Transcript Slide 1
Gene network approach in epidemiology Andrey Alexeyenko M E B edical pidemiology and iostatistics Network is just a graph! The fact that we can draw a network does not yet make it a biological reality!.. Why the network approach is an advancement compared to differential expression analysis? • • • • • • Functional context “Anchoring”, i.e. interdependence Biological interpretability Accounts for more statistical properties Data integration More data = flexibility! Gene network discovery: high-throughput experiments Gene network discovery: probabilistic data integration Bayesian inference: rps14 and rps8 coexpressed P(C|E) = (P(C) * P(E|C)) / P(E) rps14<->rps8 rps14 and rps8 functionally coupled FunCoup is a data integration framework to discover ? BHuman Mouse Fly Worm Yeast High-throughput evidence AHuman Find orthologs* functional coupling Conversion “data pieces confidence” in a Bayesian framework Data components in FunCoup D. rerio, 17.3% D. melanogaster, 9.8% C. elegans, 9.3% R. norvegicus, 5.1% S. cerevisiae, 10.2% A M. musculus, 25.4% A. thaliana, 6.5% H. sapiens, 16.5% P hylogenetic profiling, 18.6% P rotein interactions, 10.6% P rotein expression, 6.1% TF targeting, 12.3% miRNA targeting, 2.0% S ub-cellular localization, 7.3% mRNA expression, 43.1% http://FunCoup.sbc.su.se FunCoup: on-line interactome resource Andrey Alexeyenko and Erik L.L. Sonnhammer (2009) Global networks of functional coupling in eukaryotes from comprehensive data integration. Genome Research. Gene network discovery: getting rid of spurious links 0.7 0.5 0.4 Data processing inequality: “Direct links convey more information than indirect ones” Mutations: distinguishing drivers from passengers Functional coupling transcription transcription transcription methylation methylation methylation mutation methylation mutation transcription mutation mutation + mutated gene Network curation: cancer viewed by KEGG database curators Prostate cancer: recapitulated by FunCoup Network reconstruction: combination of methods Combination of methods: edges with different features A Alexeyenko, DM Wassenberg, EK Lobenhofer, J Yen, ELL Sonnhammer, E Linney, JN Meyer (2010) Transcriptional response to dioxin in the interactome of developing zebrafish. PLoS One. Verification of single gene lists •Yellow diamonds: somatic mutations in prostate cancer •Pink crosses: also mutated in glioblastome (TCGA) Subtyping cancer. Personalized medicine. Power of clinical trials AZD 2281 Salt Lake City, UT, June 19, 2007— Myriad Genetics, Inc. today announced the start of two Phase II trials for a new compound being tested to treat patients with BRCA1 & BRCA2 positive breast and ovarian cancer. Biomarker signatures in the network × Severity, Optimal treatment, Prognosis etc. Single molecular markers are often far from perfect. Combinations (signatures) should perform better. How to select optimal combinations? Candidate signature in the network Biomarker candidates Ready network-based signature RELAPSE = γ1EIF3S9+ γ2CRHR1 + γ3LYN + … + γNKCNA5 Identical genotypes can go different ontogenetic ways Gene a Protein A Disease Development Birth Adult Current gene expression results from inherited genotype, ontogenesis, and disease etiology Gene Gene a Protein A Gene Disease Gene Physiological condition Gene Pathway cross-talk Analysis of cancer-specific wiring Pathway network of normal vs. tumor tissues Edges connect pathways given a higher (N>9; p0<0.01; pFDR<0.20) number of gene-gene links (pfc>0.5) between them (seen as edge labels). Known pathways (circles) are classified as: •signaling, •metabolic, •cancer, •other disease. Blue lines: evidence from mRNA coexpression under normal conditions + ALL human & mouse data. Red lines: evidence from mRNA coexpression in expO tumor samples + ALL human data + mouse PPI. si Node ze: number of pathway members in the network. Edge opacity: p0. ss: number of gene- Edge thickne gene links. Arrow of time: network prospective A Alexeyenko, DM Wassenberg, EK Lobenhofer, J Yen, ELL Sonnhammer, E Linney, JN Meyer (2010) Transcriptional response to dioxin in the interactome of developing zebrafish. PLoS One. Thanks to: • • • • Erik Sonnhammer Martin Klammer Sanjit Roopra Joel Meyer Thank you for listening! http://FunCoup.sbc.su.se Summary: • Predicting gene networks is realistic. • Proposed applications: – Genetic heterogeneity of cancer – Communication between different cells, tissues, processes etc. – Evaluation of candidate biomarkers – Expression signatures Ask concrete practical questions, not global ones! Prostate cancer a cancer regulatory network + proteomics (HPA) data Cyan: up-regulated in normal glandular cells Red/green: up/down-regulated in malignant cells Yellow&magenta: (potential) regulators of prostate cancer