Final Cohort Filtering

[hollygene]

H0 Final Cohort Variant Info: hcm14449@n204 H0$ wc -l H0_FullCohort.vcf 1729 H0_FullCohort.vcf hcm14449@n204 H0$ pwd /scratch/hcm14449/TE_MA_Paradoxus/Illumina_Data/Out/H0

1729 raw variants

Remove sites with depth less than 90: hcm14449@n204 H0$ wc -l H0_noLow.vcf 1080 H0_noLow.vcf

Remove sites with depth greater than 218 hcm14449@n204 H0$ wc -l H0_noLow_noHigh.vcf 889 H0_noLow_noHigh.vcf

Remove hard-to-map regions called by GEM: hcm14449@n204 H0$ wc -l H0_noLow_noHigh_redGem.vcf 668 H0_noLow_noHigh_redGem.vcf

Removed sites with no calls in the ancestor: hcm14449@n204 H0$ wc -l H0_noLow_noHigh_redGem_AncCalls.vcf 654 H0_noLow_noHigh_redGem_AncCalls.vcf

Removed sites with hets in the ancestor hcm14449@n204 H0$ wc -l H0_noLow_noHigh_redGem_AncCalls_NoHets.vcf 545 H0_noLow_noHigh_redGem_AncCalls_NoHets.vcf

SNPs:

hcm14449@n204 H0$ wc -l H0_noLow_noHigh_redGem_AncCalls_NoHets_SNPs.vcf
344 H0_noLow_noHigh_redGem_AncCalls_NoHets_SNPs.vcf

Indels: hcm14449@n204 H0$ wc -l H0_noLow_noHigh_redGem_AncCalls_NoHets_Indels.vcf 223 H0_noLow_noHigh_redGem_AncCalls_NoHets_Indels.vcf

_Originally posted by @hollygene in https://github.com/hollygene/TE_MA/issues/2#issuecomment-612691832_

[hollygene]

Updates from 3/19 to present: (Had this in a Word doc, but will update on here from now on)

3/19/20

• Using vcfR to look at quality control and parse VCFs
• D0: read depth by sample violinD0.pdf heatmapD0.pdf

3/26/20

• Looking at various filtering parameters, to determine what filtering criteria I should include in my analysis
• Guidance from https://gatk.broadinstitute.org/hc/en-us/articles/360035890471-Hard-filtering-germline-short-variants o GATK suggests using the following guidelines (with guidance from your own dataset based on the various plots you generate from the annotations in the vcf file
• My annotation distributions, from D0 dataset:
• Guidelines in Otto paper were:
• Sites were excluded from further analysis if >10% of samples had no data or if the average depth across samples was >2 x the average median depth for that chromosome. Variant calls for a sample were also excluded if sample depth was <4 in haploids, <8 in diploids, if sample depth was >2 x the median depth for that chromosome and sample, if map quality was <50, if strand bias was high (P < 0.01), or if a putative variant was found in >1 reads in other samples.
• VCF files that I have don’t have a P-value for strand bias, but GATK recommends removing sites with SOR greater than 60
• Need to ask Dave about what cutoff to use for sample depth

3/27/20

• Filtered D0 cohort fully using two methods to make sure they gave me the same result: o Remove low-mappability sites first, then filter based on my criteria o Filter based on my criteria first, then remove low-mappability sites o Both gave me the exact same results

File Length D0_FullCohort.vcf 1595 Raw variants D0_reducedGEM.vcf 1042 Removed low-mappabiltiy sites D0_reducedGEM_DpGr10_Fil.vcf 827 Removed sites with depth <10 D0_reducedGEM_DpGr10_Fil_AncCalls.vcf 811 Removed sites with ancestor no-calls (./.) D0_reducedGEM_DpGr10_Fil_AncCalls_NoHets.vcf 664 Removed sites with ancestor genotype as Heterozygous

Alternative method: D0_FullCohort.vcf 1595 Raw variants D0_FullCohort_DpGr10_Fil.vcf 1128 Remove sites with depth < 10 D0_FullCohort_DpGr10_Fil_AncCalls.vcf 1094 Remove sites with ancestor no-calls (./.) D0_FullCohort_DpGr10_Fil_AncCalls_NoHets.vcf 854 Remove sites with ancestor genotype as heterozygous D0_FullCohort_DpGr10_Fil_AncCalls_NoHets_GEM.vcf 664 Removed low-mappability sites

Checked to make sure they both gave the same thing – they did (exact same results)

Using the First method, because it seems to make more sense (remove low-mapping sites first instead of at the end)

4/1/20

• Plotting depth distribution of D0 ancestor
• Median depth looks ~140
• Mean depth is 179, but this causes the depth cutoffs to be very stringent and so ~1000 variants were removed based on low depth alone
• Going to use median depth to calculate depth cutoffs instead of mean as there are several high depth sites that are skewing the mean

Chromosome I depth distribution (x axis is location on chr I, y axis is depth)

Chromosome II:

Chromosome III:

Chromosome IV:

Chromosome V:

Chromosome VI:

Chromosome VII:

Chromosome VIII:

Chromosome IX:

Chromosome X:

Chromosome XI:

Chromosome XII:

Chromosome XIII:

Chromosome XIV:

Chromosome XV:

Chromosome XVI:

Filtering of Variants D0 Low depth: 82 High depth: 206

File Length D0_FullCohort.vcf 1595 D0_noLow.vcf 962 D0_noLow_noHigh.vcf 772 D0_noLow_noHigh_redGEM.vcf 580 D0_noLow_noHigh_redGEM_AncCalls.vcf 573 D0_noLow_noHigh_redGEM_AncCalls_NoHets.vcf 471

D1 Low depth: 70 High depth: 188

File Length D1_FullCohort.vcf 2695 D1_noLow.vcf 1586 D1_noLow_noHigh.vcf 1346 D1_noLow_noHigh_redGEM.vcf 971 D1_noLow_noHigh_redGEM_AncCalls.vcf 958 D1_noLow_noHigh_redGEM_AncCalls_NoHets.vcf 812

D20 Low depth: 82 High depth: 206

File Length D20_FullCohort.vcf 1835 D20_noLow.vcf 1251 D20_noLow_noHigh.vcf 963 D20_noLow_noHigh_redGEM.vcf 741 D20_noLow_noHigh_redGEM_AncCalls.vcf 721 D20_noLow_noHigh_redGEM_AncCalls_NoHets.vcf 526

Found variants in the ancestors:

D20 ancestor variants

Chr Pos GT Type VII 61782 1/1 ALT – indel XVI 186933 1/1 ALT – indel

D1 ancestor variants

Chr Pos GT Type IV 641630 2/2 ALT – indel (het in D20 & majority of progeny) VII 61782 1/1 ALT – indel VII 986079 1/1 ALT – SNP (not in D20) XV 299062 1/1 ALT – SNP (not in D20) XVI 186933 2/2 ALT – SNP

D0 ancestor variants

Chr Pos GT Type IV 641630 2/2 ALT – indel VII 61782 1/1 ALT – indel XIII 111478 ½ ALT – indel XV 314234 ½ ALT – indel

4/3/20 Heterozygous sites shared between the 3 diploid ancestors: 40 total File: /Users/hollymcqueary/Dropbox/McQueary/Paradoxus_MA/VCFs/AncHets/D0_D1_D20_common.txt

4/7/20 H0 median depth = 153

Filtering based on depth of 150 though, so removing anything less than 90 and greater than 218