Sven Bergmann Department of Medical Genetics, UNIL & Swiss Institute of Bioinformatics WP8: Computational analysis of beta-cell modular organization EuroDia Meeting Lund, February 2009
Iterative Signature Algorithm Unsupervised large- scale data analysis tool Modularizes the expression matrix Reduction of complexity Allows for easy data integration Interactive webtool
Modular Analysis A block of the reordered expression matrix Genes and samples have scores Captures differential co- expression Transcription Module
Non-modular Analysis
Modular Analysis
Modular Analysis of Multi-tissue Gene Expression Data Gábor Csárdi and Sven Bergmann Computational Biology Group, Department of Medical Genetics, University of Lausanne, Switzerland
The Data Set Coming from WP2, Frans Schuit's group. C57bl6 mice, plus 7 S/A islet samples 23 different tissues: adrenal gland, bone marrow, brain, diaphragma, ES cells, eye, fat, fetal, gastrocnemius muscle, heart, islet, kidney, liver, lung, ovary, parotis gland, pituitary gland, placenta, seminal vesicles, small intestine, spleen, testis, thymus. 3-5 samples/tissue, 89 altogether 19 islet samples, 8 on high fat diet After filtering based on variance: 14,540 of 45,101 probesets left on the mouse4302 array
Batch and Tissue Effects Islets Pituary gland
Spearman Rank correlation between 75 Affy mouse arrays High Fat Diet (5) Standard Diet (4) Pancreatic acini (3) Adrenal (3) ES cells (3) Brain (3) Eye (3) Adipose tissue (3) Heart (3) Hypothalamus (3) Small intestine (3) Kidney (3) Liver (3) Lung(3) Parotis gland (3) Spleen (3) Testis (3) Thymus (3) Diafragm (3) Pituitary gland (5) Bone marrow (4) Seminal vesicles (3) Skeletal muscle (4) Islets TISSUE (n arrays)
Batch and Tissue Effects
Very preliminary Modular Analysis Results 977 transcription modules were identified High enrichment by tissues, GO categories, KEGG pathways and transcription factors Condition plots Show tissue specific modules
Pancreas-Specific Modules (Islets + contaminating exocrine cells) Example: #49, 35 probes, 27 Entrez genes, 43 conditions, 19 islet samples with positive scores Many pancreas related genes
Pancreas-Specific Modules Genes: Gcg, Iapp, Abcc8, Scn9a, Prss2, Pnlip, Ela3, Rab37, Cuzd1, Pnliprp2, Clps, Rnase1, Asb6, Ctrb1, BC039632, B830017H08Rik, A930021C24Rik, C04Rik, H19Rik,
Pancreas-specific Modules Module #49 Differentiates between islets and other tissues
Islet specific GO enrichment P-value#Category 5.03e-749digestion 1.25e-1210extracellular region 3.18e-520mitochondrion 1.44e-425proton-transporting ATP synthase complex, catalytic core F(1) 1.04e-664serine-type endopeptidase activity 1.49e-4151endopeptidase activity 3.33e-416structural constituent of ribosome
Islet specific KEGG enrichment P-value#Category 4.37e-616Ribosome 3.26e-4155Oxidative phosphorylation
Islet specific miRNAs P-value#Category 6.65e-3892miR-30 family
High Fat Diet Islets Running ISA on the 19 islet samples only Only 8,288 probesets after filtering Module #41 differentiates between HF/LF diets best:
High Fat Diet Islets Condition scores significantly differ, p-value: 8.5*10 -3 Significantly enriched for serine-type endopeptidase activity, p-value: 3* Enriched for regulation by Trypsin GTF, p-value: 3*10 -14
High Fat Diet Islets Module #41 58 probes, 45 Entrez genes Two outliers
High Fat Diet Islets Serine-type endopeptidase activity, GO molecular function category
High Fat Diet Islets Intersection of module #41 and “Serine- type endopeptidase activity”, GO molecular function category
Acknowledgements People: Zoltán Kutalik Micha Hersch Aitana Morton Diana Marek Barbara Piasecka Bastian Peter Karen Kapur Alain Sewer Toby Johnson Armand Valsessia Gabor Csardi Sascha Dalessi Thanks to EuroDia!