There's a widely accepted view that the genome is the computer program of life, and genes the software subroutines that code from DNA via RNA to proteins; and that like computers, when the software goes wrong bad things happen.
But biologists building computer models for testing drugs or pinning down the causes of disease say these metaphors are past their sell-by date. Instead, they're using computer models to try to see what effect different genes really have on our bodies.
A pioneer in this field, Denis Noble, codirector of computational physiology at Oxford University, has spent 47 years developing biological models. As a medical student in 1960 he developed the first viable mathematical model of a working heart cell, showing how it was possible to reproduce the heart's rhythm by modeling the changing electrical potential within it. Today, his heart cell models are so accurate that pharmaceutical companies use them to test for the effects of drugs on cardiac arrhythmia.
And the model shows effects you might not expect. For example, in Noble's heart cell model, if you remove the pacemaker gene protein -- first discovered in 1997 -- which generates 80 percent of the heart's electrical current, you might expect that the heart would stop or go haywire. In fact there is almost no change.
"The system is so robust that other mechanisms -- 40 or 50 other proteins -- ensure that if one fails the others can take over. It is beautifully failsafe," Noble says. "Nor can you conclude that if knocking out a gene has no observable effect it's not involved in a particular body function."
It takes all night on a computer with 18 processors to simulate one second of a complete beating heart, reconstructing the function down to the cellular level, so Noble is looking forward to having access to a 10-petaflop supercomputer -- the equivalent to about 5,000 consumer PCs -- being developed by Fujitsu. The machine is being designed to run simulations of complete human organ models in real time, including a whole organ heart model to be provided by Noble and his collaborators. Other teams are building multi-level computer models of all the other human organs as part of the Human Physiome Project.
Roche is one of the many pharmaceutical companies working with computer models from university groups such as Noble's, as well as using commercial mathematical modeling tools from companies such as MathWorks and Entelos.
Cristiano Migliorini, a modeling expert at Roche, suggests that biologists proposing new theories will soon submit computer models with their papers so that other scientists can re-use this knowledge.
"If someone publishes a good paper on a low-level mechanism in the liver, for instance, it would be great to be able to slot that software into a larger scale, higher-level liver model," he said.
Noble, meanwhile, has distilled his findings into "10 Commandments of Systems Biology." This list, in his paper in last October's issue of the Journal of Experimental Physiology, challenges popular perceptions about genes. Genes, Noble says, cannot be assigned specific functions; DNA is not the sole transmitter of inheritance; and there is no "genetic program." In fact there is no "program" at any level, including the brain.
Noble questions the "dogma of genetic determinism." The idea that it's rarely correct to attribute a function solely to one gene seems at odds with the headlines about discoveries of cancer genes, pacemaker genes, depression genes and so forth.