TABLE 5.

Correlation Analysis

Signature	${\mathcal{S}}_{\text{men}}$		${\mathcal{S}}_{\text{women}}$		${\mathcal{S}}_{\text{all}}$
Clinical parameter	Correlation	P Value	Correlation	P Value	Correlation	P Value
Age	0.275	3.3 × 10−4***	0.153	1.1 × 10−3**	0.21	1.3 × 10−7***
BMI	0.002	9.7 × 10−1	0.110	1.9 × 10−2	0.065	1.0 × 10−1
LDL‐C	−0.205	8.6 × 10−3	−0.069	1.4 × 10−1	−0.098	1.5 × 10−2
HDL‐C	−0.084	2.8 × 10−1	−0.078	9.7 × 10−2	−0.077	5.7 × 10−2
Triglycerides	0.243	1.6 × 10−3**	0.205	1.1 × 10−5***	0.272	5.7 × 10−12***
HbA1c	0.471	1.9 × 10−10***	0.279	1.5 × 10−9***	0.353	1.4 × 10−19***
HOMA‐IR	0.423	1.8 × 10−8***	0.307	3.2 × 10−11***	0.294	1.3 × 10−13***
ASAT	0.253	1.0 × 10−3**	0.312	1.2 × 10−11***	0.322	2.5 × 10−16***
ALAT	0.170	2.9 × 10−2	0.352	1.1 × 10−14***	0.306	7.2 × 10−15***

Correlation between prediction of RF models learnt from reference signatures on the learning cohort and clinical parameters of the HUL cohort. Spearman correlation coefficient and corresponding P value were computed in R (R Foundation for Statistical Computing, Vienna, Austria). Bonferroni’s correction was applied to deal with a multiple‐comparisons situation; P values with corresponding FWER <1% and 0.1% are tagged with ** and ***, respectively.

Abbreviation: FWER, family‐wise error rate.