Algorithm 1.

Semi-supervised hierarchical multimodal feature and sample selection

Input:
	Labeled and unlabeled data from MRI and SNP, and the number of hierarchies L.
1:	Initialize labeled sample weights in A and feature coefficients in w.
2:	for i = 1 to L do
3:	Calculate the data similarity scores in S by Eq. (4).
4:	Calculate the sample weights in  by Eq. (5).
5:	repeat
6:	Fix A and solve w in Eq. (6).
7:	Fix w and solve A in Eq. (6).
8:	until convergence
9:	Discard insignificant samples and features based on the values in A and w.
10:	Weight the remaining features by the coefficients in w.
11:	end for
Output:
	Subset of samples and features for classification model training.