Fig. 3From: ADS-HCSpark: A scalable HaplotypeCaller leveraging adaptive data segmentation to accelerate variant calling on SparkThe percentage of running time per 5% data blocks after all the data blocks are sorted. The HaplotypeCaller algorithm is separately executed on every original data blocks of dataset D1 and their respective running time is recorded. Then data blocks are sorted by execution time from high to low and the percentage of time consumption for per 5% data blocks is counted, as is shown in the figureBack to article page