Table 2. Model building before (T1k) and after re-modeling of misclassified samples (Mix) using lasso.
| SNP Dataset | Samples (AD, Ctrl) | Input SNPs | Model Length | Accuracy | Sensitivity | Specificity | AUC |
|---|---|---|---|---|---|---|---|
| T1k* | 448 (163, 285) | 1,106 | 433 | 0.801 | 0.798 | 0.804 | 0.801 |
| Missed (test)* | 89 (56, 33) | 461 | 49 | 0.929 | 0.857 | 0.952 | 0.905 |
| SNPs Mix* | 448 (163, 285) | 1,567 | 493 | 0.797 | 0.816 | 0.786 | 0.801 |
| Models Mix* | 448 (163, 285) | 482 | 358 | 0.844 | 0.834 | 0.849 | 0.842 |
| T1k (train) | 1,382 (650, 732) | 1,106 | 433 | 0.985 | 0.984 | 0.986 | 0.985 |
| Missed (train) | 21 (9, 12) | 384 | 8 | 0.833 | 1.000 | 0.750 | 0.875 |
| SNPs Mix* | 448 (163, 285) | 1,490 | 494 | 0.813 | 0.816 | 0.811 | 0.813 |
| Models Mix* | 448 (163, 285) | 501 | 391 | 0.848 | 0.828 | 0.860 | 0.844 |
* Indicates that the evaluation was performed in the test set. AD, CTRL refer to Alzheimer’s Disease and Control subjects.