Which one is a more reliable genome build to use: hg37 or hg38?
The first GWAS data was on b36
Almost all GWAS data is kept on b37
But this is because most downstream programs have not updated to b38
Hg38 is the newest build. However, many LD panels are still on hg37, so hg37 is still a bit easier to work with. Things are moving to hg38. I would say in the next 1-2 years there will be more platforms that use hg38 primarily.
It’s not so much about what’s most reliable as it is about what’s most feasible. If you have an N=500,000 sample like UK Biobank that has Hg37, it’s often most convenient to keep it on the build it’s on instead of trying to do liftover. You just have to make sure any resources you use with it match that build.
Mostly you just want to be aware of it so you don’t accidentally use reference data for the wrong build that doesn’t match your data.