Urinary Metabolite Signatures for Predicting Elderly Stroke Survivors with Depression

Figures & data

Table 1 Detailed Information of the Elderly Recruited Subjects

Variables	Training Set				Testing Set
	HCs	Strokes	PSD	P-value	HCs	Strokes	PSD	P-value
Number	44	86	82	–	34	36	42	–
Age	65.20 (3.71)	66.88 (4.84)	67.22 (4.69)	0.49^a	65.41 (2.74)	67.19 (5.42)	66.17 (4.34)	0.99^a
BMI	21.68 (2.59)	23.13 (2.80)	23.47 (2.99)	0.12^a	21.98 (3.00)	22.37 (2.60)	23.73 (3.13)	0.20^a
Sex (M/F)	18/26	26/40	39/43	0.57^b	17/17	15/21	21/21	0.71^b
HDRS	0.25 (0.61)	1.60 (1.85)	19.99 (5.42)	<0.00001^a	0.59 (1.05)	1.67 (1.56)	20.81 (5.69)	<0.00001^a

Notes: ^aP-value was obtained from one-way ANOVA test. ^bP-value was obtained from Chi-square test.

Abbreviations: M, male; F, female; HCs, healthy controls; BMI, body mass index; PSD, post-stroke depression; HDRS, Hamilton Depression Rating Scale.

Figure 1 Metabolomic analysis of urine samples from different groups: (A) OPLS-DA model built with training set showed that the elderly PSD subjects (blue dot) could be intelligibly separated from the elderly non-depressed subjects (green dot); (B) T-predicted scatter plot showed that the model could effectively predict the elderly PSD subjects (blue dot) and elderly non-depressed subjects (green dot) from the testing set.

Table 2 Important Metabolites Responsible for Diagnosing Elderly PSD Patients

Metabolites	p-value^a	Adjust p-value^b	R^c	Fold Change^d
Palmitic acid	8.11E-06	9.57E-05	0.456159	0.82
Azelaic acid	8.95E-07	1.76E-05	0.450246	1.29
Sucrose	2.41E-06	3.56E-05	0.372082	0.97
Glyceric acid	4.09E-05	4.02E-04	0.366423	1.10
α-aminobutyric acid	9.63E-05	7.10E-04	0.352394	0.64
Lactic acid	7.37E-07	2.17E-05	0.284821	0.74
Fructose	1.06E-04	6.97E-04	0.217637	0.77
3-hydroxyisobutyric acid	1.98E-03	8.35E-03	−0.271174	−0.22
Sorbitol	8.00E-02	1.52E-01	−0.301555	−0.29
3-hydroxyphenylacetic acid	1.52E-02	4.73E-02	−0.33488	−0.60
Indoxyl sulphate	6.18E-03	2.14E-02	−0.355926	−0.30
Phenylalanine	1.37E-03	6.24E-03	−0.413967	−0.40
Tyrosine	1.96E-07	1.16E-05	−0.463675	−0.52

Notes: ^ap-values obtained from nonparametric Mann–Whitney U-test. ^bAdjust p-values obtained using Benjamini-Hochberg false discovery rate. ^cCorrelation coefficient obtained from OPLS-DA model, positive and negative values indicated higher and lower levels in elderly PSD patients, respectively. ^dLog2 fold-change, positive and negative values indicated higher and lower levels in elderly PSD patients, respectively.

Figure 2 Relative concentrations of these eight urinary metabolite biomarkers for elderly PSD (mean±standard deviation).

Figure 3 Diagnostic performance of potential biomarker panel for elderly PSD: (A) AUC value in the training set; (B) AUC value in the testing set.

Figure 4 Pathway analysis using the differential urinary metabolites: (A) three metabolic pathways were found to be significantly affected in elderly PSD subjects; (B) the main metabolites involved in these pathways. Red and green ellipses represent the significantly increased and decreased metabolites, respectively, identified in this study.