Structured Matrix Completion with Applications to Genomic Data Integration

Matrix completion is the task of predicting missing entries of matrix from a subset of known entries. Notions of structured matrix completion include any setting in which whether an entry is observed does not occur uniformly at random. In recent work, a modification to the standard nuclear norm minimization for matrix completion has been made to take into account sparsity-based structure in the missing entries, which is motivated e.g. in recommender systems. In this work, we propose adjusting an Iteratively Reweighted Least Squares (IRLS) algorithm for low-rank matrix completion to take into account sparsity-based structure. We also outline an iterative gradient-projection-based implementation of the algorithm that can handle large-scale matrices. Lastly, we present preliminary numerical experiments showcasing the performance of the proposed method compared to the standard IRLS algorithm in structured settings.

We propose a low rank structured matrix completion algorithm for image inpainting problems originated from scanning microscopy. The proposed method exploits the annihilation property observed in Gaussian Markov Random Field (GMRF) or partial differential equation (PDE)-based inpainting approaches. By utilizing the commutative property of the convolution, the annihilation property is embodied into rank-deficient block Hankel structure data matrices and the image inpainting problem is converted into low-rank structured matrix completion problem. To solve the structured low-rank matrix completion problem, an alternating direction method of multiplier (ADMM) method is used with factorization matrix initialization using the low rank matrix fitting (LMaFit) algorithm. Experimental results showed that the proposed method outperforms the existing state-of-the-art image inpainting methods.

This paper considers the matrix completion problem. We show that it is not necessary to assume joint incoherence, which is a standard but unintuitive and restrictive condition that is imposed by previous studies. This leads to a sample complexity bound that is order-wise optimal with respect to the incoherence parameter (as well as to the rank $r$ and the matrix dimension $n$ up to a log factor). As a consequence, we improve the sample complexity of recovering a semidefinite matrix from $O(nr^{2}\log^{2}n)$ to $O(nr\log^{2}n)$, and the highest allowable rank from $\Theta(\sqrt{n}/\log n)$ to $\Theta(n/\log^{2}n)$. The key step in proof is to obtain new bounds on the $\ell_{\infty,2}$-norm, defined as the maximum of the row and column norms of a matrix. To illustrate the applicability of our techniques, we discuss extensions to SVD projection, structured matrix completion and semi-supervised clustering, for which we provide order-wise improvements over existing results. Finally, we turn to the closely-related problem of low-rank-plus-sparse matrix decomposition. We show that the joint incoherence condition is unavoidable here for polynomial-time algorithms conditioned on the Planted Clique conjecture. This means it is intractable in general to separate a rank-$\omega(\sqrt{n})$ positive semidefinite matrix and a sparse matrix. Interestingly, our results show that the standard and joint incoherence conditions are associated respectively with the information (statistical) and computational aspects of the matrix decomposition problem.

Genomic data integration--the process of statistically combining diverse sources of information from functional genomics experiments to make large-scale predictions--is becoming increasingly prevalent. One might expect that this process should become progressively more powerful with the integration of more evidence. Here, we explore the limits of genomic data integration, assessing the degree to which predictive power increases with the addition of more features. We focus on a predictive context that has been extensively investigated and benchmarked in the past-the prediction of protein-protein interactions in yeast. We start by using a simple Naive Bayes classifier for integrating diverse sources of genomic evidence, ranging from coexpression relationships to similar phylogenetic profiles. We expand the number of features considered for prediction to 16, significantly more than previous studies. Overall, we observe a small, but measurable improvement in prediction performance over previous benchmarks, based on four strong features. This allows us to identify new yeast interactions with high confidence. It also allows us to quantitatively assess the inter-relations amongst different genomic features. It is known that subtle correlations and dependencies between features can confound the strength of interaction predictions. We investigate this issue in detail through calculating mutual information. To our surprise, we find no appreciable statistical dependence between the many possible pairs of features. We further explore feature dependencies by comparing the performance of our simple Naive Bayes classifier with a boosted version of the same classifier, which is fairly resistant to feature dependence. We find that boosting does not improve performance, indicating that, at least for prediction purposes, our genomic features are essentially independent. In summary, by integrating a few (i.e., four) good features, we approach the maximal predictive power of current genomic data integration; moreover, this limitation does not reflect (potentially removable) inter-relationships between the features.

BACKGROUND Integrating genomic data into electronic health records (EHRs) and clinical decision support (CDS) systems marks the beginning of a new era of precision medicine. Leveraging tailored insights from patient genomic data opens the door for patient care to have enhanced diagnosis, personalized treatment plans, and optimized medication management. OBJECTIVE This systematic review aims to examine the evolution of genetic data integration into Electronic Health Records by analyzing past developments, current implementation strategies, and future opportunities. It seeks to review historical milestones, identifying key advancements in genomic data collection, storage, and early integration efforts. It assesses current clinical practice and evaluates existing technologies and policies that shape the standards of genomic data interoperability. It also aims to highlight the technical, ethical, and regulatory challenges that hinder effective genomic integration and propose potential strategies and recommendations to address them. METHODS This systematic review follows PRISMA 2020 guidelines to ensure transparency and reproducibility. A comprehensive search was completed on the PubMed database with a predefined query and relevant inclusion and exclusion criteria. RESULTS A total of 48 articles were reviewed, with 27 included in the final review. Identified themes included technical integration, clinical utility, precision medicine, and human genome sequencing. These findings emphasize the role of AI, multi-omics data, APIs, and current work in maximizing the role of genetic data in clinical practice through integration in Electronic Health Records. CONCLUSIONS The integration journey is riddled with challenges such as data standardization, interoperability, data privacy, and resource constraints, which hinder the effective utilization of genomic information. This systematic review explores historical milestones, current practices, and future implementations in genomic data integration. It emphasizes the potential of emerging technologies like artificial intelligence, multi-omics data, next-generation sequencing, APIs, and automation to bridge existing gaps. Genomic integration can revolutionize healthcare delivery by harnessing these technologies, ensuring equitable and improved patient outcomes.

Background: Epithelial ovarian cancer (EOC) is the second leading cause of death from a gynecologic malignancy in the U.S. The overall survival rate has not changed over the last 30 years thereby necessitating new therapeutic approaches. Targeted therapeutics that have come to the forefront of exploration such as anti-VEGF and anti-EGFR, have resulted in only marginal improvements in overall survival. We have used a functional genomics approach followed by analyzing TCGA (The Cancer Genome Atlas) ovarian dataset to help identify significant potential molecular targets that represent key points of therapeutic intervention. Methods: We employed an unbiased high-throughput lethality screen using a 24,088 siRNA library targeting 6,022 “druggable” genes and tested for effects on growth and/or survival of A1847 EOC cells. Candidate hits from the primary screen were rescreened in a secondary screen across six additional EOC cell lines (OVCAR5, SKOV3, A2780, C30, CP70 and UPN275), and three non-tumorigenic, immortalized human ovarian surface epithelial (HOSE) cell lines (HIO-80, HIO-117, HIO-120). We further validated and eliminated the leading hits which may be due to potential off-target effects by rescreening individual siRNAs from the initial siRNAs pools and by performing quantitative RT-PCR and western blotting. Mechanistic exploration with each of the lead candidates was performed in A2780 cells using Guava cell cycle and Nexin assays. Clinical datasets for lead candidate genes were obtained from https://tcga-data.nci.nih.gov/tcga/ (n=494) and analyzed for gene expression, DNA methylation, copy number variation (CNV) and survival (Kaplan-Meier analysis). Results: Building on the concept of “oncogenic addiction” we performed a primary lethality screen using A1847 cells and identified 300 genes that were considered essential for tumor cell growth and/or survival. These hits were defined as a gene that reduced cell viability by >15% with a false discovery rate of ≤5%. In the secondary screen, each of the ten cell lines exhibited a unique number of hits ranging from 27-94% of the total siRNAs screened. Fifty-three gene candidates were found to exhibit effects in all seven tumorigenic cell lines. Extensive validation of these hits refined the list to four top priority candidates (HSPA5, NDC80, NUF2 and PTN) that reduced the viability of all the EOC cell lines with minimal effects on the non-tumorigenic HOSE lines. In A2780 cells, HSPA5, NDC80, NUF2 and PTN siRNA treatment led to an average 2.5-fold (n=3, p<0.05) increase in apoptotic cells as compared to control siRNA. Analysis of TCGA ovarian dataset showed NDC80, NUF2 and PTN to be over-expressed (>1.5-fold increase compared with normal tissues) in 99.8%, 99.8% and 40% of patients, respectively. CNV analysis for these three genes showed that the samples with copy number gain exhibited an average 1.6-fold increase (p-value <0.0001, <0.0001 and <0.05 for NDC80, NUF2 and PTN, respectively) in gene expression as compared to samples with no copy number gain. HSPA5 was not found to be significantly over-expressed in ovarian tumors as evaluated by TCGA. DNA methylation analysis showed promoter regions of NDC80, NUF2 and PTN to be hypomethylated (beta values <0.25) in 94% of the tumor samples. KM-survival analysis showed that higher expression of NUF2 mRNA was related to poor prognosis in ovarian cancer patients. Conclusion: Our functional studies integrated with genomics data provide an important, unbiased avenue towards addressing an urgent and unmet clinical need for identification of prospective therapeutic targets for drug discovery and treatment of ovarian cancer. Citation Format: Geetika Sethi, Harsh Pathak, Hong Zhang, Yan Zhou, Margaret B. Einarson, Vinod Vathipadiekal, Sumedha Gunewardena, Michael J. Birrer, Andrew K. Godwin. Integration of synthetic lethality and genomic data to identify significant molecular vulnerabilities in ovarian cancer. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Synthetic Lethal Approaches to Cancer Vulnerabilities; May 17-20, 2013; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(5 Suppl):Abstract nr A06.

Emerging data support the hypothesis that both inherited and somatic alterations of genomes of African American (AA) prostate cancer (CaP) patients are significantly different from that of Caucasian American (CA) men. Recent studies from our and other groups have established that frequencies of the most common genomic alteration, TMPRSS2-ERG fusion, which apparently causes the high expression of ERG oncoprotein, is markedly lower in AA than in CA patients. Some works indicated that the frequency of CaP ERG fusion is the lowest in Asian patients, but this frequency changed from 11% to 78% in different Asian cohorts. Further analysis in Asian men is necessary. In order to systematically and comprehensively address this and related problems, Uniformed Services University of the Health Sciences-Center for Prostate Disease Research (CPDR), the US National Cancer Institute (NCI) and CAS-MPG Partner Institute for Computational Biology (PICB), Shanghai Institutes for Biological Sciences (SIBS), Chinese Academy of Sciences (CAS), China, launched an international collaboration project. The overall goal of the project is to evaluate difference of driver gene mutations of prostate cancer in various ethnic groups by integrating genomic and transcriptomic data from CPDR, TCGA and the Chinese Prostate Cancer Consortium (CPCC), specially focusing on somatic alterations (ERG and others), microbiome, mitochondrial sequences and microRNAs. We integrated gene fusion, copy number, and gene expression data to detect potential driver genes and, as anticipated ERG alteration was one of the most commonly detected. The frequency of ERG fusion is the highest for CA, intermediate for AA and the lowest for Asian men, while its expression is significantly higher in ERG-fusion tumors than other tumors cross over all of the three ethnic groups. We did find some genes altered more frequently in Asian patients than in American patients, subject to further validations. We explored the methodology for microbiota profiling using prostate cancer WGS sequencing data. To this date, we have developed an analysis pipeline for the detection of microbiome from prostate tumor tissues and the corresponding blood samples with high confidence but inadequate sensitivity likely due to the low availability of both microbial sequences and bacterial reference genomes. Our preliminary analyses highlight the potential of international collaboration. CPDR-PICB-NCI will work on a seamless approach for data integration, particularly of the raw data to truly enhance CaP diagnosis and treatment by providing more informative biomarker and therapy targets relevant to ethnically diverse patient populations. Citation Format: Qingyu Xiao, Yidi Sun, Hong Li, Yixue Li, Guo-Ping Zhao, Wendy Wang, Sudhir Srivastava. Collaborative analyses for delineating mutation variations among different ethnic patients of prostate cancer based on genomic data integration [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 5550. doi:10.1158/1538-7445.AM2017-5550

Screening for Familial Ovarian Cancer: A Ray of Hope and a Light to Steer by

In spite of various treatment options currently available, ovarian cancer (OC) still remains a leading cause of death in women world-wide. Diagnosis at an early stage is one of the most important factors that determines survival. Current clinical diagnostic tools have, however, a limited efficacy in early OC detection. Therefore, there is a critical need for new (early) diagnostic biomarkers and tools. Through advances in genomic, proteomic and metabolomic techniques, several novel molecular OC biomarkers have recently been identified. These biomarkers are currently subject to validation. In addition, integration of genomic, proteomic and metabolomic data, in conjunction with epidemiologic and clinical data, is considered essential for obtaining useful results. Interesting recent work has already shown that specific diagnostic biomarkers, such as BRCA mutations, may have profound therapeutic implications. Here, we review the current state of OC research through literature and database searches, with a focus on various recently identified biomarkers via different technologies for the (early) diagnosis, prognosis and treatment of OC. Multi-biomarker panels accompanied by a meticulous determination of their sensitivity and specificity, as well their validation, using multivariate analyses will be critical for its clinical application, including early OC detection and tailor-made OC treatment.

The emerging era of genomic data integration for analyzing splice isoform function

The tumor microenvironment is increasingly recognized as an active participant in tumor progression. A recent pan-cancer genomic profile analysis has revealed that gene signatures representing components of the tumor microenvironment are robust predictors of survival. A stromal gene signature representing fibroblasts and extracellular matrix components has been associated with good survival in diffuse large B-cell lymphoma (DLBCL). Paradoxically, a closely related gene signature has been shown to correlate with poor survival in carcinomas, including breast, ovarian, pancreatic, and colorectal cancer. To date, there has been no explanation for this paradoxical inverse correlation with survival outcomes in DLBCL and carcinomas. Using public gene data sets, we confirm that the DLBCL stromal gene signature is associated with good survival in DLBCL and several other B-cell lymphomas while it is associated with poor survival in ovarian cancer and several other solid tumors. We show that the DLBCL stromal gene signature is enriched in lymphoid fibroblasts in normal lymph nodes and in cancer-associated fibroblasts (CAFs) in ovarian cancer. Based on these findings, we propose several possible mechanisms by which CAFs may contribute to opposite survival outcomes in B-cell lymphomas and carcinomas.

BACKGROUND: Several lifestyle factors have consistently been associated with survival of ovarian cancer, including body mass index, analgesic medications, physical activity, and smoking. Menopausal hormone therapy (HT), which can comprise of estrogen only (ET) or a combination of estrogen and progestin (EPT), has also been studied for its impact on ovarian survival. Randomized trials as well as a recent meta-analysis have shown better survival for women who took HT after their diagnosis in comparison to women who did not (HR=0.67, 95% CI 0.47-0.97). However, the relationship between use of HT prior to ovarian cancer diagnosis and survival is less clear. METHODS: Approximately 4,700 ovarian cancer patients across nine population-based studies participating in the Ovarian Cancer Association Consortium (OCAC) were included in our analysis. Information on HT use prior to diagnosis and potential confounders was obtained via phone or in-person interviews. To estimate the effect of pre-diagnostic HT use on ovarian cancer survival, hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated using Cox proportional hazards regression models with left and right truncation to account for different observation times. Effects by type and duration of HT use as well as disease histotype were also evaluated. RESULTS: Women who used HT before diagnosis had an 11% decreased risk of death in comparison to those who did not use HT (95% CI 0.82-0.97). Similar effects were observed for both types of HT (HR=0.86, 95% CI 0.75-0.98 for ET and HR=0.89, 95% CI 0.80-0.99 for EPT). In addition, there appeared to be a duration effect with longer HT use conferring better survival (p-trend=0.07 for ET, p-trend=0.003 for EPT). When histotype was considered, the overall decreased risk was driven by serous and mucinous ovarian cancers (HR=0.85, 95% CI 0.77-0.93 and HR=0.54, 95% CI 0.28-1.07, respectively). CONCLUSIONS: Use of HT after diagnosis also affords a survival advantage based on published observational and clinical trials data. We now report that pre-diagnosis HT use may confer a survival advantage among women with ovarian cancer. Interestingly, use of ET prior to diagnosis is associated with increased risk of ovarian cancer, but better survival among women diagnosed with the disease. These observations taken together demonstrate the complex nature of ovarian cancer. Citation Format: Alice W. Lee, Siri Peterson, Ashley Wiensch, Malcolm C. Pike, Celeste L. Pearce on behalf of the Ovarian Cancer Association Consortium. PRE-DIAGNOSIS USE OF MENOPAUSAL HORMONE THERAPY ASSOCIATED WITH BETTER OVARIAN CANCER SURVIVAL [abstract]. In: Proceedings of the 12th Biennial Ovarian Cancer Research Symposium; Sep 13-15, 2018; Seattle, WA. Philadelphia (PA): AACR; Clin Cancer Res 2019;25(22 Suppl):Abstract nr DP-011.

The field of personalized disease prognosis, risk prediction and disease management has always attracted the attention of scientists and clinicians worldwide. Translating current clinical research and emerging genome-scale molecular biomarkers of ovarian and other cancers is yielding promising new insights into predictive signatures and potential clinical targets, moving treatment in favor of more personalized therapeutic approaches. Integrative genomics, big data analysis and the disease prediction modelling have become the major driving factors in understanding how to reproducibly stop a disease for individual patients - a goal of precision medicine. In clinical practice, prognostic and predictive factors could be measured as continuous or discrete variables. Here, we proposed a novel methodology to predict disease/treatment outcome via analysis of the similarities between any pair of patients who are each characterized by a certain set of pre-defined biological variables (biomarkers or clinical features) represented initially as a prognostic binary variable vector (PBVV) and subsequently transformed to a prognostic signature vector (PSV). High-grade serous ovarian carcinoma (HGSC) is the most prevalent ovarian cancer, and is one of the most lethal gynecological diseases in the world today. Many molecular signatures for prognostic and disease prediction of HGSC have been proposed, but no been implemented in clinical practice yet. In our previous study, the highly-confidence prognostic classifier, comprising 36 mRNAs that can stratify the HGSC patients into low, intermediate or high-risk subgroups with significantly distinct overall survival outcomes and sensitivity to post-surgery chemotherapy, has been proposed [1]. In this work, we used this classifier as the discriminative model of prognosis of HGSC patients. We implemented our methods to personalized prognosis of HGSC patients of TCGA database based on our molecular predictor [1]. Our results revealed that patient's age, when converted into discrete binary values, was positively correlated with the overall risk of succumbing to the disease and can be used as independent prognostic factor. The inclusion of age into the molecular survival predictor provided more robust personalized prognosis of overall survival when applied to an independent testing metadata composing 359 HGSC patients. In summary, we propose that our personalized prognosis methods and results after including age as prognostic factor could provide clinical benefits in the context of personalized prognosis and therapeutic assignment of HGSC patients. Our method can be generalized and implemented in many other diseases to accurately predict personalized patients' outcomes. [1]. Tang Z, Ow GS, Thiery JP, Ivshina AV and Kuznetsov VA. Int J Cancer. 2013; 134(2):306-318. Citation Format: Ghim Siong Ow, Zhiqun Tang, Anna Vladimirovna Ivshina, Vladimir Andreevich Kuznetsov. Big data and computational biology method for personalized prognosis of high-grade serous ovarian carcinoma. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research: Exploiting Vulnerabilities; Oct 17-20, 2015; Orlando, FL. Philadelphia (PA): AACR; Clin Cancer Res 2016;22(2 Suppl):Abstract nr A75.

Plant Genome and Systems Biology (PGSB), formerly Munich Institute for Protein Sequences (MIPS) PlantsDB, is a database framework for the integration and analysis of plant genome data, developed and maintained for more than a decade now. Major components of that framework are genome databases and analysis resources focusing on individual (reference) genomes providing flexible and intuitive access to data. Another main focus is the integration of genomes from both model and crop plants to form a scaffold for comparative genomics, assisted by specialized tools such as the CrowsNest viewer to explore conserved gene order (synteny). Data exchange and integrated search functionality with/over many plant genome databases is provided within the transPLANT project.

Ovarian cancer afflicts approximately 204,000 women worldwide, with ∼125,000/year deaths. In the United States more than 22,000 new cases of ovarian cancer diagnosed with over 14,000 disease-related deaths.[1] The standard of care in ovarian cancer includes expedient access to the health care system, consultation with a gynecologic oncologist, surgical intervention, and multiagent platinum-based chemotherapy. Prognostic factors for improved survival in women with ovarian cancer include younger age; early stage; low-grade and serous histology; good performance status; disease biology (BRCA1 and 2 mutation carriers have a better prognosis); and low volume of residual disease at the time of cytoreductive surgery. In contrast, African-American race, low socioeconomic status (SES), geographic location, and insurance status have been associated with worse survival and increased probability of not receiving appropriate standard of care treatment.[2-7] These differences in ovarian cancer outcomes linked to race, social, and economic factors indicate that health care disparities exist in the treatment of this gynecologic cancer. In this presentation, we will (1) review the standard of care for epithelial ovarian cancer treatment, (2) discuss the variables affecting the outcome for patients with epithelial ovarian cancer, (3) describe the factors contributing to healthcare disparities, and (4) discuss the evidence for clinical disparities in epithelial ovarian cancer care and outcomes. Causes of healthcare disparities are complex and include differences in race/ethnicity, SES, insurance status, education level, geographic location, culture, healthcare system factors (adherence to treatment guidelines, and access to care), and disease biology.[2-7] The following are highlighted in this presentation: Population-based studies have demonstrated that African-American women are at increased risk of ovarian cancer death compared to non-Hispanic Caucasian women.Population-based studies have demonstrated that African-American women are less likely to undergo site-specific surgery or surgical staging for ovarian cancer compared to Caucasian women.African-American women were less likely to seek care at high-volume hospitals or be operated on by high-volume surgeons.African-American women and those with Medicaid/Medicare payer status are less likely to receive standard of care therapy based on National Comprehensive Cancer Network (NCCN) guidelines.Geographic location away from a high-volume hospital is associated with increased risk of non-adherent ovarian cancer care.Geographic barriers to standard treatment disproportionately affect racial minorities and women of low-SES.In contrast to population-based studies, retrospective single academic institution studies and ancillary analysis of large cooperative group phase III trials in advanced ovarian cancer have demonstrated no difference in overall survival in women of African-American and Caucasian descent. The reported research in ovarian cancer indicates that unequal delivery of quality care, obstacles to the delivery of recommended care, and limited access to expert care may account for the disparities seen in ovarian cancer care. The U.S. Department of Health and Human Services has targeted disparities in access to health care as the centerpiece of the Healthy People 2020 campaign and the Department of Health and Human Services (HHS) has unveiled an action plan to reduce racial and ethnic Health disparities.[8, 9] The 5 goals of the HHS Disparities Action Plan are: (1) transform health care; (2) strengthen the nation's Health and Human Services infrastructure and workforce; (3) advance the health, safety, and well-being of the American people; (4) advance scientific knowledge and innovation; and (5) increase the efficiency, transparency, and accountability of HHS programs. The HHS Disparities initiative will specifically target programs to increase knowledge of, access to, and utilization of biomedical and behavioral procedures to reduce cancer disparities.[9] Listed below are future research objectives that will reduce and potentially eliminate health care disparities in ovarian cancer in the United States: Improve our knowledge of the interaction between race, ethnicity, SES, geographic location and ovarian cancer health care disparities across the United States to harmonize existing data and identify opportunities for intervention.Improve our knowledge regarding cultural barriers to care.Promote community-based participatory research to increase ovarian cancer awareness and review standards of care.Develop interventions to provide expedient access to expert care by trained specialists in Gynecology Oncology, cytoreductive surgery, and administration of multiagent platinum-based chemotherapy in order to increase the likelihood of NCCN adherent standard of care therapy.