In using in silico HLA typing programs such as OptiType/Polysolver/Athlates for the identification of neoantigens in cancer - should the control or tumour sample be used to derive HLA types? I've seen both been done in studies and am quite puzzled at how this can be. It makes sense to do HLA typing from the tumour because of immune escape mechanisms like HLA downregulation which often render the HLA type in the tumour different from normal cells. Anyone have thoughts?
Thank you for your answer. Perhaps I should clarify - for neoantigen identification generally there is a step that generates peptides and a second step that predicts binding of those peptides to the HLA types of the patient. I was referring to which HLA types to use for this, where it can only be one set - either tumor or normal.
I cannot answer with 100% accuracy and would therefore, personally, look through the methods of previous work to see what they have done. Obviously the idea is that the neoantigens provoke an immune response from normal cells. One has to of course monitor other things such as the now well known PDL1 checkpoint.