Skip to main content
. 2020 May 23;9(5):giaa052. doi: 10.1093/gigascience/giaa052

Table 7:

Summary of performance improvements obtained by parSMURF by tuning the learning parameters on the Mendelian dataset

Parameters Training set mean (SD) Test set mean (SD)
AUROC AUPRC AUROC AUPRC
Default parameters 0.99958 (0.00005) 0.53143 (0.02714) 0.99281 (0.00032) 0.42332 (0.00391)
Grid search 0.99986 (0.00009) 0.60023 (0.15977) 0.98968 (0.00140) 0.47025 (0.00585)
Bayesian optimizer 0.99989 (0.00011) 0.65388 (0.22123) 0.99264 (0.00043) 0.46153 (0.00302)

Results are averaged across 10 repetitions of the 10-fold cytoband-aware CV. AUROC and AUPRC are averaged across the 10 folds; standard deviation (SD) in parentheses. Default parameters: nParts 100, fp 2, ratio 3, nTrees 10, mTry 5.