This week in techniques

Approach

Summary

Licensing status

Publication and contact information

Computational models

An algorithm to predict cancer subtypes by simplifying complex gene classifiers

A computational model could simplify gene classifiers used to predict molecular subtypes in cancer. Patients with rhabdomyosarcoma are classified as having low, intermediate or high risk of death, which informs the type of chemotherapy they receive. In three separate sets of published gene expression data from patients with rhabdomyosarcoma, the algorithm could distinguish tumors that express the paired box (PAX)-RNA binding protein fox-1 homolog 1 (FOX1) fusion gene from those that do not in >97% of cases based on pairs of markers. The algorithm also distinguished between patients with lung cancer suffering early death and those experiencing long-term survival by using two-gene combinations with an efficiency of >85%. Next steps include generalizing the algorithm, making it available on a public platform and prospectively testing the classifier in clinical trials.

SciBX 6(35); doi:10.1038/scibx.2013.972
Published online Sept. 12, 2013

Patent and licensing status not applicable; tool will be available at http://exhaustive.msvalidator.org

Wilson, R.A. et al. Cancer Res.; published online Aug. 2, 2013;
doi:10.1158/0008-5472.CAN-13-0324
Contact: Samuel L. Volchenboum, The University of Chicago, Chicago, Ill.
e-mail:
slv@uchicago.edu