Computer program can predict breast cancer risk

Early detection of cancer risk is of the utmost importance because treatments are more likely to succeed early in the course of the disease, before the cancer spreads throughout the body. One way in which scientists are increasingly hoping to detect cancers early is through genetics.

Some cancers, including breast cancer, can be inherited or passed down from generation to generation. Either parent may carry and pass on a gene that increases the risk of developing cancer.

New methods are being developed to help doctors figure out whether a patient is likely to have inherited a cancer-causing gene, and if so, how severe it might be and what method of treatment will be most effective.

Researchers from several institutions, including Hopkins, have collaborated to create a computer tool that may one day be implemented in cancer screening processes. The research team was lead by Rachel Karchin, an assistant professor of biomedical engineering at Hopkins.

The tool, explained in the Feb. 16 issue of Public Library of Science Computational Biology, investigates certain "predictive features" of a person's gene sequence to help determine whether certain mutations are harmless variants or whether they cause an increased risk for cancer development.

The team designed a computer program that can quickly detect and characterize mutations in the BRCA1 gene, a hotspot for genetic mutations that can make some women more susceptible to developing breast and ovarian cancers. In some women, however, changes in the BRCA1 gene are harmless.

BRCA1 mutations are capable of causing cancer because their protein products are responsible for turning on TP53, a tumor suppressor gene. Tumor suppressor genes are responsible for keeping in check the cycle of cell growth known as cell proliferation.

When harmful mutations occur at the BRCA1 locus, TP53 does not function, allowing uncontrollable cell proliferation, which leads to cancer.

The purpose of the computer program is to rapidly search through many mutations, picking out those that are potentially harmful. The program has been validated with up to 94 percent accuracy and can provide results quickly, making it ideal for mass testing of many women.

The program works by identifying a set of features that are characteristic of those mutations that are potentially harmful. The calculations focus on the likelihood of any given change in the gene, and the possible impact of that change on protein function.

"The list of features is derived from computational simulations of how the amino acid substitution impacts protein structure, properties of amino acids themselves, and patterns of evolutionary conservation at the sites of substitution," Karchin said.

By tracking evolutionary conservation, scientists can determine how long a given mutation as been prevalent in the human gene pool. Advantageous genes are usually selected, thus becoming more prevalent in the gene pool as predicted by Darwin's theory of evolution by natural selection.

Mutations that have been preserved over multiple generations are usually advantageous rather than harmful. If a mutation is highly evolutionarily conserved, it is unlikely to cause disease.

By putting all of these factors together, the program compiles the chances that mutations are risky enough to potentially cause cancer.

These results alone are not enough to conclude or predict that cancer will in fact occur in a patient. However, the computer program is able to provide valuable information that can supplement traditional testing, which may fail to be conclusive or may require too much time and money.

These new findings will contribute greatly to the ability of medical teams to predict cancer risk. The tool targets the high-risk population of women who already suspect that they are more susceptible for breast or ovarian cancer based on their family histories.

"We hope that it will be useful to genetic counselors, but expect that they will also take into consideration a woman's family history, age, ethnic background and environmental factors before making any decisions," Karchin said.

Eventually the program may prove to be a valuable tool in the detection of breast and ovarian cancer. It is possible that such programs will be adapted and expanded to include information on mutations at more than one gene locus, and therefore be used to predict the risk of other inherited cancers as well.