Virtually Human

Scientists want to create distributed software that re-creates the functions of the human body.

June 10, 2002 (Computerworld)

Leave your heart in San Francisco. And while you're at it, leave your ears in Boston, your liver in Tennessee and your lungs in Seattle.
Charles DeLisi, who began the Human Genome Project, is trying to get financial support for an even more awesome undertaking: re-creating in software all the physiological processes of a human being.
DeLisi, a professor and director of bioinformatics at Boston University, hopes to create a virtual human whose organs will be dispersed among research facilities across the country.
This is no cyborg. It would be a distributed computer program that uses algorithms to re-create the functions of the human body, and in particular to study how functions in one part of the body impact others. Ultimately, the models could predict what a chemical, virus, bacterium or physical trauma would do at the cellular, organ, system and organism levels. Doctors in an emergency room could see the effects of physical trauma without opening up the body, and patients could have drugs customized to their body chemistry.
In a setup similar to that of the Human Genome Project, the Virtual Human Project would divvy up computational responsibilities. Specialists at each university or laboratory would develop software that represents a particular organ and then test what effect different stimuli have on that organ.
Software for some organs and functions would be completed far in advance of others. Development would be ongoing, and there wouldn't be a completed virtual human, but rather an ever-evolving model. That, combined with the complexity of the programming, dictates that the virtual human must be developed and run on a distributed basis, DeLisi says.
Boston University has already created a virtual cochlea, the bones inside the ear that vibrate in response to sound waves and translate them into electrical impulses for the brain. And, DeLisi says, scientists and programmers will make major progress in the next year on mapping how our sense of smell works and re-creating that in software. Vision is expected to follow over the next 10 years or so.
Physical Limitations
The virtual human would link the organ models over the Internet as they are created and eventually produce a holistic picture of the human organism. "It's been done in astronomy. It's been done in molecular biology. It's not been done in physiology," DeLisi says.
Researchers say biological information about humans has become too specialized, voluminous and dispersed to manage effectively without analytical and interactive help from computers.
"The human has got a wonderfully adaptable brain, but the input is slow, and we get distracted," says Clay Easterly, a researcher at Oak Ridge National Laboratory in Oak Ridge, Tenn., who is also promoting the virtual human. "I don't see any way for a human being, or a group of human beings, to synthesize this vast data."
While the project has drawn some interest from the government, academia and industry, no one seems inclined so far to fund major parts of it. For the time being, the virtual human is likely to grow from relatively small efforts, such as the virtual cochlea project.
Money won't solve all the problems anyway. Computer models are still relatively crude, and they're narrowly defined to a particular reaction - in a cell or in an organ, for example.
The many-to-many structure of the Internet and the use of XML would be the framework for sharing information that now exists in isolation, DeLisi says. But creating programs that accept input from other applications is a challenge.
The virtual human would require integrating processes and software across vastly different scales, from molecules to whole organisms. "I don't think we have the foggiest notion of how to go about it," Easterly says. Programming would have to be done at each level - the molecular, cellular, organ and systemic - simultaneously and progress up and down until the functions met. Biological research would influence, and be influenced by, the programming effort.
The entire process could take 100 years or more, especially since it must ultimately include the brain. "That, I think, is a ways off," DeLisi says.