GNN - Genome News Network  
  Home | About | Topics
Making predictions about cancer outcomes


One of the toughest tasks faced by physicians is assessing which cancer patients are likely to respond to treatment and which are not. Doctors currently rely on clinical information such as the size of the malignant tumor to make prognoses, but new gene-based strategies are being developed with the aim of improving the accuracy of predictions.

A cluster diagram from the microarray study.

The results of one such effort have just been published. Using DNA microarrays and algorithms, researchers were able to classify cancer patients into two groups with very different five-year overall survival rates (70 percent versus 12 percent). The subjects had diffuse large B-cell lymphoma (DLBCL), the most common lymphoid cancer in adults. This cancer kills half of the affected patients within a few years of their diagnosis.

Todd R. Golub, of the Dana-Farber Cancer Institute in Boston, Massachusetts, and colleagues used the microarrays—glass slides spotted with some 6,800 genes—to measure the activity of thousands of genes in 58 tumors. After an algorithm sorted the data, the researchers discovered patterns of gene expression associated with better or worse clinical outcomes.

The analysis uncovered some genes whose expression patterns indicate that they are important factors in the disease. These are potential targets for anti-cancer therapies. Tumors frequently are the result of overactive genes that trigger the uncontrolled growth of cells.

"This microarray classifier is a better predictor of patient survival than the International Prognostic Index," observe the authors of a News and Views piece that accompanies the study in Nature Medicine. The index is used to stratify DLBCL patients according to their likely survival on standard chemotherapy and select patients for intensified treatment. Laura J. Van't Veer and Dahne De Jong, of The Netherlands Cancer Institute in Amsterdam, wrote the commentary.

See related GNN article
»Using molecular tools to diagnose cancer

. . .

Shipp, M.A. et al. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nature Med 8, 68-74 (January 2002).

Back to GNN Home Page