I couldnt find GATK verison GenomeAnalysisTK-2.7-2-g6bda569 on your website. Kindly Help
Do you have a particular reason to seek that version? I bet you may go with GATK3 or 4 without much of a problem. Even if that is not a solution 2.7.4 seems to be available in the archived versions site.
i have added more samples for which now i need to jointly call variants that was done previously using the above version.When i used 2.7.4 i am getting less number of variations as was called by 2.7.2
It is best practices to use the latest version of GATK, even if the older versions yielded more variants. Those may not be the best variants that reflects your experiments, some of them could be spurious due to less robust statistical inference.
I notice that the 2.7.2 is not available on the download site, here is the general warning that is posted for archived versions of GATK:
"We make these available for purposes such as comparative testing and project consistency, but you should be aware that these old versions are completely unsupported and that if you choose to use them, you do so at your own risk! If you encounter any bugs in older versions and ask us for help, our only response will be please upgrade to the latest version and see if it still happens because we're not able to put effort into dealing with issues that have already been fixed in later versions. Whenever possible, you should always use the latest and greatest version of GATK. Note also that the system requirements for old versions may be different than what is documented for the current version; we do not document these explicitly here but you may find that information in the Version History."
It is possible that all of those extra variants found by 2.7.2 could be an artifact of bugs in the code, and not biologically true variants.
If you look at the Version History for updates to 2.8, all of the bug fixes are explained.
Is there a reason to not update to the latest version of GATK4?
Using the latest version of the tool will ensure that you have the most robust variants in your data set with less false discoveries, which will ultimately be more defensible in a publication.