Hello,
I am in the process of hard filtering my SNPs (non model species, 85 whole-genomes, GATK' best practices pipeline). After removing multiallelic SNPs with SelectVariant, I already applied successfully 3 (very hard) filters to my SNP call-set, in 3 separate, sequential jobs (I was interested in the intermediary outputs): AF<0.02, MQ<59.75 and FS>10. As expected, I went form ~76000000 to ~3000000 SNPs, although (strangely) no SNPs with FS>10 were filtered out (ie. the number of PASS SNPs was the same as after the MQ<59.75 filter). This is strange because I still can see SNPs with FS>10 on my FS plot ("SNPs_4filters_FS_0-10.pdf";attached). I don't know what is going wrong here.
Now I try to move on, ignoring my FS>10 filter by using the VCF produced by any of my previous filtering rounds, and applying a QD<10 filter. However, no matter which input VCF I use (including the raw, unfiltered one), when I do so I get this WARNING:
WARN 17:05:59,793 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:06,686 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:13,170 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:14,495 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:16,299 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:17,156 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
WARN 17:06:18,855 Interpreter - ![0,2]: 'QD < 10.0;' undefined variable QD
etc.
I read that this warning is not a big deal and pass incriminated SNPs. However in my case it seems that every line produces this warning, so that I end up getting no filtered SNPs at all (ie. the number of PASS SNPs in the output VCF is the same as in the input VCF). I get the same warning and output if I try to apply filters based on ReadPosRankSum or MQRankSum. The thing is, I do have QD<10 SNPs in my VCF, as we can observe on the QD plot ("SNPs_4filters_QD.pdf"; attached).
To summarise, something is definitely going wrong with my FS>10 filter. Something is also going wrong with further filtering attempts based on QD, ReadPosRankSum and MQRankSum, and this seems independent of the first problem, since no matter which VCF I use to apply a QD filter, I get the same warning and output (ie. no filtering at all).
Any thoughts?
Ben