Create a test to see if chromosome 18 is duplicated or removed more often than other chromosomes.
-
Change resampling method to bootstrap
- take N samples with replacement, where N is the the number of cells.
- 1000 bootstraps
-
Compare bootstrap distribution between each pair of chromosomes
- Wilcoxon or t-test to assess for similarity in variability between pairs of chromosomes
- Possibly KS test as well
- Generate heatmap with p-values (23x23 matrix)
-
Linear regression - identify features that predict chromosome variability
- Do this by chromosome and by bin
- Example features include # of genes, size of chromosome/segment, # of TFs in segment, gene density, open chromatin regions
- Y is the observed frequency, Xs are the observed features
-
Visualization of p-values - use all pairwise comparisons. Plot the number of significantly different chromosomes on the x axis.
-
Network - assess similarity between pairwise comparisons.
-
Use association rules to determine multiple regions that are co-deleted or co-amplified
- See "mining of massive datasets," Chapter 6.1
Brainstorming on 3/18/2018:
