Details for DMTree

Based on DMRs we can build a tree for clustering samples, termed DMTree.

image info

Start with all samples, the patterns of DMRs will be ranked by the sum of DMR weight \(\sum{W_i}\), where \(W_i\) is the absolute mean difference for DMR \(i\). Samples will be splitted into groups based on the pattern with highest \(\sum{W_i}\), until all possible pattern met the following termination criteria:

  • The number samples in the subgroup \(N_{sub}\) is smaller than \(N_{min}\) from -n, --minNSamples.
  • The sum of DMR weight \(\sum{W_i}\) is smaller than \(W_{min}\) from -w, --minSumDMRs.