I've recently learned about the reference bias issue - inability to properly map NGS reads in certain genomic regions caused by the fact that our standard genomic references are linear and do not fully represent the genomic variation present in the population. According to what I read this has multiple consequences: missing structural variants or variants in highly variable regions like HLA for humans.
Now I wonder what is the best way to address this. Of course, "best" is context-dependent but I don't have any specific context so general pros and cons will be highly appreciated. I know about two approaches but probably there are more. The first one is an alt-aware mapping that's used in DRAGEN by Illumina. As far as I understand, they just keep multiple linear alternatives for some regions and try doing a "smarter" mapping for those regions. The second approach that I've found is using graph references where the reference is not linear anymore but rather a graph and mapping is taking into account this graph structure.
Thanks in advance.