= GWAS QC pipeline =

Removed all bad samples (taken from Mathieu).

{{{
p-link --bfile GvNL_250111 --remove GvNL_bad_samples.txt --make-bed --out GvNL_good_samples}}}

Update family information

{{{
p-link --bfile GvNL_samples --update-ids GvNL_family_update.txt  --make-bed --out GvNL_good_fam}}}

Set trio relationships

{{{
p-link --bfile GvNL_good_fam --update-parents GvNL_parents_update.txt  --make-bed --out GvNL_good_fam_par}}}

Manual check for suspicious individuals (not in trios) and remove them

{{{
p-link --bfile GvNL_good_fam_par --remove GvNL_suspicious.txt  --make-bed --out GvNL_good_fam_par_susp}}}

NB: After checking suspicious individuals with Mathieu, turns out it was 1 typo and 1 filtered low call rate (93%). Therefore, only the low callrate individual was screened out after all.

Check and update sex information

{{{
p-link --bfile GvNL_good_fam_par_susp --check-sex --make-bed}}}

Crosscheck with information from Genome Studio provided by Mathieu
* 100% match between plink and Genome Studio when sex information is within plink threshold
* 2 pairs of parents swapped
* 1 individual tagged as male instead of female by both Genome Studio and plink. Tagged as suspicious in fail-sexcheck-qc.txt

Update sex information for children and swapped individuals

{{{
p-link --bfile GvNL_good_fam_par_susp --update-sex GvNL_sex_update.txt --make-bed --out GvNL_raw_final}}}

Identification of individuals with elevated missing data rates or outlying heterozygosity rate (See Anderson, NP2010, p.1569) => How does this compare with the PLINK suggested approach  [[http://pngu.mgh.harvard.edu/~purcell/plink/thresh.shtml here]]

Get missing data information:

{{{
p-link --bfile GvNL_raw_final --missing --out GvNL_raw_final}}}
 
Get heterozygocity information: 

{{{
p-link --bfile GvNL_raw_final --het --out GvNL_raw_final}}}

Calculate the observed heterozygosity rate per individual using the formula (N(NM)  −  O(Hom))/N(NM) and plot the missing SNPs vs heterozygocity rate for eyeball inspection. 
* In R: 

{{{
het <- read.table("GvNL_raw_final.het", head=TRUE)
mis <- read.table("GvNL_raw_final.imiss", head=TRUE);
het$HET_RATE = (het$"N.NM." - het$"O.HOM.")/het$"N.NM."&#10;plot (mis$F_MISS,het$HET_RATE, xlab="Missing SNPs", ylab="Heterozygocity Rate")}}}

No samples with callrate < 95% (Filtered in lab)

Put samples where heterozygocity rate is outside 3 standard dev from the mean (Anderson, NP2010, p.1569) in fail-het-3sd-qc.txt, along with their deviation from the mean

In R: 

{{{
het_fail = subset(het, (het$HET_RATE < mean(het$HET_RATE)-3*sd(het$HET_RATE)) | (het$HET_RATE > mean(het$HET_RATE)+3*sd(het$HET_RATE)));
het_fail$HET_DST = (het_fail$HET_RATE-mean(het$HET_RATE))/sd(het$HET_RATE);
write.table(het_fail, "fail-het-qc.txt", row.names=FALSE)}}}

Result:
* A total of 5 individuals failed. Worst failure has a distance of -3.74sd from the mean. 

Check inheritance and duplicates
* Prune SNPs for LD

{{{
p-link --bfile GvNL_raw_final --indep-pairwise 50 5 0.2 --out GvNL_raw_final}}}

* Generate pairwise Identity-By-State (IBS) metrics using pruned SNPs only

{{{
p-link --bfile GvNL_raw_final --extract GvNL_raw_final.prune.in --genome --out GvNL_raw_final}}}

* Check trio inheritance based on IBS: Check the Pi_HAT value for the individuals (unrelated individuals =~ 0, parent <-> child =~ 0.5, siblings =~0.5, twins =~ 1.0). Done with homemade perl script.
* 2 Families are not real families (G34c (child) is wrong and A56b (mother) is wrong)
* Duplicated family: R24 == R25

Check trio inheritance based on Mendelian segregation

{{{
p-link --file GvNL_raw_final --me 0.05 0.1 --make-bed --out GvNL_inheritance}}}

A diff between GvNL_raw_final.fam and GvNL_inheritance.fam confirms IBS version: families G34 and A56 have been removed.