E&T explains why cardiac modelling opens the door to the new era of medicine.
After nearly half a century of research investment, heart models are starting to pay dividends, providing insights into the electromechanical operation of this most complex and reliable organ and, for the first time, assisting in drug development. There is now growing confidence that heart models will within the next decade achieve their ultimate goal of helping cardiologists and surgeons in the clinic, while playing a major role in the process of discovering new drugs for a variety of potentially fatal cardiac conditions.
Heart modelling has also become a driver for systems biology, aiming to make medicine more analytical - like physics - by integrating its components: from genes and molecules up to organs - into a holistic whole. It is stimulating key systems biology technologies, such as parallel visualisation and magnetic resonance imaging and spawning initiatives, such as the Virtual Physiological Human Network of Excellence (VPH NoE) - a European project to model and simulate the whole human body.
Heart modelling has inspired such projects for several reasons. Most fundamentally, the heart is the ultimate four-dimensional organ, relying for its function on highly accurate timing and efficient mechanical operation, to draw blood in and pump it out at varying rates determined by electrophysiological inputs. The latter, in turn, are driven by changing levels of physical and metabolic activity. This four-dimensional operation is also reflected in its disease states, many of which are associated with mistiming, causing arrhythmias, in which the heart beats irregularly or at the wrong speed, increasing risk of cardiac arrest and sudden death.
The other major force behind heart modelling lies in its therapeutic potential for planning surgical treatments, installing pace-making devices, or developing new medicines. One ultimate goal is to build models capable of simulating a patient's heart in real time in the clinic, taking account of the disease state and predicting the outcome of therapies, including surgery and drugs. Another, equally ambitious, goal is to exploit the models in drug discovery, identifying compounds that work against the many types and causes of arrhythmia, while taking account of genetics and other factors that vary between individuals. Whereas these two goals are some years from being achieved, models have already contributed towards understanding how a few existing drugs for treating heart conditions work.
Side-effects of drugs on the heart
Being able just to screen candidate drugs for dangerous side-effects on the heart will itself confer huge health benefits, saving many lives and reducing costs for pharmaceutical companies. At present, side-effects on the heart are responsible for more than half of all withdrawals of therapeutic drugs after initial approval. Many drugs for treating conditions of all types - including cancer, malaria and psychoses - have been banned because, after proving effective against their primary target, they have subsequently turned out to cause arrhythmias in some patients. Models are already showing potential for identifying which people are at risk from particular drugs on the basis of genetic or other factors.
Although the mechanisms differ, these failed drugs nearly all cause arrhythmia by prolonging the so-called QT interval, which shows up on electrocardiograms (ECGs) of the heart. The QT interval is a complex measure, but equates to the duration of the action potential of pacemaker cells determining the heart's rhythm. The operation of all muscles is determined by the build-up of action potentials within their cells resulting from flows of sodium, potassium, or in the case of heart cells, calcium ions transporting electric charge into and out of cells. Skeletal muscles have to respond quickly to the brain's commands, and so action potentials in their cells have to work fast, lasting just 2-5ms. But the heart beats relatively slowly, and so pacemaker cells have action potentials lasting around 100 times longer, 200-400ms. Such duration cannot be sustained by the flows of sodium and potassium ions used in normal muscle cells, so vertebrate evolution recruited calcium ions, which allow action potentials to be sustained. Then pacemaker cells can deliver accurate timing to the heart muscle cells.
Heart models can now simulate the effect of altering ion channels and disrupting the timing of pacemaker cells that some drugs have. "With cardiac models you can simulate genetic defects causing long QT intervals, but also defects of drugs causing the same effect," says Jeremy Rice, pharmaceuticals consultant at IBM and a leading figure in whole-heart modelling. According to him, this development has convinced many pharmaceutical companies to adopt heart models within their drug development programmes.
This change in attitude has permeated the cardiac research community as a whole. Says Peter Kohl, a leading member of Oxford University's elite cardiac electrophysiology group: "Now heart modelling is not only accepted and tolerated, but the journals and funding bodies are actually quite pleased when it is there."
That change in mindset has given a big boost to research, and has naturally been welcomed by the undisputed pioneer of the field, Denis Noble, who developed the first mathematical heart model as early as 1962 and is often credited as the founder of Systems Biology long before the term itself was invented.
A major challenge
Indeed, according to Noble a major challenge now lies in managing expectations, given that the case for using heart modelling within drug development and potentially treatment has been proven.
"I am cautious," says Noble. "We oversold the Human Genome Project. We must avoid overselling the Physiome Project [loosely the worldwide effort to develop a computational framework for systems biology and modelling at all scales], particularly since it may take a long time."
As Noble points out, one of the immediate tasks is to achieve further dramatic reductions in the time taken to simulate cardiac activity. "One of the greatest challenges is speeding up the computation of the whole organ models," says Noble. "At present, it takes hours to compute seconds of real heart-time.
We are therefore collaborating with Fujitsu to implement codes on the 10 petaflop computer they are building for the Japanese Science Ministry RIKEN. The whole organ models should run in real time on that machine, due in 2011."
That will be a huge step forward, but it raises the question of what is actually being computed in a whole-organ model. Clearly, no model can simulate every single aspect of a human heart: from the movement of every single sodium, potassium or calcium ion during activation right up to the precise amount of blood pumped every microsecond through each artery. Indeed if it could, it would be not a model, but virtual reality. The key, therefore, lies in breaking down the whole heart modelling problem into a hierarchy of simplified smaller models capable of interacting and being individually tuned to deliver results for a particular overall application, whether in drug discovery or treatment. "A model should be as complex as necessary, but as simple as possible," says Kohl.
Bridging the gap
A flexible hierarchical model structure is also needed to support the requirements of diverse applications. As well as drug companies, manufacturers of pace makers are becoming interested in the potential of models to optimise the design and application of their products, according to Matthias Reumann, another leading heart modeller, who joined IBM from the Institute of Biomedical Engineering at the University of Karlsruhe in 2007. "This is not yet at the product stage, but there have already been a couple of studies to look at different ablation patterns [for example to remove bits of damaged heart to restore normal electrical activity]," he says.
Meanwhile, back on the drug development stage, heart models look like they could solve a longstanding problem in cardiac research, which is applying results obtained from mice, rats, rabbits and even dogs to humans.
These animals have hearts that beat faster, are driven by different electrical patterns, and their lives are much shorter than those of humans casting doubt on their relevance for human health.
Models could help bridge this gap between animals and humans, according to Noble. "Some drug companies have asked us whether it is possible to use the models to extrapolate from the rat or rabbit to the human," he says, adding: "Comparison across species in this way is an important application of the models, now that there is a substantial number of them in the model pool."
The next few years therefore could see heart modelling becoming essential for development of drugs and pacemaking devices, even if it does not arrive in the clinic quite as quickly. "I am confident we will usher in a new type of medical science, but the rewards will come progressively, not in an avalanche," concludes Noble.