sinanshi / ge Goto Github PK

Are we sure of this? It was not the case when we did ASHG slide.
Double check.

I applied switch error rate checking on all ~7000 trios in ~/meta/trios_for_bcftools_mendelian.ped in ~/data/phasing_test/shapeit4_full_all_snp_wrong/

1. directely on gwas.phased.vcf.gz
2. only 100 trios child_phased.vcf.gz

Issues:

• only 3625 trios are calculated, what are the missing trios? why?
• I am expecting the same number of testing sites, switch rate, and Medelian errors for the 100 trios comparing both data 1) and data 2). But it turned not to be the case. Why?? -- I need to check the sites to see where it differs first.
• switch error rate in 1) ~1.5% 2) 0.5%

This is potentially problematic. In the current version of merged data, we removed all multi-allelic data.

Did we keep the one of the allele of the multi-allelic site?

No, I can confirm that all multi-allelic sites are removed in the latest merged file /data/pipeline2018/run_0001/merge/chr20/merged_m2M2_snps_nohemi_minac1.vcf.gz

1. sanity check:

• tabulate call parents child genotype combination, fit all the flips into all the valid entries. The point being not to have abnormality in any of the entries.

• flips against AF

• Simon said: probability to flips against frequency in panel.

• Check the gelsnp dataset, because the problem can be caused by undersized chunk size. Too few het sites.

• The number is based on the data available on 2018-10-07, rare disease, cancer germline cohort, qc_passed, genome_build!=NA.
• Combining with 38874 build 38 files, it will make a release of 49166 samples.

Shorter chromosomes may need larger chunk size. We have to make sure that each chunk is sufficiently large.

• flip rate vs MAF, the purpose of it is to see how the flips are affected by singleton, doubleton, and so on.
• remove singletons.

There is a large number of VCF outputs are empty - size: 203Kb.
My first guess is there are some error occured.
Second guess is the chr22 instead of 22 issue.

• pick up an empty file and redo the subsetting.
• make a script for chr22 style files.

We observed high Mendel error and low switch error rate at trio: LP3000134-dna_c04.
Why?

• sample LP3000448-DNA_A09 has been missed.

I obtained some flip rate (number of flips/number of switches) above 50%
Can it happen? Need to investigate.