ISPO

Published in Cancer Detection and Prevention 2000; 24(Supplement 1).

Molecular pattern recognition in cancer

TR Golub

Whitehead Institute/MIT Center for Genome Research, Boston, MA todd_golub@dfciharvardedu, golub@genome.wi.mit.edu

The availability of DNA microarrays has made it feasible to measure simultaneously the expression levels of thousands of human genes. We have developed a series of analytical approaches for the identification of biologically and clinically meaningful patterns in such complex gene expression data. We tested these methods on a series of 72 human bone marrow samples obtained from patients with new diagnosis acute leukemia. We measured the expression levels of 6800 human genes in each of the samples. Clustering of the dataset using Self-Organizing Maps demonstrated the automatic identification of myeloid, T-cell and B-cell acute leukemia subtypes. Supervised learning prediction algorithms were also capable of establishing the correct leukemia subtype diagnosis with 100% accuracy, without the benefit of prior biological understanding of this distinction. We have also extended this approach to the classification of lymphoma and brain tumors, and have also used expression profiling to identify critical mediators of melanoma metastasis. Specifically, we demonstrate that the expression of the GTPase RhoC is correlated with metastatic potential in murine models of melanoma. Finally, we have demonstrated the feasibility of chemosensitivity prediction in human cancer cell lines based on gene expression pattern recognition in untreated cells. These results suggest that DNA microarray technology coupled with pattern recognition algorithms represent a feasible strategy for providing insight into cancer diagnosis and pathogenesis.

KEY WORDS: DNA microarrays, pathogenesis, T-cell, B-cell, melanoma, metastasis.

For more information, contact golub@genome.wi.mit.edu

Paper presented at the International Symposium on Impact of Biotechnology on Cancer Diagnostic & Prognostic Indicators; Geneva, Switzerland; October 28 - 31, 2000; in the section on molecular detection & therapy.

http://www.cancerprev.org/Journal/Issues/24/101/312/3819