Fig. 2

a Computational trails. Each combination of tools and parameter settings can be imagined as a unique computational “trail” that is executed simultaneously with all the other possible trails to create a collection of output objects. As an example, one of these possible trails is presented in red. The raw reads were aligned, filtered and then binned in 40 kb resolution matrices. Our own naïve matrix scaling method was then used for matrix correction and domains were called using TopDom [31]. b HiC-bench pipeline task architecture. All pipeline tasks are performed by a single R script, “pipeline-master-explorer.r”. This script generates output objects based on all combinations of input objects and parameter scripts while taking into account the split variable, group variable and tuple settings. The output objects are stored in the corresponding “results” directory. As an example, domain calling for IMR90 is presented. The filtered reads of the IMR90 Hi-C sample (digested with HindIII) are used as input. The pipeline-master-explorer script tests if TAD calling with these settings has been performed and if not it calls the domain calling wrapper script (code/hicseq-domains.tcsh) with the corresponding parameters (e.g., params/params.armatus.gamma_0.5.tcsh). After the task is complete, the output is stored in the corresponding “results” directory