Responsible: Dr. Rainer Spang, MPG, Berlin.

Background:

Expertise on how to analyse expression profiling data needs to be established in source laboratories and this implies dedicated training of scientists in the associated theory and techniques. While this process was successfully initiated in NGFN1 much remains to be done. The purpose of this subproject is therefore to provide lab-based scientists with the practical know-how and theoretical knowledge in the analysis of microarray data.

Planned work:

We plan to offer four courses a year; two taking place in Heidelberg and two in Berlin, with each course lasting four days. In theoretical sessions in the morning, the most important aspects of microarray analysis are taught on an introductory level. Only very basic knowledge in mathematics, statistics and programming is required. The workshop participants will get a first understanding of the most common problems in microarray analysis, and will be taught basic statistical and bioinformatics approaches to tackle them. We will cover data quality control issues (Huber), data normalization for various array platforms (Huber), differential gene expression and multiple testing (Spang), exploratory data analysis including clustering and imbedding methods (Lengauer), tumor classification and computational diagnostics by statistical learning models (Spang), critical evaluation of classifier performance via bootstrap and cross validation (Mansmann) and basic questions of experimental design (Mansmann). We have included specialists in all relevant fields into this training unit. During the course of the NGFN2 these topics will change as new mathematical methods emerge or other profiling methods become more prominent in the research of the clinical networks. Depending on the scientific developments, future topics could include the analysis of proteomics, matrix CGH and methylation profiling data, as well as data analysis strategies for integrated data sets from several of these platforms. The number of participants is limited to 16-18, ensuring an excellent teacher/student ratio.