Infrastructure

University Of Connecticut Sch Of Med/Dnt, Farmington CT

Investigators

Linked publications, trials & patents

Abstract

Infrastructure Project Summary The Infrastructure Core provides for overall software development and compute infrastructure for the Virtual Cell (VCell) modeling and simulation software, as well as supporting research applications that drive the development of the software. The CCAM software development group is a team of two full time software developers, and several part time developers and this core manages the formalized structure for development of the software, including management of the development tools, code repositories, bug tracking protocols, multi-platform builds, etc.. We also develop and manage the server-side infrastructure for the Virtual Cell, which now has been in place for over 17 years, but which has continuously grown in size to cope with increased user base and software features. A scalable set of services are being deployed at our High Performance Computing Facility and we have introduced recently new elements for compatibility with cloud computing protocols and prototype single-instance VM packaging. The infrastructure core also develops and maintains the VCell database (VCDB) that stores all models and model elements, as well as identifying information for compute job status and result location on the shared storage servers. We plan to construct a web application building on a recently developed VCell RESTful API to greatly expand access to the database over the web, expanding the use of the database as a knowledgebase for the entire modeling community. Part of the VCDB development is also the maintenance of the formal VCell XML data structure (VCML) and schema, as well as the updating and improvement of the import/export algorithms and tools for interfacing with community standards SBML, BioPAX, and SED-ML. A completely new requirement for development in the next project period is to create a specialized infrastructure for processing and analysis of multiple run stochastic simulations. The growing number and increase use of stochastic and hybrid solvers in VCell create requirements that go beyond the simple data retrieval of multiple trajectories and single point distribution histograms and scale in fault-tolerant and optimized fashion to support hundreds and thousands of related spatial and non-spatial runs. Last, but not least, this core is tasked with the development and maintenance of the VCell Math testing Framework. Comprised of a growing test Suite mixing random models from database with manually chosen ones (including a full complement of cases with exact solution), it automates the process of end-to-end regression testing that is critical for maintaining forward compatibility and verifying correctness and accuracy of various solvers and algorithms.