MultiViewPortal: Towards a Scalable Web Application for Multiview Learning

Abstract Recent technological advances have enabled the production of vast amounts of diverse but related data (e.g. genomics, metabolomics, proteomics) with rich information that offer remarkable opportunities to understand biological processes involved in complex diseases and to transform medicine. It is now widely-recognized that the mechanisms that underlie complex diseases are more likely to be unraveled by approaches that go beyond analyzing each type of data separately. Analyzing these multifaceted data to obtain useful information and knowledge is challenging because the data are complex, heterogeneous, and high-dimensional, and require a considerable level of analytical sophistication. Most existing software for data integration that address some of these analytical challenges are on-premises and tend to be decentralized, limiting ability to perform comprehensive integrative analysis from anywhere and on any device. Further, they are based on languages that require substantive knowledge in programming, which limits their wide-spread adoption by the research community. The few existing web applications for data integration have limited capabilities in the types of analyses that can be performed. As a first step towards providing a comprehensive, centralized data integration platform, we have developed a web application in Shiny R and are hosting the application on shinyapp.io. However, the compute and memory limitations of shinyapps.io severely constrain the problem size that can be executed. We propose to explore new cloud technologies to enhance our existing workflows for integrating data from multiple sources. Our web application, MultiviewPortal, will enable high-throughput analyses of molecular, imaging, and phenotypic data; it will make it straightforward for all users to integrate data and it will facilitate users using data from NIH-funded projects that might otherwise be complicated to retrieve, given the complexity of many data portals. Ultimately, our portal has the potential to narrow the gap from raw molecular data to biological insights, offer opportunity to expand the definition of complex diseases, and allow to stratify patients and identify those who might benefit from targeted interventions.