Table 3 Performance of logistic regression A approach (LR-A), logistic regression B approach (LR-B), pathogenicity score (PS) and the Bayesian approach (BS) on the entire Clinvitae validation set (“all” columns) and on the subset of Clinvitae variants that are interpreted as VUS by the ACMG/AMP guidelines (“VUS”) columns.

From: A machine learning approach based on ACMG/AMP guidelines for genomic variant classification and prioritization

 

LR-A

LR-B

PS

BS

All

VUS

All

VUS

All

VUS

All

VUS

Accuracy

0.9717

0.8630

0.9702

0.8507

0.9559

0.7793

0.9082

0.6161

Precision

0.9901

0.9582

0.9943

0.9703

0.9870

0.9435

0.9979

0.9679

AUC

0.9667

0.8429

0.9641

0.8270

0.9476

0.7443

0.8872

0.5489

F1

0.9643

0.8147

0.9621

0.7921

0.9433

0.6633

0.8727

0.1817

Recall

0.9398

0.7087

0.9318

0.6692

0.9033

0.5114

0.7755

0.1003

Balanced accuracy

0.9667

0.8439

0.9641

0.8270

0.9476

0.7743

0.8872

0.5489

MCC

0.9419

0.7302

0.9542

0.7193

0.9098

0.5738

0.8184

0.2357

Back to article page