Skip to main content

Table 2 DKMcost125 model performance on test sets

From: Identifying Crohn’s disease signal from variome analysis

Cutoff CD-test WTCCC and GTEx^
TP FN TN FP Prec % Rec % MCC TP FN TN FP Prec % Bal. Prec % Rec % MCC
45 9 40 13 0 100.0 18.4 0.212 384 2104 543 1 99.7 98.8 15.4 0.176
14.3 36 13 5 8 81.8 73.5 0.107 1432 1056 476 68 95.5 82.2 57.6 0.346
0 44 5 2 11 80.0 89.8 0.067 1924 564 302 242 88.8 63.5 77.3 0.279
  1. TP true positive (CDs predicted to be CD by AVA,Dx), FN false negative (CDs predicted to be HC), TN true negative (HCs predicted to be HC), FP false positive (HCs predicted to be CD), Prec precision, Rec recall of identifying CD patients (Eq. 2), Bal. Prec balanced precision, where the number of CD and HCs is standardized to represent 50% of the data, each. MCC is in Eq. 3. A more detailed performance evaluation is in Additional file 2: Tables S4 and S5
  2. ^WTCCC and GTEx panels were combined for evaluation since WTCCC contains only CD individuals and GTEx contains only HC individuals