Skip to main content

Table 12 SNV results for HG003 chr21-22, Hybrid

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

  DeepVariant HELLO
Precision Recall Errors Precision Recall Errors
20 × 15x 0.998487 0.997302 362 0.998581 0.997442 342
20 × 30x 0.998732 0.997512 323 0.998779 0.997802 294
20 × 60x 0.998709 0.997639 314 0.999221 0.998035 236
30 × 15x 0.998894 0.997407 318 0.998860 0.997570 307
30 × 30x 0.999104 0.997802 266 0.998907 0.997837 280
30 × 60x 0.999011 0.997791 275 0.999221 0.998058 234
40 × 15x 0.998929 0.997372 318 0.998953 0.997523 303
40 × 30x 0.998976 0.997651 290 0.998977 0.997826 275
40 × 60x 0.998953 0.997733 285 0.999244 0.998023 235
50 × 15x 0.999011 0.997349 313 0.999023 0.997500 299
50 × 30x 0.999057 0.997826 268 0.999128 0.997860 259
50 × 60x 0.998860 0.997826 285 0.999302 0.998093 224
  1. Values in bold indicate the best operating point for each tool for the corresponding metric