8.1 years ago by
Santiago de Compostela, Spain
I did open a discussion in the GATK forum regarding this issue, as you may check here. what I was asking was very simple, when they raised that version 2 that changed GATK as we knew it: should we stay with latest version 1.6, should we go for version 2 full, or should we go for version 2 lite? I should say that I first thought to do what you did, to stay with my 1.6 version I had already optimized and integrated in my pipeline, but Mark DePristo's answer was very clear: if you would like to continue with the 1.6 development you should go for version 2 lite, and if you would prefer to include the latest (and rather radical) changes and modules developed then you should go for version 2 full.
as you say, GATK's team has recently stated that they're about to merge both efforts, because it didn't make sense for them to maintain 2 different branches of the same concept. what they've stated is that version 2.4 would be only the 2.4 full, and that the lite version will no longer evolve. so if you want to stay with the previous algorithms then you shouldn't stay at 1.6, but at 2.3 lite. if the reason is algorithmic I could understand your point, but if the reason is the license type this will change with version 2.4, where they'll not only release it freely again, but also they'll open the source code again. as you may have already imagined, I'm currently using the latest full version, and I would honestly not see any reason why someone would prefer not to update. it's true that even if they've improved a lot the multithreading capabilities I had before (versions <2) slightly better performance, but the quality and reliability of the results that we are obtaining are not comparable.