Table 1.

Classification dataset composition, and specification of constructed OPLS-DA models with the output of the model validation. Dataset includes both positive and negative ionization mode.

Model	Total Number of Instances	Number of Model Components (to + tp) a	Model Characteristics b	Cross-Validated ANOVA c (p-Value)	Permutation d
6 HPI vs. 0 HPI	10	1+0	R2Y = 0.998; Q2 = 0.995	1.25 × 10−8	Good
12 HPI vs. 0 HPI	10	1+0	R2Y = 1.000; Q2 = 0.998	1.98 × 10−10	Good
24 HPI vs. 0 HPI	10	1+0	R2Y = 0.998; Q2 = 0.993	2.34 × 10−8	Good
36 HPI vs. 0 HPI	10	1+0	R2Y = 0.997; Q2 = 0.992	5.29 × 10−8	Good
12 HPI vs. 6 HPI	10	1+1	R2Y = 0.985; Q2 = 0.940	3.23 × 10−3	Good
24 HPI vs. 6 HPI	10	1+1	R2Y = 0.997; Q2 = 0.990	3.42 × 10−5	Good
36 HPI vs. 6 HPI	10	1+1	R2Y = 0.998; Q2 = 0.997	1.94 × 10−6	Good
24 HPI vs. 12 HPI	10	1+0	R2Y = 0.998; Q2 = 0.993	2.46 × 10−8	Good
36 HPI vs. 12 HPI	10	1+0	R2Y = 0.993; Q 2= 0.99	8.96 × 10−8	Good
36 HPI vs. 24 HPI	10	1+1	R2Y = 0.995; Q2 = 0.986	8.63 × 10−5	Good

^a with t_o the orthogonal and t_p the predictive component; ^b with R²Y the variation in Y that is explained by the model, and Q² the predictive ability of the model. Q² > 0.5 indicates good model quality; ^c a cross-validated ANOVA p-value < 0.05 indicates good model quality; ^d good permutation testing is achieved if R²Y and Q² values of the models based on the permutated data are significantly lower than those based on the real data set. “Q² > 0.5”, “p-value < 0.05” and “good permutation test” mean that OPLS-DA can successfully separate comparing groups.