Estimate relatedness in large dataset

Question

[dk]

Hi,

I am not very familiar with these types of analyses, but I am wondering if there is a (semi straight-forward) way to figure out if in a group of 500 individuals any of the individuals are related to eachother (using either SNP array or NGS data).

[GenoMax]

Are you looking for relatedness (as in familial relationship) or purely for sample identification. In fact with some large projects people may actually do independent SNP arrays to make sure there are no sample mix-ups in processing of samples (or after the fact if there is a question about provenance of some samples).

Answer 1

[chrchang523]

With plink 2.0, "plink2 --vcf [VCF filename] --make-king-table --king-table-filter 0.177" is very good at detecting first-degree relationships (use a lower --king-table-filter cutoff if you also want to estimate more distant relationships). If you want to follow that up with more sophisticated relationship inference, you may want to look at KING (http://people.virginia.edu/~wc9c/KING/ ).

If your input data is in another format, odds are good that it's either readable by plink2 or can be easily converted to be, but I can't give you specific instructions without more information.

Answer 2

[leeandroid]

VCFtools is probably what you're looking for. However you should provide more information in order to get a better answer.

Answer 3

[ashotmarg2004]

Worth checking out e.g., KING: https://www.kingrelatedness.com/manual.shtml; somalier: https://github.com/brentp/somalier; ngsrelate2: https://github.com/ANGSD/NgsRelate among others.