Time-Course Network Enrichment TiCoNE Time-Course Network Enrichment November 2016 by Jan Baumbach
vs. Jan Baumbach Computational Biology group University of Southern Denmark Odense, DK
Time series network enrichment
Time series gene expression: A network perspective Application: Time series data enrichment Network: PPI + GRN Cell type: Lung cell samples Study object: After infection with Influenza or Rhino virus over ten time points Read-out: Gene expression Human bronchial epithelial cells (BEAS-2B) Rhinovirus, Influenza virus or both, and RNAs profiled after 2, 4, 6, 8, 12, 24, 36, 48, 60 and 72hrs. Kim TK et al. A systems approach to understanding human rhinovirus and influenza virus infection. Virology 2015 Dec;486:146-57.
Time series gene expression: A network perspective Application: Time series data enrichment Network: PPI + GRN Cell type: Lung cell samples Study object: After infection with Influenza or Rhino virus over ten time points Read-out: Gene expression
Time series gene expression: A network perspective Application: Time series data enrichment Network: PPI + GRN Cell type: Lung cell samples Study object: After infection with Influenza or Rhino virus over ten time points Read-out: Gene expression Result: Temporally (in)active pathways during Influenza vs. Rhino virus infection
Time series gene expression: A network perspective Time-Series Data Clustering Network
Time series gene expression: A network perspective ... Initial Clustering Cluster prototypes
Time series gene expression: A network perspective ... Human Augmented Clustering Initial Clustering Cluster prototypes Remove non-fitting objects Split a cluster Merge clusters ... ... ... ...
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships.
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are four edges between the four blue and the four green cluster more/less than expected by chance?
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are four edges between the four blue and the four green cluster more/less than expected by chance? Calculate expected number of edges between any four nodes to any four nodes (with the same node degrees as the green/blue ones, using the joint node degree distribution). Calculate log-odds score S (#observed edges / #expected edges). Create 1,000 random networks (crossover 4*|E| edges). For each, compute log-odds score Si. Score distribution. Empirical p-value = rel. frequency of Si >= S
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are they similar between two conditions (e.g. Influenza vs. Rhino)? Influenza Rhino
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are they similar between two conditions (e.g. Influenza vs. Rhino)? Infl. Rhino # p 2 Influenza Rhino
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are they similar between two conditions (e.g. Influenza vs. Rhino)? p = empirical p-value for 1,000 networks with random color assignment Infl. Rhino # p 2 0.07 Influenza Rhino
Time series gene expression: A network perspective How many network interactions can we expect between two clusters? How many do we actually see? Discrepancy may indicate functional relationships. Here: Are they similar between two conditions (e.g. Influenza vs. Rhino)? p = empirical p-value for 1,000 networks with random color assignment Infl. Rhino # p 2 0.07 3 0.01 0.1 1 0.3 Influenza Rhino
Time series gene expression: A network perspective More/less edges between pairs of clusters of genes (of temporally similar expression) than expected by chance after Rhino virus infection. [more] [less]
Time series gene expression: A network perspective More/less edges between pairs of clusters of genes (of temporally similar expression) than expected by chance after Rhino virus infection. [more] [less] Graph representation (edge weights: log-scaled p-value).
Time series gene expression: A network perspective Subnetworks enriched with genes having network-associated temporal response patterns after Influenza infection but not after Rhino virus infection.
Time series gene expression: A network perspective Subnetworks enriched with genes having network-associated temporal response patterns after Influenza infection but not after Rhino virus infection. Known Influenza gene complex (literature).
Time series gene expression: A network perspective http://ticone.compbio.sdu.dk
Thanks! http://www.baumbachlab.net OR Jan Baumbach