Estimation of Recent Shared Ancestry Using ERSA

Question

[Caragh]

Hi there,

I am trying to estimate recent shared ancestry using ERSA2.1. However, when I run the below command, I get "no_sig_rel" for all individuals.

ersa --segment_files=mydata.match --control_files=ceu.match --output_file=mydata.out --confidence-level 0.8

From PiHAT calculations there does appear to be related individuals in the dataset, with PiHAT scores as high as 0.5.

As a sanity check I went on to run the example in the ERSA read me file (see command below).

 ersa --segment_files=data/family.match --control_files=data/ceu.match --pair_file=data/pairs.txt --output_file=example.out --confidence-level 0.8

And I received the same "no_sig_rel" for all but 1 pair of individuals.

Has anyone else had experience running this program? Am I leaving something out in the command? Any help would be greatly appreciated.

Many thanks,

Caragh

[WouterDeCoster]

ersa2.0 worked for me (well, a bit), but it seems to have different synthax...

[Caragh]

Would you be able to shared your syntax? I managed to get ersa1.0 working using their python scripts but the results are still giving me "no_sig_rel" for most of the individuals, despite their PiHAT stating otherwise.

[WouterDeCoster]

I'll have a look at my code and share it later today. However I'm not 100% sure I used it correctly (in the fifth degree everyone turned out to be related which I'm very sure of is not the case). How many samples do you have?

[WouterDeCoster]

I was using ersa for ~200 genomes. Essentially:

1. I removed all variants which weren't sequenced for all individuals and also excluded all indels (those gave some problems...). I had to split my variant file per chromosome because I had some problems in phasing "across" chromosomes. Major headache to find out what was going on.
2. I converted my variants to plink format
3. Using germline and beagle I performed phasing with the bash script "phasing pipeline" which can be found here: http://www.cs.columbia.edu/~gusev/germline/
4. I ran python ersa chr*.match (essentially ersa on all .match files generated)

I'm currently not planning to try ersa further but would be happy to hear your feedback.