Table 3.

Labels of 10 most stable features (sorted by the distance coefficient k). Features unique to each group are highlighted. k denotes the number of times out of the 50 runs a given feature was in a cluster different from the cluster it was assigned to in the best solution obtained using the complete dataset.

#	k	Controls			k	Patients
1	0	Central	SIRP	ICA	0	DMN2	SM	ICA
2	0	Central	SM	ICA	0	FPOLE	AOD	ICA
3	1	Targets	AOD	SPM	1	V1	SIRP	ICA
4	1	Central	AOD	ICA	3	Central	AOD	ICA
5	1	DMN2	AOD	ICA	3	RDLPFC	AOD	ICA
6	10	V1	SIRP	ICA	4	RDLPFC	SM	ICA
7	13	V1	SM	ICA	4	Encode	SIRP	SPM
8	13	V1	AOD	ICA	4	Targets	AOD	SPM
9	14	Novels	AOD	SPM	4	RDLPFC	SIRP	ICA
10	14	Motor	SM	SPM	4	DMN2	AOD	ICA