I am reaching out to everyone on the feasibility/usefulness of converting (liftOver) mm10 to hg19 co-ordinates for somatic point mutations, with a view to use the rich human resources for functional impact predictions for mutations.
Two specific queries related this this:
a) I have VCF files that represent somatic point mutations in mouse mm10 and I have liftedOver mm10 to hg19 co-ordinates using the corresponding chain file from UCSC. I observe that 60% to 85% of the mm10 co-ordinates were 'mapped' to human co-ordinates. I can now upload these 'humanized' VCF files to tools like Ingenuity Variant Analysis or Oncotator for functional impact predictions. Is this biologically correct ? Any caveats in this approach ?
b) For the liftover from mm10 to hg19, I had to convert the mouse VCF files to BED files with the minimal required information (chr, pos-1, pos, Ref/Alt, Strand). The output of liftOver is also in BED format with the same information. Most functional impact prediction tools (such as IVA) require VCF files as input, preferably with additional information about the calls. How do I go from BED back to VCF ? One way would be to insert the sample specific genotype/quality information from the mouse VCF files back into the human liftedOver VCF files, but then we are mixing human co-ordinates with genotype/quality evidence from mouse.
Appreciate your time and will look forward to comments.