Fig. 6
From: HELLO: improved neural network architectures and methodologies for small variant calling
Excess errors in DeepVariant and GATK variant calls compared to HELLO for HG003 hybrid chr21-chr22 data. Positive values indicate better performance by HELLO over competitor. Bars reference the left y-axis and indicate differences in FP and FN counts between tool (GATK or DeepVariant) and HELLO. The line plots reference the right y-axis, and indicate the differences in total error count as a fraction of total errors of the worse performing method at each coverage point. Please refer to Fig. 1 for additional explanation