ADDITIONAL SPEAKERS
Francesca Vitali, PhD - The University of Arizona
Novel and emerging data fusion strategies for integrating health and biomolecular data
Over the last few years, biomedical research and clinical practice have experienced incredible growth in terms of both the amount and variety of data being collected and leveraged for different types of analysis. This represents a great opportunity to increase our knowledge about many biological mechanisms as well as improve the medical process. However, not all big data is created equal, complicating the integration and analysis of such large datasets. For example, clinical record data is highly heterogeneous, sparsely annotated, and contains several measurement types and unstructured text fields comprised of ambiguous statements as well as varying levels of certainty, whereas genomic and imaging data are crisp, homogeneous, densely annotated data with a low cardinality of distinct variables. Nowadays, the development of novel methodologies capable of integrating population health data with biomolecular data is crucial, not only for enabling translational and clinical research, but for developing more effective patient care. However, integrating these data are particularly challenging when the molecular measurements are not conducted on individual subjects. In order to take full advantage of the wide spectrum of biomedical data available, advanced data integration tools need to be developed. In this context, we will discuss novel and emerging data fusion strategies for integrating health and biomolecular data to develop new research hypotheses and conduct predictive and data interpolation operations. These methods include approaches that (i) take into account comprehensive drug-exposure histories of individuals derived from healthcare data, while also including genetic, environmental, and lifestyle variabilities for each individual; (ii) integrate electronic medical records with biobank data to identify new disease pathways; (iii) combine multi-omic profiling with clinical factors from large cohorts; and (iv) perform crisp integration of bimolecular data whilst leveraging population measurements (e.g., counties, medication, diseases).
Lara M Mangravite, PhD - Sage Bionetworks
Open practices to advance biomedicine through data-intensive science
Open science practices in bio-computing have been promoted over the past 10 years under the premise that these approaches can improve confidence and, therefore, speed advancement of biomedical hypotheses stemming from computational research. In that time, we have observed wide adoption of open practices including those focused on open data, open commons(es), open source software, and open access publishing. Although many of these efforts help to establish confidence in research observations amongst computationally-savvy researchers, they often fail to support the wider acceptance necessary to inform trajectories of biological inquiry and/or to promote adoption for use in clinical care. Here, we discuss complementary mechanisms to further support the advancement of biocomputational hypotheses, including those developed using emerging digital health technologies, through the transfer and translation of knowledge across research domains.
Serghei Mangul, PhD - University of California, Los Angeles
Seeing Beyond the Target: Constructing germline research cohorts from clinical tumor sequencing
Tens of thousands of cancer patients have had their tumors sequenced to identify clinically actionable mutations. In addition to saving lives, this activity has produced valuable research data sets leading to significant discoveries in basic and translational domains. However, the targeted nature of clinical tumor sequencing has a limited research scope, especially with respect to germline genetics. In this work, we address this problem by developing a software platform (SBT: Seeing Beyond the Target) that mines discarded tumor sequences to produce rich research level data including genome-wide germline genotypes, T and B cell receptor sequences, rDNA and mtDNA copy number, and HLA types. These features have been demonstrated as potential prognostic indicators in research studies, and our methods now make them available in large-scale clinical cohorts. We validate the accuracy of our tool, by comparison, to deeply sequenced cohorts and show its utility through replication of known genetic associations. We provide a free downloadable cloud implementation and demonstrate its efficiency by constructing the largest germline-somatic cohort produced to date (n>20,000), more than doubling the size of The Cancer Genome Atlas. We believe that SBT will greatly increase the research potential of clinical tumor data sets and provide a bridge between the germline and somatic research communities. SBT is freely available at https://github.com/smangul1/seeing.beyond.target/wiki
Marina Sirota, PhD - Assistant Professor, University of California, San Francisco
Leveraging population level molecular, environmental and clinical data to study adverse pregnancy outcomes
Given the wealth and availability of genomic, clinical and environmental exposure data, computational integrative methods provide a powerful opportunity to identify population-specific determinants of disease. In this talk, I will discuss our efforts to develop computational methods and integrate large-scale genomic, transcriptomic and environmental exposure datasets to elucidate factors that affect preterm birth (PTB). Preterm birth, or the delivery of an infant prior to 37 weeks of gestation, is a major health concern. Infants born prematurely, comprising of about 12% of the US newborns, have elevated risks of neonatal mortality and a wide array of health problems. In our work, we leverage the rich multi-omic, clinical and environmental variation data to advance our understanding of biology of preterm birth as it relates to all populations. Our findings further inform precise population-specific diagnostic and therapeutic strategies bringing us closer to applying precision medicine to this important biomedical problem.
