Hello! I've asked this question in the official GATK forum but I believe it's not a very high-priority support request. I'm hoping someone can direct me to resources around acquiring or building standalone gatk tools, specifically HaplotypeCaller. All of my research has led to the monolithic gatk wrapper (either local, spark, or in docker). The big tool is brilliant and I've been using it thus far, but it's pretty hefty.

My specific use-case involves outsourcing the HaplotypeCaller step of my workflow to serverless compute instances. These accept a container image (the tools) and a payload (the alignments to call). Currently, the gatk image I have clocks in at 4.56GB, which presents a huge amount of overhead, most of which is not necessary for my use case. I'll also mention that these serverless instances are more often than not naive and therefore won't have the image cached. Ideally I'd like a container image with the absolute minimum reqs needed to run HaplotypeCaller.

I've had a look at the github repo and I suspect I can hack something together from there. Since I'm not familiar with Java I figured I'd see if anyone has traveled this path before I start banging my head against a wall...

Any help is greatly appreciated - thank you for your time and attention!