Medical scientists have been using computers to model some aspects of the human body, and its interactions with drugs, since the mid twentieth century. Early examples included the development of quantitative structure-activity relationships that relate the properties of a series of molecules to their action at a potential drug target. Another early mathematical model, of the currents caused by ions passing through channels in heart cell membranes, may now be best known as the first clear example of a new way of thinking about physiology, known as “systems biology”. Systems biologists look at physiology in an integrated way, using computers to model it on all levels from molecules to organs and eventually the whole organism. They combine models at different levels into simulations that, given the limitations on both knowledge and technology, are as appropriate and accurate as possible.  

The ultimate dream or goal of systems biologists working with medicine is a complete computer-based model of human physiology, accurate enough to use for drug testing (as just one example). Several international initiatives have been set up to work on this ambitious proposal, among them the European Commission’s Virtual Physiological Human. So far, 15 large projects have been funded through this initiative; another twelve are in the pipeline. The VPH Network of Excellence, also funded by Europe under Framework 7, provides support and coordination for these projects within the human systems biology community in Europe.

In March 2010, European Union health ministers unveiled a joint vision for e-health in 2020, in which the healthcare of the future was described as being “personalised, predictive and preventative”. This vision is, of course, covers much broader issues than just systems biology modelling. However, these words were echoed very closely by paediatric oncologist Norbert Graf in a keynote lecture at the VPH Network’s first international conference in Brussels (VPH2010) in September 2010. Graf, from the University Hospital of the Saarland, Homburg, Germany, presented a clinician’s view of systems biology modelling, with a particular emphasis on oncology. Cure rates for many cancers, including childhood ones, have improved significantly in recent decades, but a large minority remain intractable. Intensifying the trend towards personalised oncology, stratifying tumours by genetic and other characteristics, and predictive oncology, detecting cancer at an early stage, should help tackle those difficult cancers. In pursuit of these goals, systems biology modelling is already entering the oncology clinic. Graf stressed, however, that such models need to be clinically driven, fully validated and, perhaps above all, easy to use if they are to be of value to busy clinicians. Privacy and data protection may also be an important issue, particularly with models incorporating data from individual patients.

Another keynote speaker at the VPH2010 conference, Aleksander Popel of Johns Hopkins University, Baltimore, USA, touched on cancer in a wide-ranging talk on the systems biology of angiogenesis. Popel described developing a three-compartment model of tumour, blood and normal tissue that predicted that injecting the monoclonal antibody bevacizumab (Avastin™) can cause an increase in plasma VEGF concentrations. This counter-intuitive result may go part way to explain why this drug is almost always ineffective against some common cancers.

Talks and posters at the conference covered a number of projects that focus on cancer. Graf is a partner in the “Contra Cancrum” project coordinated by Konstantinos Marias from the Foundation for Research and Technology Hellas (FORTH) in Crete, the name coming from the Latin for “against cancer”. This network of eight institutions from six countries is developing computer-based models for tumour development and for the response of both tumour and normal tissue to various treatment modalities, with the ultimate goal being optimised treatment planning for individual patients in a predictive-medicine context. These multi-levels models, developed mainly by Georgios Stamatakos from the Institute of Communication and Computer Systems (ICCS) in Athens, are collectively termed the Oncosimulator. Data from cancer patients – including gene expression and imaging data, and the results of treatments – is fed back into the simulator, modelling and validating tumour response to treatment in individual patients. The project is concerned particularly with lung cancer and glioblastoma, and uses data from series of patients with these conditions.

Several projects funded through the VPH initiative are concerned with the diagnosis and prediction of cancer. The HAMAM project, coordinated by the European Institute for Biomedical Imaging Research in Austria, is combining breast images from different modalities with clinical data into models that can be used to identify and diagnose breast cancer precisely at an early stage. Digital images of breast cancer cases are being combined with the associated biopsy data and collated into a database that can be used as a diagnostic tool.   

Another VPH project concerned with cancer prediction and early detection is NeoMark, which specifically develops tools for the early detection of oral cancer recurrence. Oral cancer is a relatively common tumour; it is often curable, but in between a quarter and half of all cases the cancer recurs, and recurrent disease has a much poorer prognosis. It is therefore important to detect recurrence as early as possible, preferably before clinical symptoms become obvious. The nine partners involved in NeoMARK are pooling heterogeneous molecular and clinical data from patients diagnosed with this condition with the aim of detecting disease profiles that are able to stratify patients into high and low risk groups in terms of cancer recurrence. So far, thirty-two genes with expression patterns correlated with prognosis have been identified using data from a relatively small sample of patients. These genes and any others that may be identified using a larger dataset will be transferred to a microarray. The partners hope that this so-called “lab on a chip” may eventually become a routine diagnostic tool.

Stephen Payne, of the University of Oxford, UK gave a talk at VPH2010 that clearly illustrated the value of systems biology modelling in clinical oncology at a very practical level. IMPPACT (or Image-based Multi-scale Physiological Planning for Ablation Cancer Treatment) is developing a physiological model of the liver that will help plan treatments directly. Radio-frequency ablation is a minimally invasive alternative to surgery that kills tumour tissue through hyperthermia; its complexity reflects the anatomical complexity and variability of the liver. Its success depends critically on the skill of the clinician involved. Researchers within IMPPACT are developing computational models of liver tumours and normal liver tissue based on anatomical, metabolic and molecular data to predict how the radiation from the ablation needle will dissipate and, therefore, how the tissue temperature will change. This can be used with an individual patient’s data to derive the optimum placement of the needle to kill as much tumour and as little normal tissue as possible.

These results from a handful of varied projects show that systems biology modelling has much to offer the clinical oncology community, with some already showing clear practical use in the clinic. Computational modelling has already come a long way since those first ion channel models, half a century ago.