Questions about the software #3
Description
Hi,
It's a very interesting and innovative work！And I encountered some problems when using the software：
1.When I use the function “Predicting loops for unknown cell types”, I only have chip-seq narrowpeak data of CTCF H3K27ac H3K4me3 (or even less), should I choose the Median model in the pre-trained model?
2.When I try to use data from different organizations to predict loops of different organizations, have you evaluated whether using a pre-trained model or a new Customize model will achieve better results?
3.When I use the function “Predicting loops for unknown cell types”，When running FeatureGenerator_nobed.sh, the target_comb_final_sort.tmp file is generated. This is a file with 120 million lines. It creates a lot of memory pressure on the subsequent Rscript and annotatePeaks.pl. It cannot run successfully in Linux with 100G running memory. Is there any way to optimize it, or is there something wrong with one of my steps?
I still look forward to your reply.
Thanks!