Benchmarking HLA LOH calls on chr6 loss

[anabbi]

Hello and thank you for your great tool! It works great on my samples. I have a question regarding one of my samples, which is ultra-pure tumor (> 90% based on ASCAT) and a chromosome 6 loss.

First of all, I did a benchmarking on HLA types against Optitype. The normal sample types were complete match between SpecHLA and Optitype. For the tumor sample:
SpecHLA called HLAA01:23, Optitype called HLAA01:01
SpecHLA called HLAB44:05, Optitype called HLAB08:01
SpecHLA called HLAC05:01, Optitype called HLAC07:01
The discrepancy, I think, is because of the chromosome loss. Here is the results of LOH call:

Sample | HLA | Allele1 | Allele2 | copyratio | KeptHLA | LossHLA | Freq1 | Freq2 | Purity | Het_num | LOH
7316-219 | A | A01:23:01 | A02:01:01:01 | 0:02 | A02:01:01:01 | A01:23:01 | 0.143 | 0.857 | 0.94 | 53 | Y
7316-219 | B | B44:02:01:17 | B44:05:01:01 | 1:01 | B44:02:01:17 | B44:05:01:01 | 0.686 | 0.314 | 0.94 | 19 | N
7316-219 | C | C05:01:01:32 | C05:01:01:02 | 1:01 | C05:01:01:32 | C05:01:01:02 | 0.562 | 0.438 | 0.94 | 40 | N
7316-219 | DPA1 | DPA101:03:01:02 | DPA101:03:01:02 | 0:02 | DPA101:03:01:02 | DPA101:03:01:02 | 0.13 | 0.87 | 0.94 | 83 | Y
7316-219 | DPB1 | DPB104:02:13 | DPB104:02:01:10 | 0:02 | DPB104:02:01:10 | DPB104:02:13 | 0.186 | 0.814 | 0.94 | 43 | Y
7316-219 | DQA1 | DQA105:03:01:01 | DQA105:01:01:02 | 1:01 | DQA105:03:01:01 | DQA105:01:01:02 | 0.712 | 0.288 | 0.94 | 5 | N
7316-219 | DQB1 | DQB103:04:01 | DQB103:01:01:01 | 0:02 | DQB103:01:01:01 | DQB103:04:01 | 0.11 | 0.89 | 0.94 | 137 | Y
7316-219 | DRB1 | DRB112:01:01:01 | DRB112:01:01:01 | 1:01 | DRB112:01:01:01 | DRB112:01:01:01 | 0.29 | 0.71 | 0.94 | 18 | N

I was expecting to see all HLAs being called LOH, due to the chromosome 6 loss.

My question to you is that whether you have tried SpecHLA on a highly pure tumor with CNVs on chr6? And I would like to get your opinion on these results and some guidance on how to improve it. Is there some parameters that need to be adjusted depending on the sample set?

Thank you,
Arash

[wshuai294]

Hello,

To improve it, you could use a higher value of the MAF by -r The minimum Minor Allele Frequency (MAF), default is 0.05 for full length and 0.1 for exon typing.

[anabbi]

Thank you for your suggestion. I will give it a try!