Pictures of the Future     Spring 2005

Molecular Detectives

Siemens has developed a software-based technique that analyzes and simulates the gene activity in cells. This could enable the development of new, more individualized therapies. In particular, the pharmaceuticals industry plans to tailor medicines to groups of genetically similar patients.

Human cells are like tiny factories. Each holds some 20,000 to 25,000 genes, Each gene is continually being switched on or off, while genetic information is being read out. Messenger substances migrate to production sites that manu­facture the required type of the roughly one million proteins used by human beings. If any part of this finely tuned mechanism fails to work, functions through­out the body can deteriorate and death can follow. So it‘s no wonder that bioscientists go to enormous lengths to understand the molecular processes in cells and find key genes so they can develop new and better medicines.

"We ask ourselves, for example, what genes are more active in a breast cancer cell than in healthy cells," says Dr. Martin Stetter of Siemens Corporate Technology. As one of the biggest players in the market, Siemens’ medical technology segment is making increased use of bioinformatics, a key com­ponent of future health care. Stetter’s team has developed a method—BioSim—that could prove to be a valuable tool for the pharmaceutical industry in its search for new agents.

BioSim is currently being tested in collaboration with bioinformatics company Biomax. "We’ve created a mathematical simulation model that makes relation­ships between genes visible using known data, and identifies genes related to illnesses," says Stetter. To accomplish this, Stetter, who is a biophysicist, uses data from "gene expression analysis," which depicts the activity of genes in a cell at a certain point in time.

Modern genetic research has concluded that many medications are over pre­scribed. Because of genetic differences, for example, "beta blockers" that lower blood pressure have no effect in up to one third of all patients. Side-effects emerge with varying degrees of intensity depending on genetic disposition—or do not appear at all. For manufacturers, the financial risks associated with these differences are enormous. Experts believe that, in the future, the pharma­ceuticals industry will no longer promote "blockbuster" medications as vigorously as before but instead back medicines for patients who have similar genetic characteristics. The U.S. Food and Drug Administration is considering taking account of data from genetic analyses when approving drugs.

Virtual Cancer Patients. Stetter’s team proved that valuable information can be obtained from BioSim analyses by examining a special form of leukemia. In this instance, the team used data from St. Jude Hospital in Memphis, Tennessee. In 327 patients with acute lymphoblastic leukemia (ALL)—an illness that affects primarily children—they studied the activity of about 270 relevant genes. The resulting expression patterns—snapshots of cells as it were—result from the analysis of the messengers in blood cells. This "mRNA" is sent by genes to protein factories. The presence and concentration of mRNA allow inferences to be made regarding the activity of associated genes. The mRNA samples are marked with fluorescent dyes and applied to a biochip, where they bind to specific molecules. When optically stimulated, these molecules emit light signals. Since the position of the binder molecules is known, the light pattern shows the gene activity at the moment of sampling.

Based on the resulting data, Siemens experts used computers to construct a "Bayesian network" that makes the relationships between the genes immediate­ly visible. If gene A influences gene B, for example, this is illustrated by a connecting line. More complicated relationships can also be derived from the network, such as the fact that gene A and gene B switch on together and thus activate a third gene C. For this, Stetter’s team evaluates the structure of probability tables.

In BioSim’s simulation, the researchers increase or decrease the activity of individual genes and observe the effects. If a gene is shut down virtually, this could change the entire network and therefore the activity of many other genes.

During one of these virtual experiments, the researchers discovered a key gene for the acute lymphoblastic leukemia subtype E2A-PBX1—without knowing its biological contexts. The subtype stood out because of its unusually large number of relationships in the network; hence it may control many other genes. Next, the researchers permanently activated this gene in the computer experiment and studied its effects on the activities of the other genes in the context of their network.

The researchers found that in nearly all of the 327 patients an expression pattern occurred that was strikingly similar to that of the ALL subtype E2A-PBX1. The BioSim result matched actural observations by doctors. Through bio­molecular studies, they identified the same gene as a potential cancer trigger, indicating that the PBX-1 gene is the decisive factor in about one seventh of ALL cases. If it becomes attached to another gene through a mutation, it remains active at all times and inevitably triggers leukemia.

Growing Interest. The experiment convinced Dr. Klaus Heumann, president and CEO of Biomax. "For us, that was the proof that the simulation of genetic networks really works. After that, we finalized our partnership with Siemens." Biomax, which is based in Martinsried near Munich, is enhancing the genetic networks identified by Siemens with additional information on biochemical relationships.

Through a link to a huge database, future users will be able to click on a certain gene and immediately find out in which organs it is especially active, which genes are connected to it, and what technical literature exists on it. Biomax was founded in 1997 and now employs about 100 scientists and IT specialists worldwide. The company is financed with venture capital and is aiming for annual sales above ten millions euros—primarily with bioinformatics solutions for gene and protein analysis.

Biomax has introduced the BioSim simulation tool to a number of large pharma­ceutical companies. The response has been uniformly positive. How­ever, before the companies make their confidential genetic research data available, they would like additional demonstrations.

An initial test is to be carried out this year using mice. For this test, Siemens and Biomax want to work together on the development of a model that depicts the gene expression of mice with a genetic disposition to a certain tumor. Development of such a model would make it possible to derive experiments that could be run on real mice for verification—a procedure that is out of the question for human patients but is expected to provide the decisive proof of how powerful BioSim is.

The potential applications from such studies could be numerous. "Pharma­ceutical companies could immediately identify side-effects with BioSim," says Siemens researcher Stetter. That would require taking samples from persons who suffer from a certain illness. With the simulation, specific genes could then be switched on or off—a process that corresponds to a virtual medicine. In the network, it is possible to read which genes are being impaired and what the effects would be on real patients. New candidates for medicines could be found in a similar way, by measuring the effects of an agent in the gene net­work.

In the future, BioSim could also be used to document the impact of medicines more effectively. "Right now, research practically stops once a medicine is on the market," says Biomax CEO Heumann. If these data were universally recorded and evaluated, medications could be further optimized.

"This information has a great deal of value that isn’t being used today. Our simulations could change that picture very significantly," says Heumann.
                                                                                              Norbert Aschenbrenner