Table 3.

Machine learning models based on IgG glycans and/or inflammatory markers.

		Lasso model					Training model data					Testing model
Sex		Selected predictor(s)	n	R2	Adjusted R2	MSE	n	MSE	SD	Low 95% CI	Upper 95% CI	n	MSE	SD	Low 95% CI	Upper 95% CI
Men	IgG glycans	G0 + G0FB + G1FB + A2	253	0.23	0.21	44.19	202 (80%)	44.22	2.22	39.88	48.57	51 (20%)	45.46	9.27	27.30	63.62
	IgG glycans	G0 + G0FB + G1FB + A2	100	0.36	0.34	38.68
	Inflammatory markers*	CXCL9 + Eotaxin	100	0.02	0.00	59.42
	IgG glycans + Inflammatory markers	G0 + G0FB + G1FB + A2 + CXCL9 + Eotaxin	100	0.37	0.33	38.02
Women	IgG glycans	A1F + G2F	235	0.31	0.31	36.25	188 (80%)	36.40	1.78	32.91	39.88	47 (20%)	36.25	7.27	22.00	50.49
	IgG glycans	A1F + G2F	100	0.25	0.23	32.16
	Inflammatory markers*	CXCL9 + Eotaxin	100	0.23	0.22	32.82
	IgG glycans + Inflammatory markers	A1F + G2F + CXCL9 + Eotaxin	100	0.41	0.38	25.32

^*Pre-selected

MSE = Mean Squared Error

SD = Standard Deviation