Hi,
I know about the various remapping tools that allow you to remap features and coordinates from one genome assembly to another.
I was wondering if anyone has any ideas on how to compare these tools, apart from runtime and % of remapped features. (just to clarify, I'm specifically working on remapping variants, therefore I have VCF files as input)
•
written
by
juliette •
Liftover is what you would do if you cannot realign the data to the other reference genome, because you don't have the fastq files or don't want to process all genomes.
Ideally, I think you would also need the fastq/bam file aligned to the reference genomes you are comparing and call variants from that. That would be your truth dataset.
I do have the genomes in fasta files as well, I was just specifying what data I'm working on. This doesn't have much to do with the question though.
I was simply asking what could be some good ways of comparing how different remapping tools perform on the same dataset (CrossMap, UCSC LiftOver, LiftoverVcf etc).
Oh I see, thank you for explaining. It is an interesting idea, however I only have the variant coordinates for the "old" assembly in a VCF file. I don't know too much about variant calling but I think I'd need the original fastq reads to be able to call using the "new" assembly, correct?
You would need the fastq or the bam.
Log in to answer this question.
I am not sure how much of a difference the tools introduce, especially for single-nucleotide positions. The chain files are probably the main factor and I haven't seen many versions of those (for the same reference pairs).
I understand that, however I'm in the process of developing a remapping tool that doesn't use chain files, so I'd like to be able to compare it to existing tools in more ways than just runtime and % of remapped variants. But yeah, so far they don't seem to be that different in terms of runtime and % of remapped variants.
Interesting. How would it work without chain files?
The basic idea is to create "reads" from the flanking sequences of each variant (using the old assembly), and then map them to the new assembly using traditional mapping tools (ex: bowtie). I can then extract the coordinates and other relevant information to recreate a VCF output.