Hi all,
I am analyzing some small target (target region is around 260,000 bp) PCR-based DNA deep sequencing (~5000x) data. The data are generated by Illumina sequencer (maybe MiSeq if I am not wrong).
I did indel realignment and base quality re-calibration in pre-process steps (For details of pre-process, please see below). After BQSR, I found that for almost all reads, the BI tag (insertion quality, around phred score 90-33) is higher than BD tag (deletion quality, around phred score 80-33).
What's more, in the following variants calling step, we found much more deletions than insertions.
Is it normal behavior for the data having higher BI quality than BD quality?
Pre-process
1. BWA mapping (mem, 0.7.5a)
Per-lane data preprocess (GATK 3.5):
2. sort (Picard SortSam 1.138), remove secondary reads (samtools 1.3), AddOrReplaceReadGroups, we DID NOT mark duplicates since the data is PCR-based.
3. RealignerTargetCreator (-L, -dt NONE, -known 1000G_phase1.indels.b37.vcf, -known Mills_and_1000G_gold_standard.indels.b37.vcf), IndelRealigner (without -L, -maxReads 1000,000)
4. BaseRecalibrator (-L, -knownSites 1000G_phase1.indels.b37.vcf, -knownSites Mills_and_1000G_gold_standard.indels.b37.vcf, -knownSites dbsnp144), PrintReads(-baq RECALCULATE), samtools (1.3) calmd -Abr
Merged-lane data preprocess:
5. RealignerTargetCreator (-L, -dt NONE, -known 1000G_phase1.indels.b37.vcf, -known Mills_and_1000G_gold_standard.indels.b37.vcf), IndelRealigner (without -L, -maxReads 1000,000)
6. BaseRecalibrator (-L, -knownSites 1000G_phase1.indels.b37.vcf, -knownSites Mills_and_1000G_gold_standard.indels.b37.vcf, -knownSites dbsnp144), PrintReads(-baq RECALCULATE), samtools (1.3) calmd -Abr
AnalyzeCovariates plot
AnalyzeCovariates plot are attached here.
AHC333.L001.BQSR.pdf
AHC333.L002.BQSR.pdf
AHC333.merged.BQSR.pdf
AHC338.L001.BQSR.pdf
I randomly picked two samples (AHC333 and AHC338), for each of them, they are sequenced in 4 lanes (L001, L002, L003, L004). I attached two lanes' plots of AHC333 to check the inter-lane case and attached one lane plot of AHC338 to check inter-sample case. I also attached the plot for merged-lane plot of AHC333.
The plots are generated based on BaseRecalibrator output. (From my view, in merged data of AHC333, there is significant difference between mean insertion quality and mean deletion quality)
Thank you very much for your help.
bless~
Xingliang
I am analyzing some small target (target region is around 260,000 bp) PCR-based DNA deep sequencing (~5000x) data. The data are generated by Illumina sequencer (maybe MiSeq if I am not wrong).
I did indel realignment and base quality re-calibration in pre-process steps (For details of pre-process, please see below). After BQSR, I found that for almost all reads, the BI tag (insertion quality, around phred score 90-33) is higher than BD tag (deletion quality, around phred score 80-33).
What's more, in the following variants calling step, we found much more deletions than insertions.
Is it normal behavior for the data having higher BI quality than BD quality?
Pre-process
1. BWA mapping (mem, 0.7.5a)
Per-lane data preprocess (GATK 3.5):
2. sort (Picard SortSam 1.138), remove secondary reads (samtools 1.3), AddOrReplaceReadGroups, we DID NOT mark duplicates since the data is PCR-based.
3. RealignerTargetCreator (-L, -dt NONE, -known 1000G_phase1.indels.b37.vcf, -known Mills_and_1000G_gold_standard.indels.b37.vcf), IndelRealigner (without -L, -maxReads 1000,000)
4. BaseRecalibrator (-L, -knownSites 1000G_phase1.indels.b37.vcf, -knownSites Mills_and_1000G_gold_standard.indels.b37.vcf, -knownSites dbsnp144), PrintReads(-baq RECALCULATE), samtools (1.3) calmd -Abr
Merged-lane data preprocess:
5. RealignerTargetCreator (-L, -dt NONE, -known 1000G_phase1.indels.b37.vcf, -known Mills_and_1000G_gold_standard.indels.b37.vcf), IndelRealigner (without -L, -maxReads 1000,000)
6. BaseRecalibrator (-L, -knownSites 1000G_phase1.indels.b37.vcf, -knownSites Mills_and_1000G_gold_standard.indels.b37.vcf, -knownSites dbsnp144), PrintReads(-baq RECALCULATE), samtools (1.3) calmd -Abr
AnalyzeCovariates plot
AnalyzeCovariates plot are attached here.
AHC333.L001.BQSR.pdf
AHC333.L002.BQSR.pdf
AHC333.merged.BQSR.pdf
AHC338.L001.BQSR.pdf
I randomly picked two samples (AHC333 and AHC338), for each of them, they are sequenced in 4 lanes (L001, L002, L003, L004). I attached two lanes' plots of AHC333 to check the inter-lane case and attached one lane plot of AHC338 to check inter-sample case. I also attached the plot for merged-lane plot of AHC333.
The plots are generated based on BaseRecalibrator output. (From my view, in merged data of AHC333, there is significant difference between mean insertion quality and mean deletion quality)
Thank you very much for your help.
bless~
Xingliang