growth Factor Receptor) of metastatic transverse colon cancers (mTCCs) has usually been assumed similar to right-sided tumours; however, evidence about the clinical behaviour of mTCC is limited. Thus, to verify sensitivity of mTCC to anti-EGFRs we conducted the present study. METHODS: Patients with RAS/BRAF wild-type microsatellite stable (MSS) mTCC receiving anti-EGFR monotherapy, or in combination with irinotecan if clearly irinotecan-refractory, were included. Hypothesising an overall response rate (ORR) of 35%, 11 patients, of whom at least 3 were responders, were necessary to be able to reject the null hypothesis of an ORR of 5%, with α and ß errors of 0.05 and 0.20. PRESSING panel and consensus molecular subtypes (CMS) were assessed on tumour samples, whereas in-silico data were obtained from TCGA dataset. RESULTS: Among nine eligible patients, four and three achieved response and disease stabilisation (ORR 44%). At a median follow-up of 23.1 months, median progression-free survival and overall survival were 7.3 (95% CI 3.9 to NA) and 15.0 months (95% CI 10.0 to NA), respectively. A MET amplification and an ERBB4 S303F substitution were detected in patients with rapid disease progression, while others had PRESSING panel-negative tumours with CMS2 or CMS4 subtypes. CONCLUSIONS: RAS/BRAF wild-type MSS mTCCs may be sensitive to anti-EGFRs, as confirmed by molecular analyses.

Microenvironmental niche divergence shapes BRCA1-dysregulated ovarian cancer morphological plasticity.

How tumor microenvironmental forces shape plasticity of cancer cell morphology is poorly understood. Here, we conduct automated histology image and spatial statistical analyses in 514 high grade serous ovarian samples to define cancer morphological diversification within the spatial context of the microenvironment. Tumor spatial zones, where cancer cell nuclei diversify in shape, are mapped in each tumor. Integration of this spatially explicit analysis with omics and clinical data reveals a relationship between morphological diversification and the dysregulation of DNA repair, loss of nuclear integrity, and increased disease mortality. Within the Immunoreactive subtype, spatial analysis further reveals significantly lower lymphocytic infiltration within diversified zones compared with other tumor zones, suggesting that even immune-hot tumors contain cells capable of immune escape. Our findings support a model whereby a subpopulation of morphologically plastic cancer cells with dysregulated DNA repair promotes ovarian cancer progression through positive selection by immune evasion.

polyClustR: defining communities of reconciled cancer subtypes with biological and prognostic significance.

BACKGROUND: To ensure cancer patients are stratified towards treatments that are optimally beneficial, it is a priority to define robust molecular subtypes using clustering methods applied to high-dimensional biological data. If each of these methods produces different numbers of clusters for the same data, it is difficult to achieve an optimal solution. Here, we introduce "polyClustR", a tool that reconciles clusters identified by different methods into subtype "communities" using a hypergeometric test or a measure of relative proportion of common samples. RESULTS: The polyClustR pipeline was initially tested using a breast cancer dataset to demonstrate how results are compatible with and add to the understanding of this well-characterised cancer. Two uveal melanoma datasets were then utilised to identify and validate novel subtype communities with significant metastasis-free prognostic differences and associations with known chromosomal aberrations. CONCLUSION: We demonstrate the value of the polyClustR approach of applying multiple consensus clustering algorithms and systematically reconciling the results in identifying novel subtype communities of two cancer types, which nevertheless are compatible with established understanding of these diseases. An R implementation of the pipeline is available at: https://github.com/syspremed/polyClustR.

Patient-derived organoids model treatment response of metastatic gastrointestinal cancers.

Patient-derived organoids (PDOs) have recently emerged as robust preclinical models; however, their potential to predict clinical outcomes in patients has remained unclear. We report on a living biobank of PDOs from metastatic, heavily pretreated colorectal and gastroesophageal cancer patients recruited in phase 1/2 clinical trials. Phenotypic and genotypic profiling of PDOs showed a high degree of similarity to the original patient tumors. Molecular profiling of tumor organoids was matched to drug-screening results, suggesting that PDOs could complement existing approaches in defining cancer vulnerabilities and improving treatment responses. We compared responses to anticancer agents ex vivo in organoids and PDO-based orthotopic mouse tumor xenograft models with the responses of the patients in clinical trials. Our data suggest that PDOs can recapitulate patient responses in the clinic and could be implemented in personalized medicine programs.

Molecular or Metabolic Reprograming: What Triggers Tumor Subtypes?

Tumor heterogeneity is reflected and influenced by genetic, epigenetic, and metabolic differences in cancer cells and their interactions with a complex microenvironment. This heterogeneity has resulted in the stratification of tumors into subtypes, mainly based on cancer-specific genomic or transcriptomic profiles. Subtyping can lead to biomarker identification for personalized diagnosis and therapy, but stratification alone does not explain the origins of tumor heterogeneity. Heterogeneity has traditionally been thought to arise from distinct mutations/aberrations in "driver" oncogenes. However, certain subtypes appear to be the result of adaptation to the disrupted microenvironment caused by abnormal tumor vasculature triggering metabolic switches. Moreover, heterogeneity persists despite the predominance of single oncogenic driver mutations, perhaps due to second metabolic or genetic "hits." In certain cancer types, existing subtypes have metabolic and transcriptomic phenotypes that are reminiscent of normal differentiated cells, whereas others reflect the phenotypes of stem or mesenchymal cells. The cell-of-origin may, therefore, play a role in tumor heterogeneity. In this review, we focus on how cancer cell-specific heterogeneity is driven by different genetic or metabolic factors alone or in combination using specific cancers to illustrate these concepts. Cancer Res; 76(18); 5195-200. ©2016 AACR.