GenePattern Platform Integrates Genomics Analysis
Adapted from the 2005 Best Practices entry from Broad Institute of MIT and Harvard
In the course of developing software for the microarray analysis community, the Cancer Program Informatics group at the Broad Institute recognized that several bottlenecks were preventing the adoption, use, and distribution of new analytical techniques:
- It took too long to integrate new tools into existing software. Computational biologists were developing new methods, written in their programming language of choice, without regard for interoperability with other tools.
- There was no way to encapsulate complex methodologies. Researchers frequently used many different software tools within an analysis methodology, and there was no way to "glue" these tools together that retained the order in which they were run and the parameters associated with each step.
- Researchers could not reproduce published results without extensive communication with the authors of the original research. In fact, often they could not reproduce their own results, having lost track of the exact parameter settings and code versions they employed.
These considerations led to the development of GenePattern, a software platform that has greatly accelerated the pace of genomic analysis. Use of the GenePattern analysis environment has reduced the time required to integrate new tools into a shared platform from six months to less than one day and allowed authors to encapsulate and share all of the information required for another researcher to reproduce results of in silico analyses. The easy integration of new modules and easy construction of analytic pipelines has led to GenePattern's adoption by other organizations to support support research in metabolomics, proteomics, and analytical chemistry.
GenePattern is a software analysis platform designed to host a variety of diverse tools, manage their interactions, and create reproducible methodologies that can be easily edited and shared. GenePattern accomplishes these goals in two ways:
- A task integration environment provides a simple Web-based form that allows users to add new tools, written in any language, to GenePattern without writing additional code.
- A pipeline environment allows users to chain analysis tasks into complex analytic workflows that capture all the details of a methodology and can then be shared with others.
These environments, along with user interfaces for programming and nonprogramming users, a repository of analysis modules, and other features, have made GenePattern the main microarray analysis platform within the Broad Institute and at many other organizations.
Easy Integration of Bioinformatics Analysis Tools
The computational biologists in the Broad Institute's Cancer Genomics Program have been developing methodologies for analyzing the molecular profiles of cancer. Their algorithms are developed in a variety of languages, such as Java, Perl, C++, R, and MATLAB.
Read More . . . PK/PD Simulation Speeds Decision Making
Adapted from the Pfizer Global R&D entry to the 2005 Best Practice awards
The multimillion-dollar decision to proceed with clinical testing of a new drug often relies on preliminary -- and sometimes contradictory -- evidence of drug effects. Yet decision makers must weigh the compound's potential to compete with established therapies against other candidates vying for the same development resources.
Development teams must combine disparate knowledge and expertise to envision the likely drug profile. They rely on discussion, e-mail, and slide presentations, generally without a comprehensive, quantitative summary of knowns and unknowns. To help decide the fate of a new cardiovascular drug, Pfizer's Pharmacometrics Group built a model-based projection of the competitive landscape. The modeling effort combined internal study data with published trials and FDA submissions on competing compounds to generate an inclusive, topographical map of the drug's likely performance.
The team used a software product called Drug Model Explorer (DMX), developed by Pharsight Corporation, to evaluate and communicate the modeling results. DMX is a software toolkit for sharing modeling and simulation results through interactive plots and tables. In DMX, the team explored clinical effects, and associated uncertainty, for the drug and competitors across multiple endpoints, treatments, and patient populations.
The results showed that the new drug was unlikely to outperform its main competitor; Pfizer discontinued development. The modeling project supported a more confident decision without investment in additional trials and allowed team members to re-deploy to other programs. It provided an enduring, evolving knowledge repository to support future development projects. DMX enabled team members to evaluate complex product tradeoffs and ask targeted what-if questions in real time.
Read More . . .