Table 1.
Classification error rates (CER) for SNF models (separate and integrated models) and for PLS‐DA and DIABLO models presented for all studies.CER is validated by M‐fold cross‐validation tests (respectively, 7‐, 5‐, and 10‐fold for studies 1, 2, and 3)
| Study 1 CER ± SEM | Study 2 CER ± SEM | Study 3 CER ± SEM | ||||
|---|---|---|---|---|---|---|
| Milk versus yogurt | Non‐obese versus obese | All diets | SFA diet versus MED diet | SFA diet versus MUFA diet | MUFA diet versus MED diet | |
| SNF analysis | ||||||
| Metabolome/lipidome model | 0.21a ± 0.004 | 0.15a ± 0.003 | 0.41a ± 0.005 | 0.13b ± 0.003 | 0.29a ± 0.006 | 0.30a ± 0.006 |
| Transcriptome model | 0.14b ± 0 | 0.03c ± 0.006 | 0.38b ± 0.004 | 0.19a ± 0.005 | 0.21b ± 0.004 | 0.33b ± 0.004 |
| SNF integrated model | 0.02c ± 0 | 0.08b ± 0.002 | 0.30c ± 0.005 | 0.08c ± 0.002 | 0.23b ± 0.005 | 0.32b ± 0.004 |
| DIABLO analysis | ||||||
| Metabolome/lipidome (PLS‐DA) | 0.14a ± 0 # | 0.13a ± 0.007 # | 0.27a ± 0.003 # | 0.11a ± 0.004 # | 0.18a ± 0.002 # | 0.19a ± 0.004 # |
| Transcriptome (PLS‐DA) | 0.14a ± 0 | 0b ± 0 # | 0.06c ± 0.002 # | 0.07b ± 0.003 # | 0.01b ± 0.003 # | 0.07b ± 0.005 # |
| DIABLO model | 0.03b ± 0.007 # | 0b ± 0 # | 0.08b ± 0.003 # | 0.06b ± 0.004 # | 0.02b ± 0.003 # | 0.08b ± 0.005 # |
Different letters (a–c) indicate significant differences between CERs for comparisons between models for each study (as assessed by linear mixed‐effect models with post hoc pairwise comparisons, p adj < 0.05). #indicates a difference comparing equivalent models created by the SNF tool and DIABLO (as assessed by paired t‐test, p < 0.05).CER, classification error rate; DIABLO, data integration analysis for biomarker discovery using latent variable approaches for “omics” studies; MED, Mediterranean; PLS‐DA, partial least squares discriminant analysis; SFA, saturated fatty acid; SNF, similarity network fusion.