You are here: Home Projects Current Projects Increasing clinical usefulness …

Increasing clinical usefulness of gene signature prediction rules through simplification and validation

Duration: since 2011

Principle Investigators:
Professor Dr. Anne-Laure Boulesteix (LMU)
Professor Dr. Wilhelm Sauerbrei (IMBI)

Researchers:
Sam Dörken (IMBI)
Anika Buchholz (IMBI)

Description

Hundreds of gene signatures based on high-dimensional data (HD-D) have been proposed in the biomedical literature and hundreds of methodological papers are published on issues around deriving, validating and applying prediction methods for such data. Nevertheless, the clinical use of information from HD-D is extremely limited so far. Surprisingly, several key issues known from classical statistics and applied to low dimensional data (LD-D) are hardly considered when analyzing HD-D.

We will investigate whether some useful statistical approaches can be transferred and adapted from LD-D to HD-D. We will consider the added value of HD-D and its validation, procedures to simplify predictors and derive ‘extremely sparse’ models, the possibility to improve predictors by allowing non linear effects for continuous variables and time-varying effects of variables. This may provide insight into the role of some high-dimensional features and simpler models will help to improve the translation of results from research into practice. Biostatistical approaches will be investigated and compared in simulation studies. Available data sets will be re-analyzed with the most promising approaches. Emphasize will be on gene signatures for predicting survival of patients with breast cancer.

Systematic reviews and the results of meta-analyses play an important role to increase clinical usefulness. In the second funding period we will work on issues how to combine the results from studies and we will use the breast cancer data sets to conduct meta-analyses for HD-D.