Title: "A Genomic Random Interval Model for Functional Analysis of Genomic Lesion Data"
Speaker: Stanley Pounds; Department of Biostatistics, St. Jude Children's Research Hospital; Memphis, TN
Place: PHYS 223
Date: April 17, 2012; Tuesday,
Time: 4:30pm


New biotechnologies enable oncology researchers to identify, describe, and catalog genomic lesions of tumors. Popular methods for functional analysis of genomic lesion data determine whether the frequency at which lesions impact a set of genomic markers is significant against a model of chance. These methods can give misleading results because they use models of chance (such as permutation or the hypergeometric distribution) that do not preserve important biological characteristics of the data such as the number and continuity of the lesions and the fact that genes occupy loci along chromosomes. We propose a model of chance that represents each lesion as a genomic random interval (GRIN) that may occupy any interval locus of a given size along a specific chromosome with uniform probability. Unlike other methods, GRIN retains the important biological characteristics mentioned above and can model the full spectrum of genomic abnormalities including fusions, point mutations, and copy number alterations. Also, the GRIN model improves statistical power, diminishes multiplicity, and reduces computational effort by decreasing the number of modeled random events from the product of the number of genomic markers and the number of tumors to the number of observed genomic lesions. These advantages are observed in simulations and in practice. After Bonferroni adjustment in the analysis of early T-cell precursor leukemia whole-genome sequencing data, other methods had no significant findings while GRIN determined that the lesions' loci significantly overlap the ETV6, RUNX1, genes regulating T-cell development and the KEGG AML pathway.

Associated Reading:
Zhang et al. 2012. The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature. 481:157-163.

Click here www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL11/sem.html for a full schedule of BIOINFORMATICS SEMINARS, past and present.