Seminar 1

6July, 15:00, Seminar Room, 2nd Floor, Building VII

Alexandra Posekany, Research Unit of Computational Statistics, Vienna University of Technology

Bayesian and robust insights in data analysis and classification of health data

Outliers and systematically skewed or heavy-tailed data frequently occur in data analytical problems of many fields ranging from economics to bioinformatics. A specific notion of Bayesian robustness is robustifying the likelihood, as the backbone of the model. Constructing normally distributed likelihood models is often due to computational convenience, in the same way as classical inference with approximate normality is. Independent of sample size, data in many applied fields nowadays do not fulfill this assumption and linear and non-linear models for regression or classification suffer from that. We wish to provide robust estimation of parameters of the "main part of the data" through a normal or skewed distribution as likelihood, while simultaneously identifying the "outlying part of the data" represented by one or more skewed or heavily-tailed mixture components. Through the component labels and posterior weights we can identify the noisy or outlying parts of the data for filtering or inspecting the data quality.

Seminar 2

13 July, 15:00, Room 1.16, 1st Floor, Building VII

Sonia Torazona, Applied Statistics and Operations Research and Quality, Universitat Politècnica de València

Approaching disease through omics data: challenges and opportunities

Combining different omics data modalities such as genomics, transcriptomics, proteomics, or metabolomics measured on the same biological system has become a powerful approach for understanding the mechanisms underlying disease or obtaining diagnosis and prognosis biomarkers. Multi-omics studies are gaining popularity due to the holistic view of the biological system they provide.
The integrative analysis of such multi-omics datasets is not straightforward, though, and many methodologies are being developed following different strategies: data-driven versus biology-driven, supervised versus unsupervised, based on machine learning, etc. However, some bottlenecks in this type of analysis must still be faced, from the appropriate harmonization of the different omic modalities integrated into the same model to the complex task of interpreting the vast amount of omic feature relations derived from the integrative models, among others.
The bioinformatics tools developed by Dr. Tarazona's group aim to assist researchers in overcoming the difficulties of multi-omics analysis and gaining knowledge about disease onset, progress, or treatment. They cover different aspects and goals of multi-omics studies and are freely available to the scientific community.

Seminar 3

13 July, 16:00, Room 1.16, 1st Floor, Building VII

Antonio Gómez Corral, Complutense University of Madrid

On the exact measure of the disease spread in SIS epidemic models with horizontal and vertical transmission

Our objective is to propose a bi-variate competition process to describe the spread of epidemics of SIS type through both horizontal and vertical transmission. The main interest is in the exact reproduction number, which is seen to be the stochastic version of the well-known basic reproduction number. We characterize the probability law of the exact reproduction number by decomposing this number into two random contributions allowing us to distinguish between infectious person-to-person contacts and infections of newborns with infective parents. The methodology is mainly based on structured Markov chains and related matrix-analytic methods.