max planck institut
informatik
informatik
|
A methylated DNA molecule. Aberrant methylation in the human genome can cause cancer. | Cancer is a genetic disease Cancer is the classical example of a complex genetic disease. In that sense, cancer differs from several other groups of diseases:
|
 | |
How cancer works A simple model of cancer is as follows: Somewhere in the body a single cell or a small set of cells undergoes a genetic transformation that allows it to overcome natural growth barriers. Then, this cell proliferates and further errors happen, one by one, giving the cancer the power to grow infinitely, to invade neighboring tissue, to spread metastases(*) all over the body and, possibly, to resist chemotherapeutic anti-cancer drugs. Eventually, the tumor growth interferes with vital organs. | |
| Cancer cells overcome natural controls and multiply exponentially |
Implications for treatment
The classical understanding of cancer is that all genetic errors during tumor development involve major changes to the DNA (such as mutations, loss or duplication of critical chromosomal regions), which are irreversible, by design. Therefore, clinical cancer treatment in the form of surgery and cytotoxic chemotherapy is currently aimed at extracting or killing all cancerous cells before the collateral damage of the treatment kills the patient.
The role of epigenetics(*)
During the last few years, the dogma of irreversible genetic change at all key steps along the route of cancer progression has increasingly been challenged. In particular, researchers have found a wealth of evidence for the fundamental role of so-called epigenetic modifications in cancer cells and it is believed that these effects may contribute as much to cancer as mutations.
This emerging epigenetic model of cancer has two interesting implications. First, it provides natural explanations for many findings that are difficult to reconcile with the classical DNA mutation concept (such as the strong age-relatedness of cancer which contrasts the often fast progression after initial emergence). Second, it holds a great promise for cancer treatment: epigenetic modifications of cancer cells are reversible, in principle. Hence, it may not always be necessary to kill or extract all cancerous cells. Instead, drug therapies could recreate a normal, non-malignant epigenetic state in cancer cells.
Epigenetic cancer drugs
First steps towards epigenetic cancer drugs have been taken by several laboratories and pharmaceutical companies. One drug has already been approved for the US market, if only for a very specific leukemia-related disease.
These drugs typically act upon DNA methylation(*), and they are characterized by low degree of specificity. That is, they not only reverse the cancer-related epigenetic modifications in tumor cells, but they also modify somatic DNA methylation critical for normal cell development and they may also have a detrimental effect on the genome of potential offspring (just like classical chemotherapy).
In order to improve on the benefit of epigenetic cancer therapy, we have to strive for a better understanding why certain areas of the DNA become aberrantly methylated in cancer, while others become normally methylated in blood, lung, or brain tissue and yet others seem to evade methylation in all tissues. Only with such knowledge, we can hope to tailor more specific drugs with much fewer and less severe side effects in rational way.
Our contribution: computational epigenetics(*)
Our mission is to provide a set of bioinformatics tools that drive the research towards a genome-scale understanding of cancer epigenetics, with the ultimate goal of better therapy. We coined the term “computational epigenetics” for this approach.
Initially, we have addressed the lack of data standardization and quality control in the most widely used method of analyzing DNA methylation experimentally. To that end, we teamed up with the epigenetics group at Saarland University headed by Prof. Dr. Jörn Walter , we defined clear standards for data format and quality control, and we developed the comprehensive yet easy-to-use Java software BiQ Analyzer that guides the experimenter through the key steps of data control for the experimental analysis of methylation in DNA sequences. Furthermore, the software supports the analysis by providing interactive feedback and expert recommendations. This tool is already in use in over a hundred epigenetic laboratories worldwide.
Having tackled the data quality issue, our next step was to gain understanding on the genome-wide distribution of DNA methylation in healthy cells. Indeed, if we want to understand what is wrong with cancer cells, we need to know how a whole genome should epigenetically look like in a healthy cell. To that end, we developed a web-based data mining tool that allows for scoring DNA methylation patterns in blood lymphocytes against a wide variety of information on the human genome that is available at the genome-wide scale. This tool makes use of a variety of statistical learning methods (e.g. support vector machines) and it allowed us to identify three groups of attributes that play an important role for normal DNA methylation: DNA sequence, repetitive DNA motifs, and predicted DNA structure. While experimentalists will find it interesting to uncover the mechanisms that underlie the correlations that we identified between these attribute groups and DNA methylation, we focused on the predictive power of the method: Taken together, these attributes enabled us to predict DNA methylation at above 90% accuracy, both in a cross-validation and in an experimental validation. In the absence of suitable genome-wide DNA methylation data, such predictions are anticipated to be a useful point of reference for the analysis of aberrant methylation in cancer.
We are currently extending our work towards comparing the methylation patterns in cancer cells to those in normal cells and we analyze if we can predict preferential locations of erroneous modifications. Furthermore, we will evaluate genome-wide data with respect to the effect of one exemplary epigenetic cancer drug for unknown patterns of sequence or position specificity that may provide a starting point for improvement towards more specific drugs with fewer side effects.
Further reading
Contact
We are happy to talk with you! Please contact Christoph Bock (http://www.mpi-inf.mpg.de/~cbock/) in case of any questions, remarks or ideas, or if you just want to know more about how epigenetics may change our view of cancer and what contribution bioinformatics can make to this process.
Bibliography
Glossary
| Computational Epigenetics | Development and application of bioinformatics methods (data mining, statistical learning, pattern recognition, etc.) to biologically or medically relevant epigenetic problems. |
| DNA Methylation | An important epigenetic modification of the DNA, by which cytosine bases become methylated, often leading to deactivation of closeby genes. |
| Epigenetics | Heritable modifications of the genome that do not involve DNA sequence changes. |
| |
| Metastasis | The ability of a tumour to spread to other parts of the body. |
| Created: | Uwe Brahm/MPII/DE, 02/14/2006 12:50 PM | Last modified: | Uwe Brahm/MPII/DE, 02/22/2006 09:05 PM |
