Integrative genomic and survival analysis of breast tumors

Abstract

With the continued accumulation of genomic data at ever increasing resolution the challenge ahead lies in reading out meaningful clinical/biological information form the data that can contribute to a better understanding of the cancerous process. The need for novel approaches, new statistical methods is therefore strong. The present thesis aims to contribute to the field with three problem specific applications that that expectantly will aid researchers in a better understanding of genomic data. The first paper exemplifies the adaptation of a piecewise-linear regression framework to integrative analysis of DNA copy number aberrations and gene expression (mRNA) data. The methods allows it helps identify the association between copy number and gene expression but it takes a further step and allows detection of changing patterns and changepoints could serve as a proxy for the degree of genomic instability that causes disruptions in feedback-mechanisms. The second paper advocates the adaptation of a mediation analysis for a concomitant analysis of DNA copy number aberrations, mRNA and survival data. The paper offers ways of statistical inference by the means of Delta-method applicable concomitantly on a large number of genes. If a mediation effect is observed for a specific gene, we hypothesize that the specific gene is a driver gene. If no mediation effect is observed then possible associations between DNA copy number aberrations and the outcome are likely to indicate passenger genes. The third paper is a more applied/clinical work and using applied statistics identified a novel panel of 12-genes that can serve prognostic toll for breast cancer specific survival. The thesis concludes with a methodological description in which we describe an easy permutation based approach for testing the clonal origins of multiple tumors. The main assumption of the proposed method is that if two tumors that share a common origin or if the alleged secondary tumor is clonally related to the primary tumor then they share a higher and tumor specific amount of matching chromosomal aberrations (gains or deletions) than recurrent chromosomal aberrations can explain.