1. Systems Biology through Pathway Statistics Chris Evelo BiGCaT Bioinformatics Group – BMT-TU/e & UM Diepenbeek; May 14 2004

2. BiGCaT Bioinformatics Where the cat hunts

3. BiGCaT Bioinformatics, bridge between two universities Universiteit Maastricht Patients, Experiments, Arrays and Loads of Data TU/e Ideas & Experience in Data Handling BiGCaT LUC Diepenbeek Statistical Foundations

4. BiGCaT Bioinformatics, between two research fields Cardiovascular Research Nutritional & Environmental Research BiGCaT

5. Our usual prey: gene expression arrays Microarrays: relative fluorescense signals. Identification. Macroarrays: absolute radioactive signal. Validation.

6. Transcriptomics: The study of genome wide gene expression on the transcriptional level Where genome wide means: >20K genes. And transcriptional level means that somehow >20K mRNA sequences have to be analyzed And >20K expression values have to be filtered, normalized, replicate treated, clustered and understood Thus no transcriptomics without bioinformatics

7. No separate statistics?: Previous slide: “…have to be: filtered, normalized, replicate treated, clustered and understood” Don’t we have to know which genes really changed?

8. Changed? We need statistical prove of genes changing because… Scientist ask for it. Journals ask for it. But do we really need it?

9. No we don’t! Biologist will double check anyway Largest problem are false positives 1 in 1000 means 20 on an array! Replicate filtering gets rid of that, loosing very little power off course that needed statistical proof To understand we need pathways not single genes (or proteins)

10. Two types of arrays Single longer (>60 mer) cDNA reporters Agilent, Incyte, custom 1 value per reporter Reference variability or multi array stats Multi short (25 mer) oligo reporters Affymetrix 16-20 values per reporter Single array statistics

11. Systems Biology Triangle Systems Biology Transcriptomics Metabolomics Proteomics microarrays, 20 k (available) Large scale analytical chemistry (developing outside) 2D-gels, antibody techniques (developing inside)

12. Proteomics would be: The study of genome wide gene expression on the translational level Where genome wide would mean: >20K proteins. Then proteomics does not yet exist!

13. Protein variants derived from single genes Phosphorylation? Modification? Alternative splicing?Phosphorylation? Alternative splicing? Modification?

14. Two types of omics Transcriptomics Microarrays Values for 20 K genes Annotation difficult Proteomics Currently only 2D+MS Only 20-50 identified proteins Annotation is identification Plus modifications

15. Gene Ontology (GO) levels (I) Amigo browser http://www.godatabase.org/cgi-bin/go.cgihttp://www.godatabase.org/cgi-bin/go.cgi GO consortium: http://www.geneontology.org The Gene Ontology (GO) project gives a consistent descriptions of gene products from different databases.

16. Gene Ontology (GO) levels (II)

17. Use of GO classification -GenMAPP- GenMAPP = Gene MicroArray Pathway Profiler Program to visualize Gene Expression Data on MAPPs representing biological pathways and grouping of genes * Local MAPPs contain pathways made by specific research institutes * Gene Ontology (GO) MAPPS contain pathways with functionally related genes from the public Gene Ontology Project

18. Example Local MAPP

19. Example GO MAPP

20. Local MAPP

21. GO MAPP

22. Understanding changes Map changed genes/proteins (quantitatively or qualitatively) to known pathways. Or use information from the Gene Ontology (GO) database Steal and smartly adapt a transcriptomics tool: GenMapp/MappfinderMappfinder Rachel will show some examples