NetBioSIG2013-Talk Thomas Kelder

2,036 views

Published on

Presentation for Network Biology SIG 2013 by Thomas Kelder, Bioinformatics Scientist at TNO in The Netherlands. “Functional Network Signatures Link Anti-diabetic Interventions with Disease Parameters”

Published in: Health & Medicine, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
2,036
On SlideShare
0
From Embeds
0
Number of Embeds
1,007
Actions
Shares
0
Downloads
22
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • This analysis is like finding the right pebbles on a huge pebble beach -> can’t take them all, but want to find a representative sample to take a part of your vacation home. Molecular network underlying disease is huge, we can’t focus on everything at once, but need to find the most relevant parts that tells us about specific aspects of disease.
  • Big network underlying diseaseDrug often targets single pathwayBased on what we think (single signaling cascade)Leads to both good and bad effectsIneffective in improving health systems-wideHow should network look like for effective treatment? -> marker nodesWhat should interventions target for optimal treatment -> target nodes
  • DLI as template for good intervention16 disease parameters. These include plasma glucose and insulin, QUICKI index, body and organ weights (adipose depots, kidney, liver, heart, and total body weight), atherosclerotic lesion area, plasma cholesterol, and plasma and liver triglycerides
  • Nodes in the network with key role in linking intervention target to dyslipidemia parameters. Circumvent responses like drug signatures, since all link to disease parameters that get worse.
  • (Fasn, Axl, Fgf21, Gpd2, Cyp17a1, Pkm, Fastkd5), may point to putative targets for improved interventions mimicking the mechanisms underlying DLI. Notably, the gene products of two of these genes are already under investigation as therapeutic targets. Fgf21, encoding for Fibroblast growth factor 21, is currently being investigated as novel therapeutic agent for T2DM [29, 30], and the anti-diabetic properties of thefatty acid synthase (Fasn) inhibitor platensimycin have recently been demonstrated in a mouse model[31]. Interestingly, Axl, encoding for the AXL receptor tyrosine kinase, was found to induce T2DM afteroverexpression in transgenic mice [32].
  • Sets of 25 genes of more, outperforms DEGs. DEG finds top of iceberg, but network based method seems better when going deeper.Also complete signature has significantly higher enrichment with known disease genes than same number of genes ranked by DEG.To next slide: Signatures are not just lists of genes, but networks that provide biological context. You can study the underlying interactions that cause the genes to have a high score, this facilitates interpretation and identifying biological mechanisms. Example in next slide.
  • Network visualization of underlying interactions:See both expression, relevance and interaction together -> biological contextTopologyRed module, inflammation and consistently opposite regulation DLI vs T09. Ccnd1 top score in both DLI and T09 networks, but different neighbors. Direct regulation by 5TFs (4 inflammation related), versus indirect links and more concentrated along single path for T09. Perhaps tighter, more balanced regulation required for good effect?This module also shows a clear opposite pattern of regulation between these interventions where the majority of genes were downregulated by DLI, while upregulated by T0901317 intervention. Several nodes receive a non-zero relevance score for both interventions (Ccnd1, Lgals3, Gja1) while the network visualization provides insight in difference in their regulation by the interventions. For example, Ccnd1 has a high relevance score in both signatures, but is downregulated by DLI and upregulated by T0901317. In the DLI network, Ccnd1 is directly regulated by 5 transcription factors affected by DLI, of which 4 could be related to inflammation or immune response pathways (Nr3c1, Nr4a1, Rxra, Smarcb1; based on annotations in Gene Ontology, Ingenuity Pathway Analysis, and WikiPathways). In contrast, Ccnd1 is connected to T0901317 through a single indirect association involving multiple intermediate interactions. This difference can be observed throughout the network, as the average shortest path length from intervention to the module nodes is twice as long in the T0901317 subnetwork compared to the DLI subnetwork. In addition, the edge relevance scores for the DLI network are more equally distributed across nodes, while the scores in the T090137 network are mainly concentrated in the path through Mmp9. This may indicate a more direct and balanced activation of repression of a combination of multiple transcription factors by DLI, while the indirect regulation by T0901317 intervention leads to a less controlled mechanism.
  • WGCNA mainly applied on genetically perturbed datasets (e.g. F2 crosses)We applied to datasets where variation is induced by intervention(s) -> 10OAD, WUR-DR- Generate network from data- Identify new relations- Link to physiology or other external measurements
  • Network signatures: Utilize known network information, different datasets (transcripomic like shown before, but also genetic for causal links).Identify parts of network that are linked / determine specific disease endpoints or phenotypes -> network signaturesMarkers: Could be used as markers to distinguish subgroups (i.e. develops NASH or not), prognostic for complications or diagnostic to determine which part of the system is diseased.Specific interventions: Networks provide biological context, mechanistic insights may lead to ways to design interventions that push that specific part of the network in the right direction to cure disease.
  • NetBioSIG2013-Talk Thomas Kelder

    1. 1. Network signatures link hepatic effects of anti-diabetic interventions with systemic disease parameters Thomas Kelder Microbiology and Systems Biology, TNO, The Netherlands Network Biology SIG, ISMB 2013, Berlin
    2. 2. 2
    3. 3. 3
    4. 4. Anti-Diabetic Treatment (ADT) study 4 DISEASE PARAMETERS • Plasma glucose, insulin • Body and organ weights • Atherosclerosis lesion area • Plasma cholesterol • Plasma & liver triglycerides + Dietary Lifestyle Intervention (DLI) Fenofibrate, T0901317 Improves all disease parameters Improves glycemia Deteriorates dyslipidemia Radonjic, et al., PLoS ONE, 2012 ?
    5. 5. Intervention – hepatic mechanisms – disease parameters 5 TRIGLYCERIDES ATHEROSCLEROSIS GLUCOSE INTERVENTION
    6. 6. Which paths? 6 TRIGLYCERIDES ATHEROSCLEROSIS GLUCOSE INTERVENTION
    7. 7. Network analysis workflow 7
    8. 8. Link to disease parameters 8
    9. 9. WGCNA • Weighted Gene Co-expression Analysis* • Identify co-expressed network modules • Correlate modules to disease parameters based on their “eigengene” (1st Principal Component) 9*Langfelder et al. BMC Bioinformatics, 2008 Disease parameter Disease parameter Disease parameter ? ? ?
    10. 10. Modules to disease parameters • 14 coherent co-expression modules • 10 modules with GO annotation • 4 modules correlated with disease parameter(s) • All correlating endpoints related to dyslipidemia rather than dysglycemia despite improvement of dysglycemia by all interventions 10
    11. 11. Link to intervention targets 11
    12. 12. Prior knowledge-based networks • Curated pathways • Protein-protein interactions • Transcription factor targets • Drug targets • DLI “targets” – Ingenuity Upstream Regulator Analysis – Enrichment of known TF targets with DEGs for DLI (p<0.001) – 25 transcription factors – Some overlap with drug targets (e.g. PPARA) Total network has >12,000 (gene) nodes and >75,000 edges
    13. 13. Intervention specific networks 13 Filter by differential expression for intervention vs HFD Network Total DEGs in dataset (p < 0.05) Connected nodes Edges DLI 1,287 497 5,975 Fenofibrate 2,149 828 21,598 T0901317 2,924 1245 38,472
    14. 14. Network signatures 14
    15. 15. Random walks algorithm 15 [1] Dupont, et al. "Relevant subgraph extraction from random walks in a graph." Machine Learning (2006) [2] Faust, et al. "Pathway discovery in metabolic networks by subgraph extraction." Bioinformatics (2010) Randomwalks Intervention Nodes and edges scored by probability of being visited by the random walker Intervention Disease parameter Disease parameter
    16. 16. Network signatures 16 DLI signatures Template for successful intervention Drug signatures Circumvent this response
    17. 17. Network signatures 17 • DLI vs drug, distinct response: • Small overlap • Opposite regulation • Potential drug targets: • key nodes unique for DLI • cross-talk between processes
    18. 18. Performance assessment Improved ability to prioritize genes by relevance to disease parameters 18 8.78 fold enrichment p = 3.44E-10 3.09 fold enrichment p = 0.023
    19. 19. Cholesterol Atherosclerosis Liver weight Cholesterol Atherosclerosis Liver weight 1TF Indirect links 5 TFs Direct links
    20. 20. Conclusions Network signatures underlying effects of interventions on dyslipidemia-related disease parameters – Template for successful intervention or response to circumvent – Improves selection of genes relevant to disease parameters – Underlying interaction help interpretation 20
    21. 21. Acknowledgements • Marijana Radonjic • Lars Verschuren • Alain van Gool • Ben van Ommen • Ivana Bobeldijk Check out our poster at ISMB on Sunday Network Biology of Systems Flexibility 21 R scripts and data for this analysis available at: https://github.com/thomaskelder/ADT-liver-network igraph
    22. 22. 22
    23. 23. 23 High fat diet “diseased” control group Chow diet “healthy” control group High fat diet DLI (switch to chow) Fenofibrate T0901317 wk 9wk 16wk LDLR-/- MICE HEPATIC TRANSCRIPTOME
    24. 24. 24 DLI Fenofibrate T0901317 Hepatic transcriptome dataset: - Chow control - Dietary lifestyle intervention (DLI) - Fenofibrate - T0901317 Compared to high fat diet (HFD) at 16 Co-expression network modules ide by Weighted Gene Co-expression Ne Analysis (WGCNA) [2]. Provides high overview of relevant processes. WGCNA
    25. 25. 25 DLI Fenofibrate T0901317 WGCNA DLI T0901317 FENOFIBRATE EXTEND W FILTER F
    26. 26. 26 en modules that could be annotated to a biological proce correlated significantly with disease parameters. All s ns were with dyslipidemia related disease parameters, de mprovement of glycemic status by the interventions. NO. GENES GO TERMS SIGNIFICANT CORRE 198 Lipid biosynthetic process, Oxidoreductase activity Liver weight (-0.91), Triglycerid Atherosclerosis (-0.79), Choles 161 Cell activation, Immune system process, Inflammatory response Atherosclerosis (0.80), Cholest Liver weight (0.75) 142 Lipid metabolic process, Oxidation-reduction process Liver weight (0.88); Cholestero
    27. 27. WGCNA • Weighted co-expression network analysis* • Correlate modules to other measurements (clinical, plasma proteins, microbiome) *Langfelder et al. BMC Bioinformatics, 2008
    28. 28. glucose C how H F 16 w eeks Lifesty le R osig lita zone T0901317 0 5 10 15 20 ** ** * glucose(mM) Omics, genetics, physiological data, prior knowledge Molecular signatures of metabolic health and disease Mechanistic insight: Biological context of molecular signatures Prognostic / diagnostics molecular signatures Coexpression networks (WGCNA) Prior-knowledge networks Causality networks Variable selection methods Subgraph ID/ (K-walks) topology/ network clustering Network signatures for improved diagnostics & interventions Link to pathological endpoint Subgroup-specific molecular signatures prioritization and refinement

    ×