Experimental Design Automation

A Novel Solution for the Informatics Bottleneck in Drug R&D Productivity

Summary

Key barriers to R&D productivity in the pharmaceutical industry are the bottlenecks of data access, analysis and collaboration created by the highly fragmented and rigid landscape of informatics and data management systems. The lack of mature, generally accepted and standardized software solutions has created a situation where many IT organizations feel compelled to create laboratory-specific custom solutions or string together several off-the-shelf point solutions that make lab optimization and data sharing difficult. Where multiple point or custom solutions are cost-prohibitive, scientists are often left on their own to create ad hoc solutions using general purpose office software that are manual, error-prone and lacking domain intelligence. The resulting informatics landscape creates data bottlenecks that limit the utilization of lab resources and previous work.

Life Science Insights (LSI) (Framingham, MA) believes that the industry will move toward adoption of dynamic, user-controlled informatics to automate changing data requirements, and that the most successful of these tools will integrate both computational biology and experiment design to create a comprehensive solution for data capture, analysis and exchange. Teranode Corp. (Seattle, WA) is addressing these main points in the industry through an experiment design automation platform powered by user-defined models.

LSI conducted case study interviews with early Teranode customers to identify key factors impacting adoption and utilization of its automation platform.

This article describes the problem of informatics bottlenecks, the competitive informatics landscape, and a novel solution developed by Teranode to respond to the problem.

The Problem
Industry Productivity Dilemma

It is abundantly clear that the pharmaceutical industry faces a R&D productivity dilemma. Despite steadily increasing annual investment in R&D over the past 12 to 15 years, increasing levels of biological data and knowledge (e.g., genomics, proteomics and other “omics”), and the adoption of high throughput screening methods, the industry has not seen a correlated rise in the number of new drugs receiving approval.

Informatics Bottlenecks: Fragmentation and Rigidity

In life sciences R&D, the technology bottleneck has often been the physical throughput of a laboratory or direct access to existing data sources. While laboratory productivity has increased through miniaturization and automation and data access has been facilitated by Internet technologies, the large quantities of diverse data now available has created new problems. Large fragmented reservoirs of rigid data now require consolidation and analysis to effectively extract information from this data. As a result, data management and analytics software are now the bottlenecks to scientists’ abilities to use and optimize the increased experimental data sources and lab capacity. Even while swimming in data, too few previous experimental data sets are reused in planning lab projects, resulting in wasted effort and dead-end experiments.

The main sources of the informatics bottleneck come from fragmentation and rigidity in the current landscape of informatics, databases and analytical tools. The bottleneck arose in part from the nature of the tasks in drug discovery and development. Arising historically from academic research, the highly diverse technical disciplines that serve as the foundation for drug research carried forward specialized tools, approaches and data formats that were very different from one another. As a result, pharmaceutical companies that initially began with inherently isolated organizational structures have been required to invest considerable resources to develop common platforms and tools.

Existing Solutions

These fragmentation problems are well recognized. Scientists, informatics specialists and supporting IT groups have tried a number of approaches with varying degrees of success.

Ubiquitous General Purpose Software: Excel and Access

One of the most common approaches is the use of Microsoft Excel® and Access by scientists to develop data processing, filtering, integration and analysis routines. On the one hand, these tools are used in such an informal manner that they could hardly be called a solution; on the other hand, they are so ubiquitous that they are heavily relied on by the industry and are functional on some level.

Ad Hoc Solutions, High-level Scripting Languages

Another set of commonly used tools for integration and dataflow automation are high-level scripting languages, in particular, Perl scripts. Perl is a favorite among Unix and Linux users, who make up an important subset of scientists. Perl is designed to be a very flexible, high-level language that allows multiple approaches to accomplishing the same data manipulation tasks.

Standardization on a Single Database or LIMS Platform

Some pharmaceutical companies have strategically chosen to standardize by designating a single database platform or LIMS application as the standard within drug discovery and development. That platform or application becomes the de facto unifying technology for data automation. This solution to fragmentation is functional but gives up significant flexibility. Laboratories choosing this route typically require that all analytic tools integrate directly (out-of-the-box) or be accessible through custom integration.

Analytic Suites

An analytic suite from a major commercial vendor provides primary analytic tools, as well as a common platform. Experimental data is named and stored in formats that fit best with the analytic suite, and as much of the work as possible occurs within the confines of that suite.

Knowledge Management Solutions

Solutions that are based on knowledge management principles and technologies tackle the integration problem by a different approach: documents in place of data. Technologies used to support knowledge management approaches include: electronic lab notebooks (ELN), document management systems, intranet portals, online collaborative rooms for project teams, and indexing and search tools layered over the mostly text data that conveys results.

Custom Solutions

A number of custom solutions to the problem of fragmented informatics are currently in use. In the most radical custom solutions, the applications are also completely customized. A handful of biotech companies have invested in a completely custom informatics infrastructure, based on the unique situation where research approaches and off-the-shelf application software simply don’t work. In general, however, the mix is more typically 40 per cent custom code, 40 to 45 per cent off-the-shelf applications, and the remainder open-source type codes from academia.

Hybrid Approaches

It is rare that a pharmaceutical or biotech lab will take only one of these approaches to addressing the issues of data informatics fragmentation and rigidity. More often a lab will deploy some hybrid combination: a LIMS, a database platform, an analytical suite, some specialized point solutions, open source applications, and custom algorithms with middleware tools and scripting languages providing some pools of integration. And still, scientists will find that they resort to ad hoc methods with general-purpose tools.

A New Approach to Experiment Design Automation

Experiment design automation (XDA) is Teranode’s answer to the bottlenecks of today’s fragmented and rigid informatics landscape. The XDA provides a platform for designing, managing, analysing and collaborating on data from experiments in a wide range of labs and research disciplines. It also provides dynamic, scaleable data automation driven by user-defined models of computational biology and experiment design.

Having a new class of model-driven informatics tools that use database-driven solutions can alleviate some of the critical problems. Model-driven architectures, growing out of object-oriented programming, and Unified Modeling Language™ (UML) are gaining momentum more generally within enterprise IT departments as an approach to creating flexible and cost-efficient solutions. Model-driven informatics will have a productivity impact on design and analysis of experiments that is analogous to the productivity impact that electronic design automation (EDA) tools had on the design and analysis of integrated circuits. The key success factor for EDA was the evolution of languages that allowed scientists and engineers to search, exchange and visually analyse scientific and design data. In life sciences, the goal of XDA is to create a platform that accomplishes the same for experiment data in the drug discovery and development process.

Through focused XDA-based solutions, users can experience good short-term value for their R&D processes, as well as a foundation for long-term integration of laboratory workflow management and integration with computational model-based analytics.

For more information on this topic, please go to: http://www.teranode.com/docs
/LifeScienceInsights.pdf”>http://www.teranode.com/docs/LifeScienceInsights.pdf to access a white paper titled, “Experiment Design Automation: A Potential Solution for Fragmented Informatics in Biopharmaceutical Research and Development,” sponsored by Teranode and authored by Life Science Insights. Please note that registration is required.

Alan S. Louie, PhD is research director at Life Science Insights, an IDC company (Framingham, MA), focusing on new technology solutions in drug-discovery research. Louie brings more than 21 years of technical, product development and business consulting experience from the diagnostics and biotechnology industries, and has been responsible for the development of more than 30 commercial products. Louie can be reached at louie@lifescience-insights.com”>mailto:alouie@lifescience-insights.com.

Michael Swenson is Director of Research at CIO Executive Council and was formerly Research Manager at Life Science Insights.