Details about feature sets used:

We would put all related information about each features here if needed.

· The ninth feature group we used includes the first three singular values (SV) of the candidate subgraph

When comparing subgraphs with the same size, different shapes have different value distributions with these three SV.

For example, for four nodes subgraphs è

Shape	Linear	Clique	Star	Hybrid
First Three SVs	% 1.6180 % 1.6180 % 0.6180	% 3.0000 % 1.0000 % 1.0000	% 1.7321 % 1.7321 % 0.0000	% 2.1701 % 1.4812 % 1.0000
Graph (binary)

· The eighth feature group we used is derived on topological coefficient.

The topological coefficient (TC) is a relative measure of the extent to which a protein shares interaction partners with other proteins. It reflects the number of rectangles that pass through a node.

For node p, we assume it has q partners, one of them is node s and there are N(p,s) number of nodes to which both p and s are linked (plus 1 if there is a direct link between p and s).

Thus TC=average_q(N(p,s)/q). We use mean, variance and maximum of the property and the expected value for different shapes also varies depending on the ratio of rectangles within.