When you’re inside a maze, it’s difficult to know which path to take and which direction to go next. Wrong turns and dead ends in the labyrinth are discovered through experimentation. You backtrack, repeat steps, and try new routes without knowing if any of them will lead you closer to your ultimate goal. But if you can elevate your view of the maze, you have better information about which paths are unproductive and which represent progress.
When scientists endeavor to optimize a compound for drug development, it’s as if they are trying to find the shortest path through a maze. But they don’t need to rely on purely experimental approaches. They can use knowledge to elevate their view so they can make informed decisions and find the shortest path to a successful outcome. Predictive data models and information about previous experiments can help scientists recognize which choices are likely to accelerate progress and which will lead to dead ends. The goal is to achieve better insights sooner by making data available throughout the entire workflow.
Achieving a Collaborative Data Model
Most life science organizations have correctly identified gaps and inefficiencies in their processes and targeted them successfully. Individual software applications can streamline specific activities and find patterns in the data to help scientists make decisions. But the data in these solutions are typically not readily shareable in today’s highly externalized business model that includes geographically distributed internal departments and collaboration with partner organizations. Thus the value of these targeted ‘best-of-breed’ applications is diminished, as they don’t fit the overall drug development workflow.
To truly transform the drug development process, we need to think of the data generated as part of the continuous workflow and not as discrete silos. Data analysis creates knowledge, and that knowledge needs to flow both upstream and downstream in the process. As such, life science companies should consider adopting a paradigm that focuses on the organizational workflow as a whole. Discovery, development, and manufacturing must be treated as a continuum rather than as independent activities.
To enable this holistic approach, organizations need to standardize data and capture it everywhere. Because externalized research, development, and manufacturing are now the norm in pharmaceutical development, companies need also to implement data standards that facilitate seamless and secure collaboration throughout the extended enterprise. This strategy provides a foundation that enables computational scientists to mine the data for patterns and insights that accelerate progress.
Fortunately, the pharmaceutical industry is making headway on developing a collaborative data model. This is a challenge for software developers as well as an objective for the industry as a whole. Unsolved issues include unifying data formats and standardizing semantics. IT leaders at pharmaceutical companies are coming together to share pre-competitive strategies in projects like HELM (Hierarchical Editing Language for Macromolecules) from the Pistoia Alliance. Strategies like these are critical for producing the predictive models and continuity of knowledge that can elevate the view of the maze.
Holistic Workflows Enhance Insight
The drug development process gets increasingly costly as it moves downstream.
Because each wrong turn or dead end in the maze can potentially be very expensive, some organizations wisely strive to minimize costs by failing potential drug candidates early in the process. For example, if molecular biologists can understand early in the discovery process that a certain antibody will be a bad candidate for development and manufacturing, they can avoid investing time in it. Better still, if they discover this early enough, they can design that undesirable property out.
Combining predictive models with physical experiments in early stages can facilitate this level of insight. Using in silico studies along with wet lab experiments in these early-discovery stages can save companies time and money by helping them to virtually look ahead and avoid the wrong paths in the maze. Scientists working in discovery can foresee what is likely to succeed in late-stage clinical trials and endeavor to design with that outcome in mind.
Companies can support these strategies by adopting holistic workflows. Of course, companies may not want to abandon their existing specialized software applications. These tools typically function well within their defined niches. Often they represent technology investments that support unique business practices. Any change could detrimentally affect business efficiency in the short to medium term. Therefore, an open technology platform is needed—one that allows an organization to integrate these point solutions so they become part of a unified workflow. Each task and the associated results can then be made available throughout the continuum.
Once data, applications, and workflows are unified, life science organizations can fully leverage the benefits of virtual design strategies. For example, when a chemist creates a model of a molecule that will be synthesized, a scientifically-aware system could submit data into an in silico screening process based on that model. This process would predict biological end-points and biophysical properties that would inform the chemist about anything that might make the compound undesirable for use in drug development. Unified processes would enable scientists to validate molecular models at the point of design.
A scientifically-aware system doesn’t simply inform researchers about whether an idea is good or bad; it should also provide sensible suggestions for alternatives. The system drives the process forward by suggesting alternate ideas from a similar starting point for experimentation and study.
A Unified Approach to Drug Development
The long-term objective for pharmaceutical organizations should be to achieve competence in both virtual and physical domains. Scientists need to consider in silico studies to be on a par with physical experimentation and they should be able to switch between them seamlessly.
The first step involves seamless collaboration with partners to help experts identify patterns and insights. Laboratory software captures data. Collaborative science applications bring the data together and enable people to visualize it. Predictive sciences enable researchers to mine data, make predictions, and gather actionable insight to move the process forward. An integrated technology solution allows experts to uphold best practices that span individual point solutions. By applying these principles, life science companies can elevate their view of the drug development continuum to find the shortest path through the maze.
About the Authors:
Adrian Stevens is director of product management for predictive science applications at Dassault Systèmes BIOVIA. Prior to joining BIOVIA in 2008, he spent more than 12 years working in the domain of computational chemistry and biology within the pharmaceutical industry. Dr. Stevens has published approximately 20 papers and is named on three patents.
Tim Moran is director of life science research product management at Dassault Systèmes BIOVIA. His early work in the industry focused on immunomodulation and imaging studies of effects on T-cell lymphocyte homing. He has held several managerial positions in image informatics as well as roles in life science research, next- generation sequencing, sequence analysis, biotherapeutics, and registration.