Computing technologies are providing powerful tools to investigate the development of cancer, diagnose the disease at an early stage, and select an appropriate treatment plan.

A technician prepares a patient for a mammogram, which can reveal signs of breast cancer-the leading type of cancer affecting women.

doctor in Virginia, relying on the latest available technology, diagnosed two of his patients with what appeared to be an early stage of the same type of cancer. He surgically removed the cancerous tissue from both patients and placed them under the same follow-up regimen of radiation therapy. One of them recovered fully, while the other continued to suffer from the disease and had to receive additional treatment.
        This example is not unique in health care, particularly in cancer care. For cancer is an extremely complex disease, and the formulation of a treatment plan appropriate for each patient often eludes even the most brilliant doctors.
        What, then, is cancer? It is a term applied to a group of related diseases that affect a variety of organs in the body. Usually, each form of cancer is named after the type of organ or group of cells where it originates. The common denominator for the various types of cancer is that the affected cells are abnormal and multiply out of control.
        Moreover, each type of cancer develops in stages, following a progressive pathway. During the early stages, the cancerous tissue is localized, and treatment of the disease is more likely to succeed. Over time, however, some of the cancerous cells may break away, enter the bloodstream or lymphatic system, and spread to other parts of the body--a phenomenon known as metastasis.
        Unfortunately, most cancers are detected during their advanced stages, when it is nearly impossible to effect a cure. Medical researchers have therefore sought to develop methods that allow early diagnosis and tracking of the disease. In so doing, they are increasingly relying on computers to perform important tasks.
        For example, computers are an integral part of imaging devices such as CT (computerized tomography) and MRI (magnetic resonance imaging) scanners, which are used to distinguish cancerous tissue from normal tissue in a patient's body. These techniques, which produce three-dimensional images, have been refined to reveal tumors as small as a few millimeters in width.
        On another level, scientists have been using computers to help them understand the molecular mechanisms that underlie the development of cancer. Having discovered a variety of cancer-related genes and proteins, they have found it necessary to sort through enormous quantities of data and build mathematical models to predict the course of the disease. For this purpose, they have turned to the field known as bioinformatics--a field in which computers and computational methods are used for the organization and analysis of biological data [see "The Biologist Meets the Computer Scientist," The World & I, March 2002, p. 136].

Cancer-related genes

he first step in the development of cancer is referred to as the initial insult. It is thought that in most cases, the initial insult is a process that introduces one or more mutations in certain genes. The genetic material consists of DNA (deoxyribonucleic acid), and genetic mutations correspond to alterations of the DNA structure.
        Some mutations may be inherited, others may arise randomly, and yet others may be caused by exposure to ultraviolet radiation or harmful chemicals. If the mutations occur in genes that are involved in the control of cell growth and division, then those gene functions may be disrupted and the cells may grow and multiply abnormally, leading to cancer.
        Researchers have identified a number of genes that, when mutated, influence the development of various

A researcher uses a computer to analyze gene chips, each of which consists of thousands of DNA fragments on a glass or silicon plate. By this technique, DNA samples can be rapidly screened for cancer-related genetic mutations.

cancers. For example, mutations in the BRCA-1 and BRCA-2 genes are associated with certain types of breast and ovarian cancers, while defects in the p53 (or TP53) gene are linked to many different tumors. Generally, each type of cancer is associated with mutations in several genes, and each gene may be mutated at more than one site.
        Based on this knowledge, genetic tests have been developed to determine whether a person's DNA happens to carry cancer-related mutations. For the rapid screening of large numbers of genes and their mutated variants, researchers use DNA microarrays, often called gene chips [see "Genes on a Chip," The World & I, September 1997, p. 189].
        Each microarray consists of thousands of DNA fragments (of known structures) attached to a glass or silicon chip. The DNA that needs to be tested is extracted from the appropriate tissue, tagged with a fluorescent dye, and evaluated for its ability to bind (hybridize) specifically to fragments on the chip. Each chip yields an enormous amount of data that must then be processed by algorithms in a computer, eventually revealing the identity of mutations in the test DNA. In most cases, the genetic information is useful only if the gene's function has been studied and understood.
        The presence of cancer-related genetic mutations is one factor that increases the risk of a person developing the disease, but it generally does not predetermine the onset of the disease. Other factors that increase this risk include the individual's age and lifestyle--such as smoking or heavy drinking.

Cancer-related proteins

hile genes are important, they do not directly participate in carrying out the cell's functions. Rather, they contain the information from which proteins are synthesized, and it is the proteins that are directly involved in many of the body's functions [see "Proteins, Proteins, Everywhere," The World & I, May 2002, p. 130]. Some proteins are structural components of cells; others function as enzymes, hormones, or antibodies.
        The synthesis of each protein occurs through a complex mechanism, in which the information within a gene is first used to synthesize RNA (ribonucleic acid); this RNA then directs the formation of the corresponding protein [see "Unraveling the Human Thread of Life," The World & I, September 2001, p. 136]. The protein may then undergo a variety of modifications by additional processes in the cell.
        If a gene becomes mutated, its protein product will be structurally altered, and the protein may be unable to function normally. Alternatively, if the gene is normal but the protein-synthesizing machinery is defective, then the resultant protein's structure and function will be adversely affected. Thus, to track the development of a cancer, it is important to check the proteins associated with the disease.
        Some proteins, known as tumor markers, are present at elevated levels in the blood, urine, or tissues of individuals with

Cancer cells that originate in one part of the body may break away and metastasize (spread) to other parts. Here we see breast cancer cells (brown) that have metastasized to the lymph nodes (top) and liver (bottom). The tissues were stained using a technique known as immunohistochemistry. Inset: A breast cancer cell is visualized at high magnification with a scanning electron microscope.

certain types of cancer. Those found in the blood include PSA (prostate-specific antigen), for prostate cancer in men, and CA 125 (cancer antigen 125), for ovarian cancer in women. An elevated level of a marker in the bloodstream suggests but does not prove the existence of cancer because it may occur in noncancerous conditions as well. Moreover, many tumor markers are associated with more than one type of cancer, so their detection in the blood does not indicate the location of the cancerous tissue.
        For these reasons, it is important to test proteins in tissue samples obtained by biopsies. In this process, proteins that occur in cancerous tissue need to be compared with those in normal tissue from the same patient and organ, to look for abnormalities that may otherwise be missed.
        One common technique used in identifying proteins in tissue samples--both normal and cancerous--is known as immunohistochemistry (IHC). In this case, the researchers begin by preparing antibodies that will bind specifically to each type of protein that they wish to detect. The antibodies (which are tagged with a dye or other type of label) are allowed to interact with the tissue sample, and the tissue is stained to reveal the sites of antibody-bound proteins. Using this technique, the development of cancer can be tracked by examining the tissue for tumor-specific protein markers.
        Recently, Dr. William Grizzle of the University of Alabama Medical School in Birmingham developed a new adaptation of IHC, by which the concentrations of proteins in tumor cells can be determined. In this method, an image of the stained tissue (after antibody binding) is scanned into a computer, which then evaluates the stain intensity, percentage of cells stained, and the protein concentrations. This method has sparked new research because the concentrations of protein markers are thought to reveal the aggressiveness of the disease.
        Some important protein markers that are tracked using IHC include p53, BRCA-1, and BRCA-2, which are specified by the p53, BRCA-1, and BRCA-2 genes, respectively. Under normal conditions, the p53 protein helps inhibit the growth of tumors. When its structure is altered (by mutations in its gene), it may lose this function. Deformed versions of the p53 protein have been found in most solid tumors.
        The BRCA proteins have several functions, including the repair of damaged DNA and the regulation of RNA synthesis from various genes. When the BRCA-1 and BRCA-2 genes are mutated, the corresponding proteins may not function properly. As a result, damage to cellular DNA may go unrepaired and the cells may become cancerous.
        It should be noted that the detection of a single defective protein marker does not give sufficient information. Cancer progresses in a complex, nonlinear fashion, and multiple protein markers need to be checked to grasp the behavior of the disease. Moreover, the structure of each marker may be modified at a number of sites, producing different variants of the protein.
        For the rapid testing of many proteins or many variants of the same protein, researchers now use protein chips (protein microarrays), which are modeled after gene chips. Each chip carries an array of sensors that function like antibodies--that is, each sensor can bind specifically to a protein (or protein fragment) with a particular structure. The protein sample that needs to be tested is allowed to interact with the chip under special conditions, and the markers (or their fragments) can be identified from their pattern of binding to the chip's sensors. Here again, computers are needed to evaluate the binding patterns.
        While the above methods of screening individuals for cancer-related genetic and protein markers provide useful information for the diagnosis and tracking of cancer, they also raise some social and ethical issues. For instance, if a person who seems healthy is found to carry cancer-related gene mutations or defective proteins, indicating a predisposition to cancer, he needs to receive the information with appropriate counseling and emotional support. In addition, this information needs to be kept secure, so it is not used to discriminate against him when he applies for insurance or seeks medical care. To protect the person's identity, many research institutions separate his medical data from his personal information.

Pathways of cancer development

he identification of protein markers does not provide sufficient information about the disease they are associated with. Scientists also need to know the functions of these proteins, how they relate to one another, and how they participate in the steps of cellular growth and multiplication. The course of these steps is described as a pathway. If a clinician is able to predict the pathway along which a patient's cancer will progress, he would be able to select a treatment plan with the best chance of success.
        Computer-based technologies--such as data mining, decision trees, mathematical analysis, and simulations--offer assistance in discovering, understanding, and predicting these pathways. In the pool of data produced from a single tumor, not all the points are useful, and patterns or abnormalities need to be identified. Data mining is the process of evaluating a large data set and extracting relationships between data points, as a way to obtain useful information.
        A decision tree is a means of organizing the information to help the physician decide on a treatment plan for the patient. If two patients with the same symptoms have the same protein markers at approximately equal concentrations, both individuals would be placed on a common branch of the decision tree and assigned the same treatment plan.
        An example of such a case was given at the beginning of this article. In that example, the two patients were initially placed on the same branch of a decision tree, but they responded differently to their treatments. Consequently, the patient who did not recover had to be moved to another branch of the decision tree, and researchers had to look for a new, distinguishing protein marker.
        The data sets obtained from looking at protein markers may also be subjected to mathematical analyses to extract useful information. In this regard, significant research is being done to identify statistical methods that describe relationships between protein markers. A nonstatistical method, called artificial neural networks (ANN), has also met with success. An ANN system uses artificial intelligence to learn patterns within the data presented to it, and it may help identify relationships (between data points) that may be overlooked by typical statistical methods.
        When we see a pattern, our mind has the ability to remember it and identify a similar pattern on another occasion. Likewise, an ANN system can be trained to learn different sets of data about cancer patients, along with the corresponding treatments and outcomes. The system creates connections within its memory that even the scientist is unaware of. Later, when presented with new data, the system will associate the information with certain patterns that it has recognized and will predict outcomes and suggest treatments.
        Some scientists have been using high-speed computers to develop simulations of biological pathways. For example, Mandri Obeyesekere of the M.D. Anderson Cancer Center at the University of Texas has published a complex mathematical model that simulates the behavior of human cells, including some mechanisms of cancer development. Such simulations are difficult to formulate because of the large number of proteins in each cell. Obeyesekere's model has a number of limitations, but it can mimic most known behaviors of early cell growth. Its potential usefulness ranges from early diagnosis of cancer to treatment evaluation and development.
        Recently, Emanuel Petricoin III of the Food and Drug Administration published a system in which computer-based analysis of protein patterns can help with the early detection of ovarian cancer. More than 80 percent of women who are diagnosed with ovarian cancer are already in a late stage of the disease, when the survival rate is only 35 percent. On the other hand, if the cancer is detected at an early stage (stage I), the survival rate reaches 90 percent.
        Petricoin and his research team started with a group of 116 women, 50 of whom had ovarian cancer while the rest did not. They analyzed the women's blood proteins using protein chips and a technique known as mass spectrometry (to read the microarrays). They also trained a computer algorithm (similar to an ANN system) to recognize protein patterns that were associated with ovarian cancer. By this method, they correctly identified all 50 cases of ovarian cancer, and the computer found a protein pattern that was specific for early-stage ovarian cancer.
        Thus we see that computer-aided analysis of multiple protein markers can help with the diagnostics and treatment of cancer, as well as improve our understanding of the development of cancer through different stages. The new methods, however, are limited by the amount of data collected from cancer patients. The development of each new data set from a patient commonly takes at least five years.
        As the median age of our population rises, the need to deal with cancer is likely to become a more pressing issue. New technologies will be eagerly sought to detect the disease as early as possible and predict its behavior. In addition, computers will be increasingly used to model cancer-related molecules and their interactions and to design cancer-fighting drugs.
        The recently developed systems are available to many researchers, but they have yet to gain acceptance by practicing physicians. Dr. Don Colvin, a colorectal surgeon, offered this assessment: "If the new techniques don't change how I treat a patient, then the research is useless in a clinical setting." In other words, research results should lead to new treatment options or significant modifications of current treatment programs. Toward this end, many bioinformatics projects have begun including medical doctors in the acquisition and analysis of data.
        The above examples are just a small slice of the pie, as many additional technologies are in the works. Someday soon, computers may finally help scientists find the way to cure cancer.

On the Internet

CancerQuest
www.cancerquest.org

National Cancer Institute
www.cancer.gov

OncoLink
www.oncolink.upenn.edu