Skip to Main Content
The current advances in array-based techniques allow the measurement of copy number at very large number of locations in the genome. The copy number data from a single sample is segmented to identify gains and losses which are frequently found in cancer. The availability of large sample-size datasets encourages researchers to identify recurrent aberrations. recurrent aberrations happen within the same chromosomal region across multiple cancer samples. In this paper we propose a new algorithm based on non-negative matrix factorization (NMF) and circular binary segmentation (CBS). The proposed algorithm uses sparse NMF and CBS to detect the recurrent regions candidates. Then we adopt cyclic shift which is used to permute the data to distinguish between recurrent and sporadic copy number aberrations. We applied the proposed algorithm to two real datasets of glioblastoma and ovarian cancer. The results show the ability of the proposed algorithm to identify recurrent regions and to provide useful information to study genesis of cancer.
Date of Conference: 2-4 Dec. 2012