Tumour normal comparison - different normal reads reported in VCF between tumours

[laurenmcauley]

Hi, I am using ClairS (v0.4.0) for my analysis. I have two tumour samples from one individual plus a matched blood. I've used ClairS to find somatic variants between T1 vs normal and T2 vs normal. I noticed that the read depth info in the VCF outputs varies for the normal between T1 vs normal and T2 vs normal;

T1 vs normal:
GT:GQ:DP:AF:AD:NAF:NDP:NAD:AU:CU:GU:TU:NAU:NCU:NGU:NTU
0/1:16:34:0.2647:0,9:0:29:0,0:9:0:25:0:0:1:28:0

T2 vs normal:
GT:GQ:DP:AF:AD:NAF:NDP:NAD:AU:CU:GU:TU:NAU:NCU:NGU:NTU
0/1:6:40:0.2000:31,8:0.0286:35:31,1:8:0:32:0:1:1:33:0

There's higher overall depth for the normal in the T2 vs normal run, with more reads supporting both the reference allele (G) and alternative alleles A and C. I have checked in IGV and these reads are present in the normal sample. I am confused as to why different reads from the same normal sample are included in the outputs of one run of ClairS and not another? The same normal BAM is provided in each comparison.

I don't believe this is due to downsampling as the coverages in IGV (including secondary, supplementary alignments, MQ > 0) don't reach above 54 in tumours or 40 in the normal.

This is the command I ran:

run_clairs \
--tumor_bam_fn ${TOI_BAM} \
--normal_bam_fn ${NORMAL_BAM} \
--ref_fn ${HG38} \
--platform ont_r10_dorado_hac_5khz \
--threads 30 \
-q 8 \
--enable_clair3_germline_output \
--output_dir ${OUTPUT_DIR} \
--output_prefix ${TOI_ID}.clairs.snvs \
--enable_indel_calling \
--indel_output_prefix ${TOI_ID}.clairs.indels \
--sample_name ${TOI_ID}

Any insights would be appreciated!

[zhengzhenxian]

Hi @laurenmcauley,

Did two variant records output the same variant? The read depth in your case shouldn't be due to a downsampling issue. We'll conduct some internal tests and get back to you soon.

[laurenmcauley]

Hi,

Yes, the same variant was output (G -> A). In T1 vs normal, the variant passed while it was flagged with LowQual in T2 vs normal. Thanks!