Sorting Algorithm

library(ggoncoplot)

How are oncoplot axes sorted?

The most common approach to sorting an oncoplot is a hierarchical sort, which follows these steps:

Step 1: Identify Sort Order for Y Axis

  1. Identify Recurrence:
    • Determine the genes mutated in the greatest proportion of samples (multi-hits in a gene don’t count towards this quantification of recurrence).
  2. Select Top Genes:
    • Identify the top N genes (default is 10) based on recurrent mutations across samples.
  3. Rank Genes:
    • Rank these genes based on the degree of recurrence. This ranking determines the sort order for the Y axis.

Step 2: Identify Sort Order for X Axis

The X axis is sorted based on the Y axis.

Goal:

Rank each sample such that samples with mutations in the most recurrently mutated gene appear furthest left on the oncoplot. This is followed by samples with mutations in the second most recurrently mutated gene, and so on. Within each group of samples with mutations in a specific gene, we want to sort based on whether the next most mutated gene is mutated, and continue this process hierarchically.

Method:

The X axis hierarchical sort is achieved using a base-2 scoring system. If we display the top 5 recurrently mutated genes on the Y axis, we allocate a weighting for each gene based on its ranking in order of recurrence:

  • The most recurrent gene would have a weight of \(2^{(5-1)} = 16\).
  • The second most recurrent gene would have a weight of \(2^{(4-1)} = 8\).
  • And so on.

Each sample’s X axis sorting rank is calculated by summing the weights of all genes mutated in that sample.