Table 4.

Baseline logistic regression model and generalized regression elastic net models on the prediction of colorectal cancer from gene-metabolite interaction, with one interaction term.

	Logistic Regression Original Model		Generalized Regression Elastic Net Model
			AICc Validation		Leave-One-Out Validation
Parameters	Estimate	p (X2)	Estimate	p (X2)	Estimate	p (X2)
(Intercept)	−5.6	0.93	0.4	0.78	1.1	0.45
MMA * Gene mutations	−42	0.68	−30	<0.0001	−11	<0.0001
Homocysteine	−15	0.77	−12	<0.0001	−5.7	<0.0001
Methyl-folate	14	0.69	9.1	<0.0001	3.4	0.0019
Gene mutations	14	0.86	11	<0.0001	4.0	0.0188
Vegetable intake	28	0.62	17	<0.0001	5.6	0.0005
Age	−14	0.63	−8.7	<0.0001	−2.9	0.0024
MMA	−0.4	0.996	−1.7	0.28	0	1.0
Misclassification Rate	0.2	–	0.03	–	0.04	–
AICc	27	–	26	–	–	–
Area under the curve	1.0	–	0.998	–	0.997	–

MMA: Methylmalonic acid; *: Interaction; –: Not available; AICc: Akaike’s information criterion with corrections: AUC: Area under the curve.