Sir

Without minimizing problems presented by large-scale ranges, as indicated in H. Kitano's good overview article in your Computational Systems Biology Insight (Nature 420, 206–210; 2002), I think that the potential leverage offered by a 'layered platform' approach to modelling biological processes deserves some mention.

Members of the Biomedical Information Science and Technology Initiative (BISTI) of the US National Institutes of Health (NIH) concluded in 1999 that if we properly develop software tools, “time spent reinventing the same [software] processes in one laboratory after another will be freed for basic research” (see http://www.nih.gov/about/director/060399.htm). The NIH subsequently started a programme focusing on tool development, called Innovations in Biomedical Information Science and Technology, and, more recently, established the National Institute of Biomedical Imaging and Bioengineering.

How exactly does one avoid reinventing the wheel? As an illustration, transport phenomena are involved in biological/ physiological processes whose scale ranges from individual protein production, via tissue nourishment and organ functionality, to whole-organism homeostasis. While transport is the common underlying theme, each system differs in geometry and the properties of materials involved.

A single 'transport platform' that allows a user to configure the geometry with finite element analysis and to specify 'material' behaviour with specific constitutive equations could simulate applications in all biological transport systems. The remaining 'layers' of the total solution — fluid mechanics, chemical reactions, heat transfer, matrix inversions and so on — would be supplied by experts in these disciplines and managed automatically.

The feasibility of this approach, and its compelling economic logic, is supported by the success of analogous platforms in other venues, for example ANSYS in structural mechanics (see http://www.ansys.com). What is perhaps not so obvious is that biologists, physiologists, clinicians and others can model phenomena that they understand without being required to make a major time investment in such matters as analytical mechanics, optimum matrix inversion techniques, database management or graphical user interfaces.

As we begin the effort to develop computationally intensive models of biological phenomena, it is worthwhile to pause for a moment and think about the structure of the tools that are needed.