Integrative clustering methods for high-dimensional molecular data.

<div>Abstract<p><b>Purpose:</b> Lower-grade gliomas (WHO grade II/III) have been classified into clinically relevant molecular subtypes based on <i>IDH</i> and 1p/19q mutation status. The purpose was to investigate whether T2/FLAIR MRI features could distinguish between lower-grade glioma molecular subtypes.</p><p><b>Experimental Design:</b> MRI scans from the TCGA/TCIA lower grade glioma database (<i>n</i> = 125) were evaluated by two independent neuroradiologists to assess (i) presence/absence of homogenous signal on T2WI; (ii) presence/absence of “T2–FLAIR mismatch” sign; (iii) sharp or indistinct lesion margins; and (iv) presence/absence of peritumoral edema. Metrics with moderate–substantial agreement underwent consensus review and were correlated with glioma molecular subtypes. Somatic mutation, DNA copy number, DNA methylation, gene expression, and protein array data from the TCGA lower-grade glioma database were analyzed for molecular–radiographic associations. A separate institutional cohort (<i>n</i> = 82) was analyzed to validate the T2–FLAIR mismatch sign.</p><p><b>Results:</b> Among TCGA/TCIA cases, interreader agreement was calculated for lesion homogeneity [<i>κ</i> = 0.234 (0.111–0.358)], T2–FLAIR mismatch sign [<i>κ</i> = 0.728 (0.538–0.918)], lesion margins [<i>κ</i> = 0.292 (0.135–0.449)], and peritumoral edema [<i>κ</i> = 0.173 (0.096–0.250)]. All 15 cases that were positive for the T2–FLAIR mismatch sign were <i>IDH</i>-mutant, 1p/19q non-codeleted tumors (<i>P</i> < 0.0001; PPV = 100%, NPV = 54%). Analysis of the validation cohort demonstrated substantial interreader agreement for the T2–FLAIR mismatch sign [<i>κ</i> = 0.747 (0.536–0.958)]; all 10 cases positive for the T2–FLAIR mismatch sign were <i>IDH</i>-mutant, 1p/19q non-codeleted tumors (<i>P</i> < 0.00001; PPV = 100%, NPV = 76%).</p><p><b>Conclusions:</b> Among lower-grade gliomas, T2–FLAIR mismatch sign represents a highly specific imaging biomarker for the <i>IDH</i>-mutant, 1p/19q non-codeleted molecular subtype. <i>Clin Cancer Res; 23(20); 6078–85. ©2017 AACR</i>.</p></div>

Multi-omics data integration for subtype identification of Chinese lower-grade gliomas: A joint similarity network fusion approach

BackgroundCharacterizing cancer molecular subtypes is crucial for improving prognosis and individualized treatment. Integrative analysis of multi-omics data has become an important approach for disease subtyping, yielding better understanding of the complex biology. Current multi-omics integration tools and methods for cancer subtyping often suffer challenges of high computational efficiency as well as the problem of weight assignment on data types.ResultsHere, we present an efficient multi-omics integration via weighted affinity and self-diffusion (MOSD) to dissect cancer heterogeneity. MOSD first construct local scaling affinity on each data type and then integrate all affinities by weighted linear combination, followed by the self-diffusion to further improve the patients’ similarities for the downstream clustering analysis. To demonstrate the effectiveness and usefulness for cancer subtyping, we apply MOSD across ten cancer types with three measurements (Gene expression, DNA methylation, miRNA).ConclusionsOur approach exhibits more significant differences in patient survival and computationally efficient benchmarking against several state-of-art integration methods and the identified molecular subtypes reveal strongly biological interpretability. The code as well as its implementation are available in GitHub: https://github.com/DXCODEE/MOSD.

Breast cancer is one of the most frequently occurring female cancer types and represents a major cause of death among women worldwide. Breast cancer is heterogeneous in both molecular characteristics and clinical outcomes for its different molecular subtypes. High-throughput technologies facilitated the fast accumulations of the multiple Omic data for cancer patients. These data sources posed a computational challenge for the efficient integrated multi-Omic analysis. The existing studies usually investigated the differential representation or machine learning problems using a single type of Omic data. This study hypothesized that different Omic types contributed complementary information to each other, and their integrated analysis may improve the single-Omic models. An efficient logistic regression-based multi-Omic integrated analysis method (ELMO) was proposed to integrate the RNA-seq and DNA methylation data to detect the breast cancer intrinsic subtypes. ELMO achieved the highest accuracy with a smaller number of features compared with the existing filter and wrapper feature selection methods in this study. The experimental data supported our hypothesis that multi-Omic models outperformed the single-Omic ones.

Despite the improved prognostic relevance of the 2016 WHO molecular-based classification of lower-grade gliomas, variability in clinical outcome persists within existing molecular subtypes. Our aim was to determine prognostically significant metrics on preoperative MR imaging for lower-grade gliomas within currently defined molecular categories. We undertook a retrospective analysis of 306 patients with lower-grade gliomas accrued from an institutional data base and The Cancer Genome Atlas. Two neuroradiologists in consensus analyzed preoperative MRIs of each lower-grade glioma to determine the following: tumor size, tumor location, number of involved lobes, corpus callosum involvement, hydrocephalus, midline shift, eloquent cortex involvement, ependymal extension, margins, contrast enhancement, and necrosis. Adjusted hazard ratios determined the association between MR imaging metrics and overall survival per molecular subtype, after adjustment for patient age, patient sex, World Health Organization grade, and surgical resection status. For isocitrate dehydrogenase (IDH) wild-type lower-grade gliomas, tumor size (hazard ratio, 3.82; 95% CI, 1.94-7.75; P < .001), number of involved lobes (hazard ratio, 1.70; 95% CI, 1.28-2.27; P < .001), hydrocephalus (hazard ratio, 4.43; 95% CI, 1.12-17.54; P = .034), midline shift (hazard ratio, 1.16; 95% CI, 1.03-1.30; P = .013), margins (P = .031), and contrast enhancement (hazard ratio, 0.34; 95% CI, 0.13-0.90; P = .030) were associated with overall survival. For IDH-mutant 1p/19q-codeleted lower-grade gliomas, tumor size (hazard ratio, 2.85; 95% CI, 1.06-7.70; P = .039) and ependymal extension (hazard ratio, 6.34; 95% CI, 1.07-37.59; P = .042) were associated with overall survival. MR imaging metrics offers prognostic information for patients with lower-grade gliomas within molecularly defined classes, with the greatest prognostic value for IDH wild-type lower-grade gliomas.

Intrahepatic cholangiocarcinoma (iCCA) is the second most prevalent primary liver cancer, characterized by insidious onset and high malignancy. Many patients are diagnosed at an inoperable stage, and the effectiveness of chemotherapy and radiotherapy remains limited. This study aimed to provide a comprehensive review of the histological classification, genetic alterations, molecular subtypes, and corresponding imaging signatures of iCCA, highlighting its heterogeneity and offering insights into targeted therapy and personalized treatment. The heterogeneity of iCCA poses significant challenges to both targeted therapy and immunotherapy, necessitating in-depth exploration at the molecular and subtyping levels. Investigating genetic variations, signaling pathway alterations, and molecular subtypes can aid in patient stratification. Stratifying iCCA patients allows for more precise treatment selection, ultimately improving survival outcomes. Imaging, as a non-invasive tool, holds substantial potential for predicting subtypes and molecular profiles. It is possible to infer histological and molecular features from imaging, or to interpret imaging signatures in light of known histological and molecular data. This integrative approach, combining external imaging with internal molecular insights, fosters a comprehensive understanding of iCCA's characteristics and enhances clinical management.

Breast cancer comprises several molecular subtypes with distinct clinical features and treatment responses, and a substantial portion of each subtype remains incurable. A comprehensive analysis of multi-omics data and clinical profiles is required in order to better understand the biological complexity of this cancer type and to identify new prognostic and therapeutic markers. Thus, there arises a need for useful analytical tools to assist in the investigation and clinical management of the disease. We developed Cancer Target Gene Screening (CTGS), a web application that provides rapid and user-friendly analysis of multi-omics data sets from a large number of primary breast tumors. It allows the investigation of genomic and epigenomic aberrations, evaluation of transcriptomic profiles and performance of survival analyses and of bivariate correlations between layers of omics data. Notably, the genome-wide screening function of CTGS prioritizes candidate genes of clinical and biological significance among genes with copy number alteration, DNA methylation and dysregulated expression by the integrative analysis of different types of omics data in customized subgroups of breast cancer patients. These features may help in the identification of druggable cancer driver genes in a specific subtype or the clinical condition of human breast cancer. CTGS is available at http://ctgs.biohackers.net.

Breast cancer is a heterogeneous and polygenic disease that can be divided into different molecular subtypes based on histological and genomic features. To date, numerous susceptibility loci of breast cancer have been discovered by genome-wide association studies and may expand the genetic features. However, few loci have been further studied according to molecular subtypes. We genotyped 23 recently discovered single nucleotide polymorphisms using the Sequenom iPLEX platform in a female Chinese cohort of 3,036 breast cancer patients (2,935 samples matched molecular subtypes) and 3,036 healthy controls. Through a stratification analysis, 5q11.2/MAP3K1 (rs16886034, rs16886364, rs16886397, rs1017226, rs16886448) and 7q32.3/LINC-PINT (rs4593472) were associated with Luminal A, and 10q26.1/FGFR2 (rs35054928) was associated with Luminal B. In our study, breast cancer-specific molecular subtype-associated susceptibility loci were confirmed in Chinese Han women, which contributes to a better genetic understanding of breast cancer in different molecular subtypes. To date, genome-wide association studies have identified more than 90 susceptibility loci associated with breast cancer. However, few loci have been further studied according to molecular subtype. The results of this study are that breast cancer-specific molecular subtype-associated susceptibility loci were confirmed in Chinese Han women, which contributes to a better genetic understanding of breast cancer in different molecular subtypes.

With the advent of high throughput genome-wide assays it has become possible to simultaneously measure multiple types of genomic data. Several projects like TCGA, ICGC, NCI-60 has generated comprehensive, multi-dimensional maps of the key genomic changes like MiRNA, MRNA, proteomics etc. from cancer samples[2,4]. These genomic data can be used for classifying tumour types[5]. Integrative analysis of these data from multiple sources can potentially provide additional biological insights, but methods to do any such analysis are lacking. One of the widely used solutions to handle high dimension data is by removing redundant information in the integrated sample. Most of the expressed genes are overlapped and can be projected onto lower dimension, and then be used to classify different tumor types, without the loss of any/much information. Sufficient dimension reduction (SDR) [1], a supervised dimension reduction approach, can be ideal to achieve such a goal. In this paper, we propose a novel integrative SDR method that can reduce dimensions of multiple data types simultaneously while sharing common latent structures to improve prediction and interpretation. In particular, we extend the sliced inverse regression (SIR) technique, a major SDR method, to integrate multiple omits data for simultaneous dimension reduction. SIR is a supervised dimension reduction method that assumes that the outcome variable Y depends on the predictor variable X through d unknown linear combinations of the predictor[3]. The predictor variable is replaced by its projection into a lower dimension subspace of the predictor space without the loss of information. The aim is to find the intersection of all the subspaces δ called the central susbspace (CS) of the predictor space satisfying the property Y ╨ X| Pδ X. To integrate multiple types of data, we propose and implement a new integrative sufficient dimension reduction method extending SIR[3], called integrative SIR. The main idea is that we take into account all the multi-omics data information simultaneously while finding a basis matrix for each data type with some sharing latent structures. Finally, we get d dimension data which is much smaller than the original data dimension. The reduced dimension d was achieved by cross validation. To demonstrate the integrated analysis of multi-omics data, we applied and compared conventional SIR and integrative SIR to analyze MRNA, MiRNA and proteomics expression profile of a subset of cell lines from the NCI-60 panel. The data used is taken from [6]. The outcomes we have to classify are CNS, Leukemia and Melanoma tumor types. We pre-screened 400 variables from each data type with the criteria of high variance. To find classification error, we performed random forest classification after we applied to each method with leave-one-out cross-validation. As a result, we found out that integrative SIR leads to less classification error as compared to conventional SIR. To summarize, we proposed a new integrative SIR method, a supervised dimension reduction technique for integrative analysis of multi-omics data types. Unlike conventional SDR methods, the new approach can reduce the dimensions of multiple omics data simultaneously while sharing common latent structures across data types without losing any information in prediction. By efficiently capturing the common information, our numerical study shows that integrative SIR classifies tumor types more accurately as compared to conventional SDR methods.

11090 Background: Classification into molecular subtypes is important for the selection of therapy for patients with early breast cancer. Here we determine rates of pathological complete response (pCR) in early stage breast cancer to neoadjuvant capecitabine plus docetaxel, +/- trastuzumab, and investigate MammaPrint together with the molecular subtyping profile BluePrint as markers of pathological response in comparison to other biomarkers. Methods: This analysis was carried out on data from 122 patients enrolled in a multicenter study (XeNA) of neoadjuvant therapy for four 21-day cycles with capecitabine 825 mg/m2 plus docetaxel 75 mg/m2if HER2-, and a standard trastuzumab dose if HER2+ (Glück , BCRT 2011). Clinical and pathological features, TP53 mutation analysis and PAM50 results were collected through GEO at NCBI (GSE22358). MammaPrint and BluePrint outcomes were determined from the available gene expression data and resulted in 4 distinct molecular groups: Luminal A (MammaPrint Low Risk/Luminal-type), Luminal B (MammaPrint High Risk/ Luminal-type), Basal-type and HER2-type. Results: In patients who completed 4 cycles of chemotherapy and surgery the overall pCR rate was 16%. Stratified by BluePrint pCR was observed in 1/15 (7%) of the Luminal A and 2/44 (5%) of Luminal B, in 10/22 (45%) of the HER2-type and in 7/41 (17%) of the Basal-type. The response rate among TP53 mutated patients was 6/61 (26%), which was significantly higher than TP53 wild-type patients (3/54 4%; p=0.012). Concordance of BluePrint/MammaPrint with PAM50 molecular subtyping was 61%. Conclusions: Molecular Subtyping with BluePrint and MammaPrint can identify better outcomes of patients in the neo-adjuvant setting. Patients with Luminal A breast cancer have a good baseline prognosis with excellent survival and may not benefit from chemotherapy (Glück, SABCS 2013). MammaPrint and BluePrint provide predictive information for patients treated with treated with docetaxel-capecitabine +/- trastuzumab.

A novel method for integrating multi-omics data, including gene expression, copy number variation, DNA methylation, and miRNA data, is proposed to identify biomarkers of cancer prognosis. First, survival analysis was performed for these four types of omics data to obtain survival-related genes. Next, survival-related genes detected in at least two types of omics data were selected as candidate genes. The four types of omics data only composed of candidate genes were subjected to dimension reduction using an autoencoder to obtain a one-dimensional data representation. The mRMR algorithm was used to screen for key genes. This method was applied to lung squamous cell carcinoma and 20 cancer-related genes were identified. Gene function analysis revealed that the genes were related to cancer. Using survival analysis, the genes were verified to distinguish between high- and low-risk groups. These results indicate that the genes can be used as biomarkers for cancer.

Background: Epithelial ovarian cancer (EOC) affects nearly 22,000 women annually and is the leading cause of death from gynecologic cancer in the United States. Five year cure rates are &amp;lt;40% and approximately 14,000 will die each year. Large-scale efforts are currently underway to use molecular profiling via next generation sequencing (NGS) technology to guide treatment in cancer patients with poor prognosis, but limited application of NGS in ovarian cancer has been reported. In this report, our previously described Ex Vivo 3D Drug Response Profiling was used to identify response differences between newly diagnosed and relapsed ovarian cancer patients and correlated with NGS of primary tissue. Materials &amp; Methods: Processing of Ovarian Cancers: Under informed consent, ovarian cancer samples were obtained and processed using standard mincing &amp; digestion. 3D spheroids were developed and 3D perfused Ovarian Microtumors were cultured using the 3DKUBE™. Ex Vivo Testing &amp; Analysis: Cultured cells were exposed to clinically relevant concentrations of cytotoxic or targeted agents. Relative IC50s and total percent inhibition were used for ranking compounds. Isolated DNA was sequenced in a CLIA laboratory using a 37-gene NGS panel (GeneTrails®) on the Ion Torrent PGM. Results: Tissues from both newly diagnosed, treatment naive subjects and relapsed subjects were obtained and processed after IRB-approved tissue consent was provided. Spheroid formation was uniform across all malignant tumor types. There was a statistical difference for 3D spheroids treated with Carboplatin formed from either naïve or relapse tissue. The relapse samples had a significantly higher median IC50 than did the naïve samples (70.8 vs. 17.4). This significant difference was not apparent in matched 2D treatment groups. Gemcitabine response varied across tissue type, but did not correlate with traditional biomarkers (i.e. hENT mRNA expression). NGS testing turn-around time was a median of 9 days (range 7-14). Ovarian 3D microtumors were successfully perfuse and tested with targeted agents guided by NGS results, as evidence by a tumor with a mutation of EGFR (p.P265T, clinical significance unknown) with both erlotinib and afatanib demonstrating activity (3.3uM and 0.7uM, respectively). Conclusions: EV3D DRP successfully differentiates carboplatin response and 3D perfusion of microtumors permits ex vivo culture of primary ovarian samples for genotypic-phenotypic response determination. Clinical response data and clinical correlation is ongoing. EV3D DRP permits phenotypic drug response correlation with molecular profiling in real time and may be a clinically relevant functional assay for driver mutation identification and maximal patient response to targeted agent(s). Citation Format: Tessa Desrochers, Stephen Shuford, Christina Mattingly, Lillia Holmes, Matt Gevaert, Jeff Elder, David Orr, Christopher Corless, Larry Puls, Hal E. Crosswell. Ex vivo 3d drug response profiling of primary human ovarian cancer differentiates treatment-naive and relapsed patients and molecular subtypes. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr LB-282. doi:10.1158/1538-7445.AM2015-LB-282

BackgroundOne major goal of large-scale cancer omics study is to identify molecular subtypes for more accurate cancer diagnoses and treatments. To deal with high-dimensional cancer multi-omics data, a promising strategy is to find an effective low-dimensional subspace of the original data and then cluster cancer samples in the reduced subspace. However, due to data-type diversity and big data volume, few methods can integrative and efficiently find the principal low-dimensional manifold of the high-dimensional cancer multi-omics data.ResultsIn this study, we proposed a novel low-rank approximation based integrative probabilistic model to fast find the shared principal subspace across multiple data types: the convexity of the low-rank regularized likelihood function of the probabilistic model ensures efficient and stable model fitting. Candidate molecular subtypes can be identified by unsupervised clustering hundreds of cancer samples in the reduced low-dimensional subspace. On testing datasets, our method LRAcluster (low-rank approximation based multi-omics data clustering) runs much faster with better clustering performances than the existing method. Then, we applied LRAcluster on large-scale cancer multi-omics data from TCGA. The pan-cancer analysis results show that the cancers of different tissue origins are generally grouped as independent clusters, except squamous-like carcinomas. While the single cancer type analysis suggests that the omics data have different subtyping abilities for different cancer types.ConclusionsLRAcluster is a very useful method for fast dimension reduction and unsupervised clustering of large-scale multi-omics data. LRAcluster is implemented in R and freely available via http://bioinfo.au.tsinghua.edu.cn/software/lracluster/.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-015-2223-8) contains supplementary material, which is available to authorized users.

Undifferentiated and dedifferentiated endometrial carcinomas are rare and clinically aggressive variants of the disease. We sought to define the molecular subtypes and genetic alterations affecting cancer-related genes of these rare histologic endometrial cancer types. Patients with undifferentiated/dedifferentiated endometrial cancer subjected to clinical tumor-normal panel sequencing between January 1, 2014, to June 1, 2023, were retrospectively identified, and relevant demographic and clinicopathologic data were extracted from medical records. All cases underwent central pathology review. Endometrial carcinomas with mixed histology and synchronous tumors were excluded. Genomic data including somatic mutations, copy number alterations, and microsatellite instability (MSI), in addition to immunohistochemistry results, were extracted and utilized for molecular subtyping. A total of 35 patients met inclusion criteria, with a median age at diagnosis of 60 years (range, 36 to 85). Of these, 16 (46%) were undifferentiated and 19 (54%) dedifferentiated, with undifferentiated being more frequently of International Federation of Obstetrics and Gynecology (FIGO) 2009 stage IV at diagnosis than dedifferentiated (7/16, 40% vs 3/19, 17%, p = .05). All 4 molecular subtypes were present, with the majority being MSI-high/mismatch repair-deficient (n = 25, 71%); the remaining cases were of POLE molecular subtype (n = 2, 6%), copy number (CN)-high/TP53abnormal (n = 3, 9%), and CN-low/no specific molecular profile (n = 5, 14%). The most recurrent genetic alterations were found in PTEN (26/35, 74%), ARID1A (26/35, 74%), PIK3CA (21/35, 60%) ARID1B (14/35, 40%), and SMARCA4 (9/35, 26%). Pathogenic mutations in chromatin remodeling genes, including ARID1A and ARID1B, were absent in endometrial cancers of CN-high/TP53abnormal subtype. Undifferentiated/dedifferentiated endometrial cancers are heterogeneous at the molecular level; however, the majority are MSI-high/mismatch repair-deficient, which may have therapeutic implications. Loss-of-function alterations in chromatin remodeling genes were present in all molecular subtypes except CN-high/TP53 abnormal undifferentiated/dedifferentiated endometrial cancers.

Purpose:Endometrioid ovarian carcinoma (ENOC) is the second most-common type of ovarian carcinoma, comprising 10%–20% of cases. Recently, the study of ENOC has benefitted from comparisons to endometrial carcinomas including defining ENOC with four prognostic molecular subtypes. Each subtype suggests differential mechanisms of progression, although tumor-initiating events remain elusive. There is evidence that the ovarian microenvironment may be critical to early lesion establishment and progression. However, while immune infiltrates have been well studied in high-grade serous ovarian carcinoma, studies in ENOC are limited.Experimental Design:We report on 210 ENOC, with clinical follow-up and molecular subtype annotation. Using multiplex IHC and immunofluorescence, we examine the prevalence of T-cell lineage, B-cell lineage, macrophages, and populations with programmed cell death protein 1 or programmed death-ligand 1 across subtypes of ENOC.Results:Immune cell infiltrates in tumor epithelium and stroma showed higher densities in ENOC subtypes with known high mutation burden (POLEmut and MMRd). While molecular subtypes were prognostically significant, immune infiltrates were not (overall survival P > 0.2). Analysis by molecular subtype revealed that immune cell density was prognostically significant in only the no specific molecular profile (NSMP) subtype, where immune infiltrates lacking B cells (TILB minus) had inferior outcome (disease-specific survival: HR, 4.0; 95% confidence interval, 1.1–14.7; P < 0.05). Similar to endometrial carcinomas, molecular subtype stratification was generally superior to immune response in predicting outcomes.Conclusions:Subtype stratification is critical for better understanding of ENOC, in particular the distribution and prognostic significance of immune cell infiltrates. The role of B cells in the immune response within NSMP tumors warrants further study.