Plasma metabolomics provides new insights into the relationship between metabolites and outcomes and left ventricular remodeling of coronary artery disease

Table 2 Independent metabolic signature selection using LASSO

Terms	Coefficient (β)	HR	Frequency
LASSO based signature selection for death
Dulcitol	0.22	1.25	200
4-Acetamidobutyric acid	0.29	1.34	200
N6-succinyl adenosine	0.05	1.05	195
l-Cystine	0.08	1.08	191
β-Pseudouridine	0.05	1.05	173
2-(Dimethylamino) guanosine	0.02	1.02	137
Kynurenine	0.04	1.04	43
3,3ʹ,5-Triiodo-l-thyronine	− 0.12	0.89	43
d-Sorbitol	0.04	1.04	21
DL-P-hydroxyphenyllactic acid	0.02	1.03	21
Phenyllactate (PLA)	0.01	1.01	11
Cyclic AMP	0.02	1.02	5
S-(5-Adenosy)-l-homocysteine	0.02	1.02	2
LASSO based signature selection for MACE
4-Acetamidobutyric acid	0.06	1.06	200
l-Cystine	0.06	1.06	200
l-Tryptophan	− 0.24	0.79	200
Dulcitol	0.10	1.10	200
5-Methyluridine	0.28	1.33	200
Kynurenine	0.22	1.25	200
Phenyllactate (PLA)	0.10	1.11	200
LysoPC 20:2	− 0.51	0.60	200
d-Sorbitol	0.02	1.02	199
LysoPC 20:1	− 0.04	0.96	193
N6-succinyl adenosine	− 0.01	0.99	2

The regression coefficients were calculated by averaging the coefficients obtained from tenfold cross-validation lasso Cox regression with 200 repeats, adjusted for 17 main clinical confounders. The confounders included age, sex, AST, eGFR, DM, HyperT, CHOL, HDLC, PPI, ACEI, BB, CCB, current smoking, family history of CVD, SYNTAX, SBP, and GLUC. The variables that appear zero times were removed and the variables left were further selected to develop a predictive model, abbreviations are as in Table 1

ISSN: 2045-3701