Seminarium Wydziałowe: Olga Vitek
W imieniu Dziekana Wydziału serdecznie zapraszamy pracowników oraz studentów na seminarium wydziałowe, które odbędzie się we wtorek 23 czerwca o godz. 12:30 (Instytut Informatyki, sala 119). Tradycyjnie, przed seminarium, o godzinie 12 Dziekan zaprasza na kawę i ciasto (przed salą 119).
Prelegentem będzie Olga Vitek, będącej Raymond Bradford Bradstreet Professor w Khoury College of Computer Sciences oraz dyrektorką Barnett Institute for Chemical and Biological Analysis na Northeastern University.
Olga Vitek wygłosi wykład: Statistical methods increase the accuracy an d the interpretability of quantitative proteomic investigations.
Abstrakt: Proteins are the main functional molecules in cells and play a central role in understanding disease mechanisms, identifying drug targets, and evaluating the effects of therapeutic interventions. Proteomic experiments aim to measure, on a large scale, which proteins are present in a biological sample and how their abundance changes across conditions, such as healthy versus diseased tissues, treated versus untreated cells, or different stages of a drug-discovery experiment. As a result, modern proteomic technologies generate large, noisy, and highly structured data sets that require careful quantitative analysis.
Although statistical methodology in proteomics is often associated mainly with hypothesis testing, its role in understanding proteomic data is much broader. In this lecture, I will discuss how statistical ideas contribute to proteomic data analysis in three complementary ways.
First, statistical models provide principled methods for aggregating many low-level measurements, such as peptide-ion or fragment intensities, into accurate summaries at the protein level. This may be viewed as an estimation problem in which repeated, heterogeneous, and partially missing measurements must be combined while accounting for measurement error and systematic variation.
Second, flexible statistical models can improve the detection of changes in protein abundance in complex experimental settings. This is particularly important when the experimental design goes beyond standard two-group comparisons, for example in screening experiments whose goal is to identify proteins affected by a chemical or biological perturbation.
Third, statistical methods help connect observed quantitative changes with broader scientific knowledge. In this way, they support not only prediction or detection, but also interpretation: they help identify patterns in the data, relate them to known biological mechanisms, and lead to more meaningful scientific conclusions.
The lecture will illustrate these ideas using recent methodology implemented in the MSstats software framework, with examples from quantitative proteomic experiments, including chemoproteomic screens. The emphasis will be on the statistical and computational principles underlying the methods, rather than on proteomic details.
