Sample quality control

In this notebook, we perform quality control on samples, removing samples with very low depth or elevated heterozygosity.

Coverage data

How many samples fall below the threshold for total reads?

Removing 70 samples due to low total depth

Total reads per target SNP

Which target SNPs have lower total depth than the amplicon threshold?

Removing 2 target SNPs due to low total depth

	mutation
56	DSX1
57	DSX2

Number of missing calls

48 samples have more than 60 missing calls overall out of all possible target SNPs
48/48 of these are also present in the low depth samples to be excluded

Autosome / Sex chromosome coverage ratios (ag-vampir only)

Females will have a lower ratio of autosomes:x, and males will have a higher ratio. Its not clear whether we can use this yet to sex samples.

Sample heterozygosity

Locate heterozygosity outliers

We then find samples within each cohort which have a heterozygosity (2.5 * IQR) higher than the 75% quantile, to exclude samples with very high heterozygosity for their cohort.

For Obuasi the heterozygosity threshold is 0.145, out of 96 samples, 1 are outliers
For Gambia_URR the heterozygosity threshold is 0.125, out of 96 samples, 5 are outliers
For VK7 the heterozygosity threshold is 0.118, out of 96 samples, 1 are outliers
For Siaya the heterozygosity threshold is 0.375, out of 264 samples, 0 are outliers

Removing 7 samples in total due to high heterozygosity

Preliminary PCA - remove outliers

removing any invariant and highly missing sites

Obuasi - Found 2 PCA outliers in 96 samples using Z-scores
removing any invariant and highly missing sites

Gambia_URR - Found 2 PCA outliers in 96 samples using Z-scores
removing any invariant and highly missing sites
VK7 - Found 2 PCA outliers in 96 samples using Z-scores
removing any invariant and highly missing sites

Siaya - Found 11 PCA outliers in 264 samples using Z-scores

Summary of samples to exclude

	location	count
0	Gambia_URR	29
1	Obuasi	22
2	VK7	22
3	Siaya	12
4	total	85

Sample QC complete!

A new metadata file with low-quality samples removed has been written to results/config/ :)