Skip to content

Exposició GRASS

Ponents: Jose Luís Jiménez (UAM) // Carles Serrat - Dijous 5 de març 2015

Jose Luís Jiménez.

Multivariate Shared Kernel Bayesian Screening.


In gene expression studies, it is common to test for differential expression between groups for each gene independently without taking into account within-subject dependence or similarities across genes in the shape of the expression distribution. We are motivated by a related application in which the focus is on assessing differences between occupational groups in the levels of a marker in different brain regions. We would like to identify brain regions showing differences in the marker distribution between occupation groups, while taking into account within-subject dependence in the multiple measurements as well as similarities in the shape of the distribution across brain regions and groups. We propose a method based on characterizing the marker distribution for each subject in each brain region as a mixture, using a common set of kernels, with a subject-specific intercept and weights that are region and group-specific. The multiple hypothesis testing problem is accomplished in a Bayesian manner, formulating hypotheses of equivalence between groups for a particular brain region as corresponding to equivalence in weight vectors. We show that the proposed approach gains power over usual screening methods, while maintaining low false positive rates and adjusting for multiple comparisons.

Carles Serrat.

Joint model per al cas del càncer de pròstata.

The paper describes the use of frequentist and Bayesian shared-parameter joint models of longitudinal measurements of prostate-specific antigen (PSA) and the risk of prostate cancer (PCa). The motivating dataset corresponds to the screening arm of the Spanish branch of the European Randomized Screening for Prostate Cancer study. The results show that PSA is highly associated with the risk of being diagnosed with PCa and that there is an age-varying effect of PSA on PCa risk. Both the frequentist and Bayesian paradigms produced very close parameter estimates and subsequent 95% confidence and credibility intervals. Dynamic estimations of disease-free probabilities obtained using Bayesian inference highlight the potential of joint models to guide personalized risk-based screening strategies.