Changes in the liver transcriptome of farmed Atlantic salmon (Salmo salar) fed experimental diets based on terrestrial alternatives to fish meal and fish oil

Table 7 Results obtained from the stepwise discriminant analysis performed on log₂-transformed qPCR-analyzed transcripts

Grouping variable	Biomarkers included	Function	Eigenvalue^a	Canonical correlation^b	Wilks’ Lambda^c				Accuracy (%)^d
Grouping variable	Biomarkers included	Function	Eigenvalue^a	Canonical correlation^b	λ	χ²	df	p-value	Accuracy (%)^d
Diet (MAR, ABP, VEG)	sqs, gck, gadd45ba, sgk2b, fabp3a	1	4.50	0.904	0.075	67.2	10	< 0.001	90
Diet (MAR, ABP, VEG)	sqs, gck, gadd45ba, sgk2b, fabp3a	2	1.41	0.765	0.415	22.9	4	< 0.001	90
Growth (High, Low)	sgk2b, gck, mxa, fabp3b	1	2.02	0.818	0.331	29.9	4	< 0.001	94
EPA + DHA (High, Low)	fabp3a, igd	1	0.394	0.532	0.717	7.98	2	0.019	67

^aRatio between sum of squares among groups and within groups calculated from the discriminant scores of each function. Strong discriminant functions present eigenvalues above one
^bMeasure of the association between the groups in the dependent variable (i.e., diet, growth, and EPA + DHA levels) and the discriminant function
^cWilk’s lambda is used to test the significance of the discriminant functions
^dAccuracy of the function(s) in predicting an individual’s membership to its corresponding phenotypic class

ISSN: 1471-2164