Modeling life

Simulation helps scientists explore and predict

William J. Holstein
18 November 2014

3 min read

Scientists are learning how to use data to predict which patients will respond well to particular drugs. Even more advances are occurring in understanding human cells, organs and – ultimately – the unique variations of each individual’s body.

If a drug is tested in clinical trials on 10,000 people who suffer from the same ailment, it can be rejected by regulators if it has a seriously negative effect on just one of those 10,000 people.

The costs of such a failure are huge, both for society and for the company that invented the drug.

“If the drug was improving the health of 9,999 patients out of 10,000, it means that those 9,999 patients will not benefit from the drug anymore,” said François Képès, who represents Genopole France, a top synthetic biology organization based in Paris, at the BioIntelligence Initiative, a consortium of life science companies, technology providers and research institutes.

To address challenges like these, the consortium has developed a software platform dedicated to collaboration, modeling and simulation in the life sciences industry. From predicting how molecules will react in the human body to understanding the complex functions of organs, virtual modeling and simulation are helping scientists to understand the body, design new treatments and choose the best therapy for each individual.


Researchers expect that the costs of sequencing the human genome will continue to plunge – it now costs about US$1,000 per person, with results available in about 24 hours – making it possible to affordably map the specific genetic makeup of virtually any patient. “Now we have increased our capacity to predict that this person should not take this drug, but other people can,” Képès said.

The implications of the consortium’s work could be enormous for life sciences companies by helping them to precisely define their target populations, ensure the right drugs go to the right people and allow the development of treatments for small groups with unique conditions.


Scientists also are working to build models of organs from massive amounts of data and then asking those models to understand the impact of a particular treatment – without the need for human or animal testing. The Living Heart project, co-sponsored by Harvard University and the Massachusetts Institute of Technology, both located in Cambridge, Massachusetts (USA), is an example (see the video at the end of this article).

Meanwhile, Bernhardt Trout, a professor of chemical engineering at MIT and head of the Novartis-MIT Center for Continuous Manufacturing, is attempting to tackle the challenge of modeling individual processes within cells. Using massive supercomputers, Trout has developed and tested algorithms that predict how antibodies will behave when they reach key surface areas of a protein, known as “hot spots.“

The algorithms, which are based on Trout’s knowledge of the physical and chemical properties of proteins, are simple enough that he can demonstrate them on a laptop computer in a couple of minutes.


Researchers are using software based on Trout’s algorithms to narrow down a large list of promising compounds into a list of the most likely candidates for drug research. His research will impact the three basic stages of drug commercialization – discovery, development and manufacturing. “I’d like to see the industry transformed, ideally within 10 to 20 years,” Trout said.

He calculates that the pharmaceutical industry in the United States alone spends US$200 billion annually on drug discovery and production, a figure he believes his research could reduce by 30%, saving US$60 billion that could be invested in additional drug discoveries. “The industry as a whole should get on the bandwagon,” Trout said.

When creating protein-based drugs, scientists must avoid those with a tendency to aggregate, or clump. Aggregates can reduce the medicine’s effectiveness, make it difficult to manufacture or administer, or potentially trigger an adverse immune response in the patient. Predictive software helps to identify protein aggregation “hotspots” (red areas), alerting scientists to either mutate the protein or substitute a different one. (Image © BIOVIA)


The work that scientists are doing on the building blocks of cells, including proteins and enzymes, are setting the stage for software tools that fully predict how substances will affect human tissue.

“If the software makers could produce software that is predictive, then it would be very disruptive of drug R&D,” said Bernard Munos, a senior fellow at FasterCures, a center of the Milken Institute, a think tank based in Santa Monica, California. “Instead of scratching our heads, we could actually interrogate the software. I could say, ‘OK, I want to inhibit a particular enzyme. Tell me what’s going to happen.’ The software runs a simulation and tells you whether your intervention is likely to produce the desired outcome or whether it will throw the metabolism of the cell out of kilter and kill it.”

Massive increases in computing power and human knowledge lie ahead before that day arrives, Munos believes. But when it does, he said, “we will have transformed biology into a predictive science.”  ◆

Watch simulation of the human heart :

Related resources