The patient, let’s call her Irene Mayer, has a severe cough that just won’t go away. To be on the safe side, her doctor refers her to a lung specialist. The specialist tests a blood sample for signs of lung cancer. The results indicate, with an 85 percent probability, that Irene has lung cancer even though the 61-year-old has never smoked. A lung endoscopy and a DNA test provide conclusive evidence and information for a treatment plan.
A few days later after Irene is admitted to the hospital, the principal consultant and his team at the cancer unit gather for their weekly tumor conference. Colleagues from other wards join via videoconference. The team discusses the most urgent cancer cases and decides what treatments they will use. With only one hour for four patients they have to work efficiently.
Extracting information from the entire patient database
After a while, Irene’s case is up. Looking at a split screen where all relevant patient-specific information about Mayer is displayed, the doctors start making cross-comparisons. How were other patients with similar diagnoses treated? What was the average prognosis? Which treatments have had the best success rates? In just a few minutes, the doctors extract key information from the entire patient database about current and former lung cancer patients and thus get a first valuable indication of possible treatment options. But Mayer’s case is complex. The DNA analysis shows that she has a very rare combination of gene mutations. Standard treatments would most likely not help her. There are, however, new treatments available via clinical trials. Mayer’s particular details meet the eligibility criteria for several of these trials.
So the doctors take the next step. The knowledge gained from decades of cancer research and the patient’s data have been fed into a complex mathematical model which allows the computer to produce a virtual twin of Irene Mayer. All possible medical treatments, including those still in the trial stage, are then simulated on Mayer´s virtual self. The computer uses her “avatar” to determine whether a particular treatment will be effective and what side effects it may have. Fifteen minutes later, the doctors determine that a new type of chemotherapy tested in one of the available trials is the most likely to stop the tumor from growing and keep side effects to a minimum. Mayer joins the clinical trial. The chances of successfully treating her cancer are good.
The above scenario with Irene Mayer is still just a vision. But with eight million people worldwide dying of cancer each year, the pressure is on to make it reality. Studies show that by 2030 cancer will account for 17 million deaths worldwide and a further 80 million people will have the disease. The cost of developing cancer treatments is rising. At the same time, the number of new cancer drugs coming onto the market is declining. Cancer is a very complex disease: there are many differences between individual tumors and they respond very differently to treatments. Furthermore, the currently available drugs are only effective in about a quarter of treatments. Scientists from various disciplines and from across the world are therefore trying hard to increase the success rates in order to avoid unnecessary therapies, reduce patient suffering, and save health-care expenditures.
IT modelling to increase the success rates of cancer treatment
Valentin Flunkert and Morten Ernebjerg are two such scientists. They are not physicians but physicists. Flunkert’s academic career took him from computer simulations to IT. Ernebjerg likewise transitioned from computational biology. Both work on IT to model virtual patients at the SAP Innovation Center in Potsdam, Germany.
Next page: SAP helps bring together knowledge of cancer
“IT is helping to move medicine toward personalized diagnostics and treatment,” says Flunkert. Ernebjerg adds: “Medical research has advanced enormously in recent years through combining techniques and findings from molecular biology, genetics, bioinformatics, mathematics, and systems biology. Rapid progress in human genome sequencing generates vast amounts of data that have to be analyzed using mathematical models and powerful computers.”
A few miles away from the sleek new building of the SAP Innovation Center, Alexander Kühn is working on the systems biological foundations of the virtual model in a former Berlin hospital. What was once a dermatology clinic is now the headquarters of Alacris Theranostics, a spin-off from the Max Planck Institute for Molecular Genetics in Berlin. Alexander and his 15 fellow researchers are working on a model that brings together knowledge of cancer from publications and databases with information about the structure of molecular networks. The model contains data about several hundred genes that have been identified as cancer-relevant, genes which can play a role in the development of tumors.
DNA sequencers help with drug tests
“Every tumor is different,” says Kühn. For instance, a drug that is licensed for the treatment of diabetes or arthritis may help treat someone with skin cancer. “We look at a patient’s genome and transcriptome data. In other words, we look at which of a patient’s genes have changed and which are active. Genes have to be active to influence tumor growth,” says Kühn. As part of this analysis, DNA sequencers help scan a patient’s entire DNA, or certain segments thereof, for typical mutations. The researchers then test various drugs on the virtual patient. Kühn points out: “This could be a combination of drugs in the market place and drugs that are not yet licensed.”
Next page: The Virtual Patient Platform and SAP HANA
Every couple of weeks Alexander Kühn and the developers at the SAP Innovation Center meet to improve the software application called the Virtual Patient Platform. Their vision is to create a system that every doctor can use, in real time, to determine which medication will best treat a patient’s cancer. Such calculation-intensive processes need a powerful platform like SAP HANA.
Kühn expects that more and more hospitals will be using their own sequencers to analyze DNA. If lots of physicians want to access the same data at the same time to create virtual patients, they will need a lot of processing power.
Software assesses treatments tested on virtual patients
Valentin Flunkert and Morten Ernebjerg are working on the software that will meet the above requirements. They are designing a system that “enables doctors to work efficiently and produces results that are intuitive to use,” says Ernebjerg. The software is designed to help doctors, not replace them. For example, it ranks the treatments tested on the virtual patient and the effects of those treatments. How well do the drugs suppress tumor growth? What effect do they have on secondary cancers and on the resistance of the tumor to treatment? What are the likely side effects? “Doctors can evaluate this information before selecting the medication and dosage,” says Flunkert.
Next page: Reducing the cost of developing drugs
Flunkert and Ernebjerg are working closely with other teams at the SAP Innovation Center in Potsdam and other SAP development units in Palo Alto, Vietnam, and Walldorf. This collaboration across teams and geographies also applies to the Patient Data Explorer project which recently received special recognition by the White House Office of Science and Technology Policy for its contribution to accelerating the progress towards personalized medicine. In this project, a team in Potsdam, the newly opened Design and Co-Innovation Center in Walldorf, and a development team there are working on a tool that doctors can use to quickly and efficiently analyze and compare groups of patients, so called cohorts. Until now, comparing data from different cohorts – for example one that has received a new type of chemotherapy and another that has been treated with conventional drugs – has been a laborious and time-consuming process. The Patient Data Explorer, which runs on SAP HANA, does this in seconds. It can also display a patient’s complete medical history, such as that of Irene Mayer.
In Walldorf, Frank Kilian is doing his part putting the diagnosis and treatment of diseases like cancer on a new technological basis with SAP HANA. He explains how SAP HANA integrates unstructured data such as admission notes, diagnoses, and research publications and makes them available in the Patient Data Explorer tool. If combined with historical data and algorithms for DNA sequencing to generate treatment recommendations, Frank adds: “It could spark interest not only from hospitals and physicians, but also the pharmaceutical industry,” as this step forward would significantly reduce the costs of drug development. Only those patients most likely to benefit from a new drug would be selected for clinical trials. Moreover, the pharmaceutical industry might also find new uses for what it calls Fallen Angels – drugs that fail at the trial stage because too few patients responded to them.
Patient Data Explorer to be available as part of SAP HANA for Healthcare platform
Germany’s National Center for Tumor Diseases (NCT) in Heidelberg, the SAP Innovation Center in Potsdam, and SAP employees in Walldorf have developed a prototype of the Patient Data Explorer which the NCT will soon put into live operation. Soon, the Patient Data Explorer shall be made available as a product for other cancer centers and hospitals as part of the SAP HANA for Healthcare platform.
The Virtual Patient Platform is still a prototype. But the underlying model is already showing great potential. For instance, a terminally ill patient, who failed to respond to other treatments, was stabilized over a number of months by a drug developed for an entirely different medical purpose. It was the model that identified this drug as suitable.
Let’s go back to Irene Mayer, our lung cancer patient. Since she never smoked, she is not in the group of patients at high risk of developing this cancer. Yet she has the disease. This shows just how complex cancer is and how hard it is to predict who is likely to develop it. Here too, IT has helped medical science make significant advances.
Cancer diagnostics project at SAP Innovation Center
Joos-Hendrik Boese, a computer scientist and database expert, leads a project known as Proteome-Based Cancer Diagnostics. He and his team, who are based at the SAP Innovation Center in Potsdam, are working with the Freie Universität Berlin on a new method to diagnose serious diseases like lung cancer at an early stage. One promising approach is proteomics. The proteome, which is the human body’s entire set of proteins, can act as a kind of early-warning system of the human body.
“Diseases leave telltale marks on our proteomes,” says Boese. “Researchers in proteomics are working on identifying protein activities that indicate cancer and on finding correlations between protein patterns and certain diseases.”
Mass spectrometers are used to extract information about which proteins are present in a patient’s blood and in what concentrations they occur. These data are collected for a cohort of sick patients and for a healthy cohort and then compared. If many patients in the group of lung cancer patients show the same specific differences in the protein pattern when compared to the healthy group, this difference is referred to as a fingerprint. If this fingerprint is now found in the proteome of a new patient, it can indicate the presence of lung cancer and so provide another way to detect cancer early. Boese points to his laptop screen and explains that the traffic light symbol shows whether there are any indications that this particular patient has lung cancer. The peaks show that the probability of the patient developing the disease is high.
“This information is a preliminary assessment before the doctor orders a lung endoscopy. It is an inexpensive test that can be used to detect cancer early and avoid more expensive and complex diagnosis techniques that can be very unpleasant or even harmful for the patient. Since we can now recognize the fingerprints of diseases like lung cancer, we can develop much more targeted drugs,” says Boese.
Fast analysis of fingerprints with SAP HANA
With a mass spectrum involving some 2.4 gigabytes of data, comparing disease fingerprints with a patient’s data used to take a very long time. Such volumes of data can now be analyzed fast and accurately with SAP HANA. SAP and Munich’s Technische Universität also announced in May that SAP HANA is the technological basis for a database called ProteomicsDB, which, according to Judith Schlegl, aims to contain the data for the entire human proteome. Schlegl, who has a doctorate in biochemistry, is responsible for development support in SAP’s HANA quality assurance team. Together with Boese, she also works on database design, application administration, and project management. Though she points out that: “We still have much more to learn about our cells, about how proteins interact, and their influence on disease,” she believes that SAP’s innovative database technology can help bring major advances in medical research.
Next page: A complete cure for cancer in sight?
Managing cancer a benefit for science and society
Neither Schlegl nor any of the other SAP colleagues we interviewed about personalized medicine would claim that one day there will be a complete cure for cancer. Yet if doctors were able to detect cancer much earlier and to manage cancer as a chronic disease, then everyone would benefit: science, society, and most of all patients like Irene Mayer.
SAP Innovation Center http://www.sap-innovationcenter.com
Virtual Patient Platform
This article first appeared on SAP Milestones