Computational Diagnosis

Nested Effects Models Rainer Spang 22.11.2007 Annual Conference of the Working Group Statistical Methods in Bioinformatics Signal transduction in the cell S External Signal Non-transcriptional signaling Difficult to observe Transcription / Gene Expression Apoptosis Easy to observe with Microarrays Signaling and External Interventions S Blocking signaling flow by RNAi Introducing constitutive signals by transfecting activated genes into the cells Apoptosis Traces of a Bifurcation S Module 1 Bifurcation Module 2 Module 1 Module 2 From: Boutros M, Agaisse H, Perrimon N. Sequential activation of signaling pathways during innate immune responses in drosophila. Developmental Cell, 3(5):711–722, 2002 E-Genes S-Genes Nested effects models (NEMs) Negative Controls (C-) Positive Controls (C+) Interventions in S-Genes (RNAi) Observations in E-Genes (Microarray) Silencing Effect: An E-gene goes from a C+ level back to a C- level Data: Binary matrix D = (eik), where eik = 1 if E-gene Ei shows in experiment k the same expression as in the negative controls. The core model and the extended model E E E We only want to infer the core topology E E E E E E E E The position of the yellow edges is unkown too, but we treat it as a nuisance parameter core model extended model How does data generated by the model look like? Model Assumptions: - The core model is transitive - Every E-gene is connected to exactly one S- gene - Independent binary noise Scoring observed silencing effects Marginalization Markowetz F., Bloch J., Spang R., Non-transcriptional pathway features reconstructed from secondary effects of RNA interference. Bioinformatics 21, 4026-4032, 2005 The model search space We have a likelihood for each core model  find the maximum likelihood model Pairs of genes Fit a model for every pair of genes Pick the best of the four possible models Problem: transitivity lost Triples of genes Fit one model for each triple of genes Pick the best from the 29 possible models … Edgewise model averaging: Edges that are frequently used in the triple models are included in the final model Reconstruction accuracy in simulations Exhaustive Triples Pairs The Boutros et al Data Raw Data Estimated Effects ML-Model The Likelihood Landscape Parameter Dependence 100% accuracy used parameters Nested Effects and Clustering BCR-signaling in immature B-cells Tze et al PLoS Biology 2005 Posterior E-gene position Posterior E-gene position NF-KB Posterior E-gene position Burkitt Lymphoma Marker Acknowledgements Claudio Lottaz Juby Jacob Stefan Bentink Inka Appelt Christian Kohler Mohammad Sadeh Mathias Maneck Maria Bartolim ------------------------------------------------------Florian Markowetz NEMs Jörn Tödling Xinan Yang Jochen Jäger Stefanie Scheid Dennis Kostka BCR-Signaling, COMAPs Benedikt Anchang Thank You