Table 2.

Characteristics of PLS-DA models

Cohort (number of variables in the matrix)	Components number	Observations number	R2Y (%)	Q2Y (%)	CV-ANOVA
Panel A
Grison et al. (2013) (1376)	2	20	91.9	55.2	9.40e−03
Grison et al. (2013) (95)	2	20	88.9	74.2	4.60e−04
Present article (1718)	3	39	95.5	75.2	1.70e−06
Present article (95)	2	39	88.0	80.2	7.77e−12

Discrimination	Model	Components number	Observations number	R2Y (%)	Q2Y (%)	CV-ANOVA
Panel B
A vs. B
All time points	1	2	120	26.9	16.7	2.23e−05
3 month	2	0	40	–	–	–
6 month	3	0	40	–	–	–
9 month	4	0	40	–	–	–
A vs C
All time points	5	2	120	57.3	40.3	6.42e−10
3 month	6	0	40	–	–	–
6 month	7	2	40	46.6	16.7	0.0340864
9 month	8	2	40	51.6	30.3	0.00127167
A vs D
All time points	9	3	120	53.2	33.3	6.04e−05
3 month	10	2	40	61.8	33.5	0.0105006
6 month	11	0	40	–	–	–
9 month	12	0	40	–	–	–
A vs E
All time points	13	3	117	76.7	69.6	2.44e−23
3 month	14	3	39	81.4	58.3	0.00698012
6 month	15	2	39	81.5	64.4	6.87e−07
9 month	16	3	39	83.7	70.3	4.00e−05

Panel A Models discriminating the control rats from those contaminated for 9 months at the dose 40 mg L⁻¹ in the present study and in our previous proof-of-principle study

Panel B Analyses performed on the “dose matrix” after feature selection (126 variables) to investigate the dose effect; models are discriminating the control from the contaminated rats for each dose (dose B: 0.015 mg L⁻¹; dose C: 0.15 mg L⁻¹; dose D: 1.5 mg L⁻¹; dose E: 40 mg L⁻¹) after 3, 6 and 9 months of contamination and all time-points together