Fig. 4

Repeat count estimation on 11 Huntington’s disease samples with CAG repeats and on NA12878. a The correlation of estimated repeat counts against benchmark repeat counts on ~15X downsampling data (3 times for each data). b The difference of estimated repeat counts minus benchmark repeat counts for the four methods. c Averaged absolute difference (Error) of estimated repeat counts and benchmark repeat counts. “All” for all repeat counts, while “pathogenic” for only pathogenic repeat counts. d The distribution of estimated repeat counts from all long reads for a sample (ND30422) whose benchmark repeat counts are 18 and 40 (shown by a down arrow in magenta). e The distribution of estimated repeat counts from all long reads for another sample (ND30626) whose benchmark repeat counts are 21 and 41 (shown by a down arrow in magenta). The distribution for more samples is in Additional file 1: Fig. S5. d, e: x-axis for estimated repeat counts; y-axis for the number of supporting reads for each estimated repeat count. f The difference of supporting reads between DeepRepeat and other methods (in percentage = (o-D)*100/D where D is the number of supporting reads by DeepRepeat, while “o” is the number of supporting reads by other methods (RepeatHMM, Tandem-genotypes and STRique)