Skip to main content
. 2022 Jan 14;109(2):195–209. doi: 10.1016/j.ajhg.2021.12.007

Figure 4.

Figure 4

StrVCTVRE performance is consistent across SV size and performance improves on more reliably classified data

(A) Across three size ranges, StrVCTVRE accurately classified variants in an independent test set. In this ROC comparison of StrVCTVRE (solid line) and SVScore (dotted line), three size ranges of SVs were considered. StrVCTVRE performed very well on large and small SVs, while performing well on mid-sized variants. AUC with 95% confidence interval is in parentheses.

(B) When presented with data that are more reliably classified, StrVCTVRE’s performance improved. ROC plot showing StrVCTVRE’s performance increased as SV contribution to proband phenotype increases from set 3 (includes less confidently classified SVs) to set 2 and from set 2 to set 1 (most confidently classified SVs). The performance of SVScore did not significantly differ between the sets. AUC with 95% confidence interval is in parentheses.