Algorithm 1: Method of finding the top-ranked node in discriminating 3-level ASD subtypes and NC using mean value of PSD.

1.Input data = matrix ($m\times n$) m = 36 (number of subjects), n = 4 (number of group) 2.BOLD signals = $t\times R_j$ t = time points, $R_j$ = node number (j = 1, 2, …, 116) 3.PSD = estimate PSD for each BOLD signal 4.$PS{D}_n$ = normalize and determine the average of the PSD values 5.$G_j$ = cluster the outcome of (4) into matrix ($m\times n$) 6.p-value = run ANOVA test for each cluster ($G_j$) While j ≤ 116 repeat step 6 If p-value ≤ 0.05 save p-value at T else T = empty 7.F = Find top-ranked node based on T-index