91av

GLENN ALLGOOD’S victim is heading for a battering. “I’d like to subject it to
carcinogenic materials,” he muses. “I think you should be able to give it a
suntan or a blister. I’m serious. I mean, you should be able to cut it. You
should be able to have it cough up a hairball if you want to.”

Soon he might be able to do all that and more—without going to jail,
and without causing anyone or anything a moment’s suffering. Allgood, a
computational engineer at Oak Ridge National Laboratory in Tennessee is involved
in creating the most complex computer model ever attempted: a virtual human
being. It could take decades and billions of dollars to build it, but the dream
of a Virtual Human is now becoming reality.

This won’t be a virtual cartoon character, like cyber-newsreader Ananova. Nor
will it be a merely visual representation of a human body, like the National
Library of Medicine’s Visible Human image archive. The Virtual Human will have a
living, breathing body whose cells will replicate and die, and whose blood will
flow under Allgood’s knife. True, the Virtual Human will be alive only inside a
computer, but the simulated gash will provoke the same cascade of physiological
reactions a real person experiences. Immune cells and clotting factors will rush
to the wound, and biochemical stress reactions will reverberate throughout the
body. The Virtual Human’s accurate simulation of human biochemistry will also
mean it could test new drugs for us, and its realistic physical responses will
allow the military to test the effects of the latest weaponry. In short, it will
be given the worst job in the world: pharmaceutical guinea pig, crash test dummy
and biologists’ action toy, all combined in one unlucky cyber-human.

Not only will the model work like a human, it will also be fully explorable
inside and out. Virtual-reality software being developed will allow researchers
to see inside the Virtual Human’s organs, or watch its blood vessels dilate in
response to drugs. They will even be able to set out on a journey inside the
body using interactive “total immersion” software that will let them stand in
the airflow entering the lungs or listen to the flapping of faulty heart
valves.

Being able to sit and watch the workings of the interconnections between
different organs, map the whole body’s reaction to different chemicals or
physical stimuli, or watch how a disease affects different areas of the body
will also uncover invaluable new medical information.

The idea of building, using and abusing this virtual re-creation of humanity
was born in 1996 when Clay Easterly, a health physicist at Oak Ridge, was
approached by representatives of the US Marine Corps. They were looking for a
way to test experimental non-lethal weapons without having to shoot them at real
people. Could the weapons’ effects be simulated in a computer, they wondered. No
chance, said Easterly. “We don’t have that much knowledge about the overall
human,” he told them. “We have knowledge in bits and pieces, but we don’t have
integrated knowledge”.

But the enquiry got Easterly thinking about the diversity of modern
biological research. Many of the tens of thousands of journal articles published
each year must impinge on each other, he realised, but they are often considered
in isolation. There’s no sure way for the implications of one study to be
applied to another. Computer models of human cells, organs and
systems—ranging from the most basic representations to relatively
sophisticated simulations of real biochemistry and function—were also out
there, scattered among various academic and commercial laboratories.

The virtual heart developed by Denis Noble at the University of Oxford shows
vividly how useful biological modelling can be (91av, 20 March
1999, p 24). Noble’s heart is a carefully assembled mass of virtual cells, each
processing virtual sugar and oxygen and behaving just like the real thing. Noble
can watch his virtual heart beating on a computer screen. He can make it develop
diseases, then treat it with virtual drugs. Drugs companies have used his heart
to test for adverse reactions (see “Heart of the matter”).

Noble is a co-founder of Physiome Sciences, a company that specialises in
modelling human organs and cells for the pharmaceuticals industry. Physiome has
crude but functioning models of dozens of types of human cells, and is working
on building them into complete immune, endocrine and bone systems. It’s
part of a whole new industry that is springing up, offering pharmaceuticals
companies virtual versions of everything from single receptors to multi-organ
systems for testing potential drugs. Entelos, for example, a modelling company
based in the heart of Silicon Valley at Menlo Park, California, specialises in
simulating disease states. Its computer model of asthma includes elements of
every relevant structure and process, from airways to immune cell reactions. If
companies like these, along with academic researchers, could stitch their
efforts together—and find a big enough computer to run the resulting
model—a rudimentary Virtual Human could begin to take shape tomorrow.

But it’s not going to be easy. On top of the huge technical challenge, there
is a major organisational barrier. Turning today’s models into a Virtual Human
will involve standardising the way biological information is collected, stored
and shared, since the current models are mostly incompatible. Each represents a
particular scale, from cell to tissue or organ, and is generally incapable of
receiving input from another model. Most of them were custom-designed by their
creators right down to the way they input their data and the programming
language they employ. At present, one model’s kidney would be deaf to what’s
happening in another model’s liver. Anyone hoping to assemble the various
simulated organs and systems faces a digital Babel.

But this may be about to change. In March this year, Physiome Sciences
announced that it will allow non-commercial researchers to use its
cell-modelling software free of charge. The company says it wants to help
promote the idea of biological modelling, though researchers who use Physiome’s
software will find there are strings attached. The company gets access to the
researchers’ data to improve its own models, and the researchers must agree not
to reveal the company’s algorithms. Nevertheless, Physiome has already had more
than a dozen applicants, ranging from academic researchers and biotech companies
to a group within the National Cancer Institute and another at the US Department
of Agriculture. While some may fear this raises the spectre that some of
biology’s most useful information could fall into private ownership, Easterly
welcomes Physiome’s initiative as “a tremendous step”. Anything that makes
biological modelling technology accessible to more researchers is welcome, he
says.

Easterly, Allgood and their colleagues at Oak Ridge have been hoping that
government agencies will fund the infrastructure work, such as establishing
standard data formats and programming languages. They would like to see a more
complete and realistic model than can be built with the crude commercial models.
But the scale and ultimate expense of the Oak Ridge ideal seems to alarm those
holding the public purse strings. Modelling the whole human body will mean
dealing with billions of megabytes, much more data than has come out of the
Human Genome Project. Scientists are still struggling with the glut of genome
data, so government agencies are reluctant to embark on an even more ambitious
venture.

Meanwhile, Physiome Sciences is forging ahead. “We’re not going to wait for
them,” says CEO Jeremy Levin. But he insists this doesn’t signal another
genome-style race between private enterprise and a publicly funded effort, not
least because no single company could possibly afford to build its own Virtual
Human. If a public version ever gets going, Physiome Sciences will “commit
heavily” to it, Levin adds.

Whoever funds the Virtual Human, it won’t be much use to researchers unless
they can interact with it in a way that makes instant sense. Oak Ridge computer
engineers are already working on this. For starters, they are building a Virtual
Human portal—a website to act as a door to the model from the outside
world. The idea is that researchers will sit at their own computer terminals and
see, for example, how the Virtual Human’s blood pressure responds to being given
varying doses of a particular drug.

Eventually, things could get much more ambitious. The developers hope to pipe
the Virtual Human’s output into a total-immersion environment in which
researchers can explore inside the body. The idea is based on systems already
being developed as training and diagnostic tools for doctors. The user wears
goggles that create the impression of a 3D environment from images projected
onto four walls. Eventually, the images will respond to touch. “This
allows you to get a real close-up feel of what’s happening with the data and
interact with it,” explains Oak Ridge engineer Richard Ward.

Stepping into the immersion chamber, researchers could embark on a “fantastic
voyage” into a single cell, use a haptic wand to poke a ribosome or just move
some molecules around to see what happens. Others could wade into an
artery and watch close-up how tissue properties and the fluid dynamics of blood
change in response to some simulated stimulus. A surgeon might prefer to stay
outside the body and try a new procedure to see how the Virtual Human copes.
And, of course, military researchers will be eager to shoot it with virtual
versions of their experimental weapons to see what damage they do.

But perhaps the most exciting prospect is that the Virtual Human may give us
new and unexpected insights into the way our bodies work. When biologists
program in some new observations about the pancreas, say, it might influence
parts of the Virtual Human that no one had ever guessed would be affected, such
as the lungs or the heart. The answers to mysterious medical problems could
emerge every time the model is examined or altered, Easterly believes.

The Virtual Human could also be customised to isolate the effects that drugs
have on different individuals, allowing researchers to investigate the influence
of sex, age, racial background or any other factor, without the need for
elaborate, time-consuming and expensive clinical trials. With that technology in
place, Easterly says it should eventually be possible to create a virtual
version of every one of us, tailored to our individual genome, medical history
and other specifics.

But even the most ambitious modellers are steering clear of one important
organ: the brain. “There is every possibility of modelling a human neuron, and
perhaps a cluster of neurons,” explains Levin, “but modelling the human brain is
outside the realm of our reality.” As long as they model the brain’s basic
outputs—the autonomic nervous system and endocrine signals, for
example—the researchers believe the body will still function normally.

The Virtual Human promises to be a revolutionary tool. Once biologists get a
taste of the advantages to working “in silico”, Easterly believes demand will
escalate, continually improving the model and encouraging more researchers to
use it. Eventually, he says, many scientists will wonder how they ever lived
without the Virtual Human.

Perhaps it’s for the best that the Virtual Human will be devoid of brain
power. As it vomits up new antibiotics, takes a hail of bullets in the chest,
and develops a particularly nasty skin cancer, it can do without the added
burden of realising what a rotten day it’s having.

When pharmaceuticals giant Hoffman-LaRoche was putting its heart drug
mibefradil through clinical trials in 1997, a blip appeared in the
electrocardiograms of some subjects. It looked like a cardiac malfunction known
to be lethal. Fearing the drug might never make it to market, the company turned
to Physiome Sciences’ virtual heart.

Denis Noble, Physiome’s co-founder and a physiology professor at the
University of Oxford, spent decades developing the heart model, along with
collaborators Raimond Winslow of Johns Hopkins University and Peter Hunter at
the University of Auckland. When Noble “administered” the drug to his virtual
heart, it too experienced the blip, and he was able to trace it back through the
simulation to a benign phenomenon unrelated to the deadly condition they feared
it revealed. The US Food and Drug Administration accepted Noble’s explanation of
the anomaly and went on to approve the drug.

But that’s not the end of the story. A year later, new data showed that
mibefradil was interfering with certain liver enzymes, creating a risk that
other medication might build to dangerous levels within a patient’s body. The
FDA threatened to rescind mibefradil’s approval, and although it is still
available elsewhere, Hoffman-LaRoche voluntarily withdrew it from the US market.
Had Noble been able to link his heart model with a liver model—as in a
virtual human—the problems with mibefradil would have shown up before it
reached any real-world trials.