By Jim Tung

The introduction more than five years ago of systems biology tools for modeling pathways delivered an enormous gap between user expectation and tool capability. Those first-generation tools had modest simulation capabilities at best, despite their accompanying hype. In some cases, they were just drawing tools that did not simulate models at all; in other cases, they performed simulations using less robust and less accurate mathematical techniques, which have been surpassed by more recent work. In general, the analytical techniques underlying the tools could not answer biologically relevant questions.

Numerical robustness and ease of modeling weren’t the only issues. Experimentalists complained that the simulations didn’t correlate with experimental data. When forced to choose between simulation results and experimental results that differed, they would go with the latter every time. And the results differed pretty often. This disparity highlighted the need for algorithmic techniques that define or refine models so that key model aspects can be derived from, or correlated with, experimental data.

Beyond the tool limitations, the disparate nature of the community didn’t advance the cause of simulation. To succeed with simulation, you needed collaboration between a computation/math person and a biologist, which wasn’t typical at the time. Today, we see more and closer collaborations, a primary reason why tools are improving and breathing renewed hope in the field. Another positive change has been academic curricula that, over the last few years, have broadened the training of biologists in modeling, mathematics, statistics, and systems, and the training of computation/math students in scientific topics.

Recently, a new generation of tools has emerged for modeling and simulating complex pathways that uses graphical and textual representations to simplify building and understanding complex models. These tools, equipped with new mathematical solvers that robustly simulate the models, address some of the roadblocks that limited the value and applicability of previous modeling tools. Techniques for model-fitting or parameter estimation have helped to create a bridge between experimental results and models, providing ways to correlate the two and verify model simulations against experimental data.

Researchers are focusing on the kind of data (and accompanying experimental techniques) that can effectively help model development. Some modeling tools can connect directly to algorithm development tools, which enables research in that area. This important research needs to be supported and encouraged, both by the community and through more rapid incorporation of techniques in tools.

As I compare systems biology to other scientific and engineering areas in which simulation plays a key role, I am intrigued by how the market has segmented such that simulation platform vendors are sometimes also the IP vendors for investigating a particular disease. Certainly, this business model can help establish value for the customer, at least initially, and it occurs frequently in industries when simulation is first being applied.

However, while this approach may generate some value initially, I have observed that simulation really flourishes when the simulation platform vendors are not the same as the IP vendors. Such a market structure creates clean distinctions that facilitate collaboration without conflicting interests. Each party can (and should) evolve its respective piece in support of the other parties: the tools, the data underlying the models, the IP represented in the models, and the application of the tools and IP within the end-user organisation, which in turn drives the ongoing development of the tools and modeled IP.

Providing a simulation platform as a vehicle for others to deliver IP or knowledge bases is a business model we’ve used for many years, and certainly is the way the computer and software industries have evolved. Although I understand that our approach is not yet typical in the computational biology area, I think it serves the current systems biology simulation market.

The community needs it to:

• produce results that benefit the customer now
• rapidly evolve the tools
• take advantage of increased interdisciplinary collaborations for larger longer-term benefits

The immediate and long-term benefits can differ according to customer objectives and priorities. They might include:

• The ability to represent and communicate systems biology knowledge acquired through experimental data analysis or small-scale simulations so that fundamental understanding can grow over time
• A reduction in the number of experiments needed by making some decisions based on simulations that have been verified against data
• In the longer term, the identification of candidates through simulations as a more cost-effective manner than relying solely on experiments

The first two benefits are being realized today, and the third benefit is still evolving.

In the past few years we’ve seen simulations progress down the pipeline across applications ranging from systems biology to pharmacokinetics. Companies are exploring where and how simulations deliver the biggest ROIs for their processes and planning horizons. That’s proven very interesting to watch.

Increased collaboration – between life scientists and computational developers, between experimentalists and algorithm developers, between tool vendors, modelers, and users – are combining with today’s stronger collection of tools to fulfill some of the potential promised by that first generation of tools.  We look forward to the discussions and collaborations to make simulation a more valuable and critical approach for the computational biology community.