Longitudinal measurements are receiving much attention in clinical research where patients are followed up to an event of interest. In these studies, the progression of disease can be closely monitored for a period of time and outcomes can be predicted more effectively by assessing temporal changes in responses. The analysis of follow-up studies with time-to-event observations is called longitudinal and survival data. For longitudinal gene expression data, it is observed that the measurements are often dependent on previous values, that is, they are auto‐correlated. The baseline expression value portrays a significant role and often the recent expression value is compared to the baseline value to draw the necessary inference.

Time-varying biomarkers have become an effective indicator for identifying the disease progression in Oncology. This tool has been proven advantageous to detect tumor development at early stages and also benefits in treatment decision making. In practice, the information of expression values is collected for a large number of biomarkers, producing a high-dimensional structure of the dataset. However, these datasets often contain missing observations due to patient drop-outs which undermines the validity of the research results. Our research aims towards several challenges that appeared in the biomarker studies and develops an effective statistical procedure for classification and modeling of gene expression data.

Organiser: Judit Szigeti

Chair: Dominik Rozkrut

Speaker: Ábel Czékus

Speaker: Peter Toth

Speaker: Peter Palosi

Speaker: Zsolt Kovari

Speaker: Marek Rojíček