Problem with minimum indel AF filter

[keremozdel]

Hello,

I'm facing an issue while trying to filter indels based on minimum allele frequency. After I set the --indel_min_af parameter to 0.3, I can still observe indels with lower AF. I'm working with amplicon data and here is my code: run_clair3.sh --gvcf --platform="ont" --indel_min_af=0.3 --snp_min_af=0.08 --model_path=${PATH}/r941_prom_sup_g5014 --bam_fn=${BAM} --bed_fn=${BED} --ref_fn=${REF} --var_pct_full=1 --ref_pct_full=1 --var_pct_phasing=1 --output=${OUTPUT}

The problem is that there are some questionable indels with lower af than the specified threshold in my result. Here are some examples:

CHROM | POS | ID | REF | ALT | QUAL | FILTER | INFO | FORMAT | SAMPLE

-- | -- | -- | -- | -- | -- | -- | -- | -- | -- chr17 | 43123064 | . | G | GA | 2.44 | PASS | AD=226,61;DP=719 | GT:GQ:DP:AD:AF:PL | 1/1:2:719:226,61:0.0848:17,5,0 chr17 | 43095603 | . | C | CA | 4.44 | PASS | AD=148,58;DP=408 | GT:GQ:DP:AD:AF:PL | 0/1:4:408:148,58:0.1422:16,0,9 chr17 | 43071847 | . | C | CA | 9.09 | PASS | AD=164,95;DP=654 | GT:GQ:DP:AD:AF:PL | 0/1:9:654:164,95:0.1453:13,0,36

I suspect most of the reads are filtered due to low MQ and BQ, thus I'm getting these problematic results. And that's why I want to filter them. I was wondering why Clair3 doesn't filter them as expected.

Thanks in advance!

[aquaskyline]

The AF in snp_min_af and indel_min_af means the sum of AF of all alternative alleles at a site. The AF of each VCF record is just an alternative allele at a site.

A post-processing step to remove variants with unexpected AF is suggested.

[keremozdel]

Thank you for your response.

If I set the snp_min_af filter to 0.3 and observe 100 bases at a specific position, with the reference allele being G, and I find 20 T's and 5 A at that position, Clair3 would not call any variant because the alternative allele frequency is 0.25, which is below the specified filter threshold of 0.3. Is my understanding correct?

[aquaskyline]

Yes.