Skip to main content

Table 5 Indel results for HG003 WGS, Hybrid

From: HELLO: improved neural network architectures and methodologies for small variant calling

  DeepVariant HELLO
Precision Recall Errors Precision Recall Errors
20 × 15x 0.987302 0.988139 12,661 0.994011 0.993572 6381
20 × 30x 0.989490 0.991368 9894 0.996103 0.995849 4137
20 × 60x 0.989477 0.993166 9012 0.997052 0.996892 3114
30 × 15x 0.993222 0.993308 6939 0.996341 0.995970 3951
30 × 30x 0.993277 0.994482 6324 0.997269 0.997177 2856
30 × 60x 0.992232 0.995110 6570 0.997664 0.997566 2453
40 × 15x 0.995176 0.995084 5016 0.997139 0.996838 3095
40 × 30x 0.995033 0.995756 4755 0.997790 0.997649 2345
40 × 60x 0.993734 0.995974 5338 0.998024 0.997841 2125
50 × 15x 0.996197 0.995881 4077 0.997726 0.997257 2576
50 × 30x 0.995982 0.996470 3895 0.998047 0.997816 2126
50 × 60x 0.994690 0.996468 4582 0.998192 0.998016 1949
  1. Values in bold indicate the best operating point for each tool for the corresponding metric