Table 3.

Top 34 features (5%) showing the most discriminative biomarkers for multi-biological predictions.

Number	Feature name	Feature type	Number	Feature name	Feature type
1	SOD	Blood	18	Undefinedb	GM
2	MLR	Blood	19	Anaerostipes	GM
3	Lactobacillus	GM	20	PLT	Blood
4	MON	Blood	21	alpha2_aNLe_P4	EEG
5	Haemophilus	GM	22	Dialister	GM
6	Prevotella	GM	23	beta1_aLambda	EEG
7	NEU	Blood	24	Slackia	GM
8	CRP	Blood	25	Undefined	GM
9	Megamonas	GM	26	Odoribacter	GM
10	theta_aNLe_T6a	EEG	27	Ruminococcusc	GM
11	theta_aNe_T6	EEG	28	theta_aDc_FP1	EEG
12	theta__aNCp_T6	EEG	29	alpha2__aNCp_P4	EEG
13	WBC	Blood	30	beta2_aNLe_FP2	EEG
14	NLR	Blood	31	beta2_aDc_O2	EEG
15	Collinsella	GM	32	Gemmiger	GM
16	gamma_aDc_F7	EEG	33	alpha2_aNLe_T4	EEG
17	Clostridium	GM	34	alpha2__aNCp_T4	EEG

The top 34 features are listed in the descending order of their weights.

GM gut microbiota, EEG electroencephalogram, SOD superoxide dismutase, MLR monocyte–lymphocyte ratio, MON monocyte, NEU neutrophil, CRP C-reactive protein, WBC white blood cell, NLR neutrophil–lymphocyte ratio, PLT platelet, aNLe nodal local efficiency, aNe nodal efficiency, aNCp nodal clustering coefficient, aDc degree centrality.

^aThe EEG features are represented as a_b_c, where a represents the frequency band, b represents brain network attributes, and c represents the electrode channel.

^bUndefined Lachnospiraceae.

^cUndefined Ruminococcaceae.