|
Having a virtual copy of
a patient’s blood in a computer would be a boon to researchers and doctors.
They could examine a simulated heart attack caused by blood clotting in a
diseased coronary artery and see if a drug like aspirin would be effective in reducing
the size of such a clot.
Now, a team of biomedical
engineers and hematologists at the University
of Pennsylvania has made
large-scale, patient-specific simulations of blood function under the flow
conditions found in blood vessels, using robots to run hundreds of tests on
human platelets responding to combinations of activating agents that cause
clotting.
Their work was published
in Blood.
Patient-specific
information on how platelets form blood clots can be a vital part of care.
Normally, clots prevent bleeding, but they can also cause heart attacks when
they form in plaque-laden coronary arteries. Several drugs, including aspirin,
are used to reduce the size of such clots and prevent heart attacks, but, as
platelets differ from person to person, the efficacy of such drugs differs as
well.
“Blood platelets are like
computers in that they integrate many signals and make a complex decision of
what to do,” said senior author Scott Diamond, professor of chemical and
biomolecular engineering in the School
of Engineering and
Applied Science. “We were interested to learn if we could make enough
measurements in the lab to detect the small differences that make each of us
unique. It would be impossible to do this with the cells of the liver, heart,
or brain. But we can easily obtain a tube of blood from each donor and run
tests of platelet calcium release.”
When blood platelets are
exposed to the conditions of a cut or, in a more dangerous situation, a
ruptured atherosclerotic plaque, they respond by elevating their internal
calcium, which causes release of two chemicals, thromboxane and ADP. These two
activating agents further enhance calcium levels and are the targets of common
antiplatelet drugs such as aspirin or clopidogrel, also known as Plavix. By
preventing platelets from increasing their calcium levels, these drugs make
them less able to stick together and block blood vessels, decreasing the
likelihood of a heart attack.
Since blood is a liquid,
the liquid-handling robots originally developed for drug screening tests were
ideal to test platelet function.
“We used a technique
developed in our lab called ‘pairwise agonist scanning’ on platelets from three
different donors to generate a massive data set of how their cells responded to
all different pairs of these activating agents,” Diamond said. “Then we trained
neural network models for each donor based on this data to simulate how each
and every cell in a blood clot is responding.”
Neural networks are a way
of looking at the relationship between inputs and outputs for very complex
processes, rather than at the details of the process.
“They summarize the
overall function of all the chemical reaction networks that are occurring
within a single platelet,” Diamond said.
Graduate student and lead
author Matt Flamm developed a powerful multiscale computer model that populates
a simulation of blood flowing over a site of vessel damage with thousands of
platelets whose behaviors derive from the neural network model developed for
each patient.
“This is the first time
that it has been possible to predict blood clotting under flow using patient-specific
platelets,” Flamm said. “We were able to predict the ranked potency of several
drugs.”
To show that the computer
simulations allowed them to make accurate predictions about an individual
donor’s platelet behavior, the researchers performed physical test as well.
Using microfluidic devices, they ran scores of blood tests with each blood
sample at venous and arterial flow conditions using different drugs.
The multiscale computer
simulation for each donor predicted the drug responses very accurately.
“We even identified one
person who was resistant to aspirin,” Diamond said, “and then discovered a
novel genetic mutation in their thromboxane receptor gene. The computer
simulation for that donor identified the functional defect before we even
sequenced the gene.”
Multiscale,
patient-specific simulation of blood function is an example in the rapidly growing
field of systems biology. Multiscale models require the understanding of the
intracellular signaling in thousands of individual cells activating at the site
of damaged blood vessel as well as detailed calculations of blood flow and
molecular diffusion.
“Fields like weather
prediction and airplane design simulate the flow of air,” Diamond said, “In
cardiovascular medicine, we encounter the individually unique and complex fluid
of human blood. Research areas involving traumatic bleeding, stroke, and deep
vein thrombosis may benefit advanced simulations of blood function.”
The development of
equations and algorithms to model reactive blood flow will be very helpful in
predicting clinical risks, drug responses, and new disease mechanisms and in
designing biomedical devices.