A confirmation analysis method for identification of chromosomal fragile sites.

Fragile sites are chromosome bands that do not manifest a presumed breakage pattern. Identification of fragile sites is a way to investigate the mechanism of carcinogenesis because the fragility at a specific chromosome position may be the causation of an associated cancer. A problem in the identification of fragile sites is the high false positive rate arising from simultaneously carrying out a large number of significance tests. To control it, we propose to find a reference study to confirm the identification result of an objective study. We utilize the Bayesian concept for linking two studies. Basically, our method demonstrates a conservative way to take account of the prior information of a binomial parameter. The derived estimate of breakage probability can be interpreted as a resampling weighted sample-pooling method. It is applied to confirm the identification of fragile sites for a data set of neuroblastoma patients.