Isofox and Very Polymorphic Human Genes

[DarioS]

Like T1K, can Isofox accurately infer HLA and KIR mRNA abundances given that hg38 poorly represents most people for such genes?

[p-priestley]

Hi Dario,

We have developed LILAC for HLA typing (https://github.com/hartwigmedical/hmftools/tree/master/lilac and https://pubmed.ncbi.nlm.nih.gov/37165135/). This can be run on both RNA and DNA.

With respect to ISOFOX, we internally tested against both Salmon and RSEM during development on ~100 samples. For HLA class 1 genes we found highly concordant abundance estimates with RSEM, but Salmon often differed quite markedly from both RSEM and ISOFOX.

Peter

[DarioS]

Can Isofox somehow take as input LILAC-inferred HLA alleles and quantify those instead of gene-level quantification?

[p-priestley]

I think it will be similar. Do you have an example where it goes wrong?

[DarioS]

No just wondering conceptually if instead of

        Sample A Samples B
HLA-A       10      0
HLA-B       20      30

something like

              Sample A   Sample B
HLA-A A*01:02     5         0
HLA-A A*02:01     5         0
HLA-B B*01:01     20        15 
HLA-B B*02:01     0         15

is possible for quantification. I am wondering about allele-specific expression (A.S.E.) of polymorphic genes.

[p-priestley]

This is very similar to what LILAC outputs. Here is the picture from our paper: (https://www.nature.com/articles/s41588-023-01367-1/figures/1)

[DarioS]

Indeed, it does.

RNA ‘expression’ of alleles
LILAC also optionally accepts a RNA bam. As per the fragments in the bam are counted for each allele in the determined type. This can be interpreted as a proxy for allele specific expression.