Create clusters based on pairwise columns (iteratively)

Question

I have a long pairwise table with two columns. I would like to convert these columns into clusters. For example, I would like:

input:
A B
A C
A D
A E
A F
B G
Y Z

output:
Cluster1 A,B,C,D,E,F,G
Cluster2 Y,Z

The part that is difficult for me is iterating through. For example, I can use datamash to create this (datamash -W -g1 collapse 2):

A B,C,D,E,F
B G
Y Z

But, I also want G to get grouped with A (through B) and to go through enough iterations to saturate. Any help is appreciated. Thanks!