CancerNetsTier3

Compbio results

Main thesis

Signed nested effects models find better network models than nested effects models and find novel pathway members better than other pattern detection methods.

Nested effect models have the advantage over Bayesian networks due to cyclical modeling, and advantages over DBNs due to lower requirements on data, in number of samples, time requirements.

Primary Material

ZIP Archive of all Figures: figs.zip

ZIP Archive of all Figures in TIFF format: figs-tiff.zip

Figure 1. Scoring pair-wise interaction modes with quantitative data.

Figure 2. Structure of factor graph for model inference

Figure 3. Results on artificial network data

Figure 4. Yeast knockout compendium predictions

Figure 5. Invasive colon cancer network predictions

Table 1. Top frontier genes for colon cancer invasiveness

Supplemental Material

ZIP Archive of all supplemental material: supp.zip

Text S1 Additional methods used in generation of artificial networks and data.

Figure S1. Observed inhibitory effects and signaling in yeast compendiums Evidence for inhibition from measured responses of knockdown, and from annotation in curated pathways.

Figure S2. Comparison of uFG-NEM and exhaustive NEM model search for structure recovery. Each bar shows the mean and standard deviation of precision or recall for S-gene links over 100 artificial networks with five S-genes. Networks were generated with no inhibition and an average of 20 E-genes per S-gene. Expression was generated for one replicate using standard deviation of 1 and a mean for the down distribution of -1.75. Precision and recall for both uFG-NEM and exhaustive NEM search do not differ significantly.

Figure S3. Network recovery as a function of increasing inhibition in E-gene attachment. Accuracy of FG-NEM and uFG-NEM network recovery on artificial networks with inhibition only in E-gene connections.

Figure S4. Estimating difference between activation and inhibition distributions using proteasome- and ribosome-related genes as contrast sets. Estimated difference in expression from three species and over 5000 microarrays.

Table S1. Yeast Knockout Compendium Pathway AUC AUC and AUC ratios for expansion of Yeast pathways.

Table S2. Ion Homeostasis Frontier Genes. Genes most likely to be attached to the ion homeostasis network for both the FG-NEM and uFG-NEM methods. Genes are sorted by LAR.

Table S3. Invasiveness E-gene LAR Scores. Connection point, connection strength, and connection significance of E-genes in colon cancer network.

Dataset S1. Selected E-genes and SAM selection. Sheet 1. Selected E-genes and their expression. Sheet 2. Input to SAM for determining parameters of the Gaussian mixture in the Expression Factors. Sheet 3. SAM results used for determining the parameters of the Gaussing mikter in the Expression Factors.

Outdated Material

KnockoutNetsPreviousResults