imputation only choose one population as reference panel

Question

[Peter Chung]

I have a gwas data to do the imputation by shapeit2 and impute2. I would like to use only east asians population in 1,000 Genomes haplotypes -- Phase 3 integrated variant set release in NCBI build 37 (hg19) coordinates from impute2 reference data:

https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html

impute2 -filt_rules_l 'EUR!="0"' 'EUR=="0"' 'AMR=="0"' 'AMR!="0"' 'SAS=="0"' 'SAS!="0"' 'AFR!="0"' 'AFR=="0"' -known_haps_g gwas.phased.haps -m genetic_map_chr22_combined_b37.txt -h 1000GP_Phase3_chr22.hap -l 1000GP_Phase3_chr22.legend -int ${startchr} ${endchr} -Ne 20000 -o chr22test.impute2

I use filt_rules_l options to remove any other population, however, it doesn't work.

any suggestions to filter out the population ? thank you very much.

[ShirleyDai]

Hi Peter, Please check this website. https://mathgen.stats.ox.ac.uk/impute/using_multi_population_reference_panels.html I have also been confused about this for a long time. As I understand, before 1000Gp1v3, it is suggested to use "best match" population as reference. However, nowadays, for "modern imputation", IMPUTE2 can provide all available reference haplotypes then let the software choose a "custom" reference panel for each individual to be imputed. Then I also found some papers used this mixed population as reference. You can check the number of reference population in those papers.

Shirley

[zx8754]

I think you need to drop double quotes - impute2 -filt_rules_l 'EUR!=0' ...

Also, from the manual, it says "Filtering rules are always applied in 'OR' fashion".

I would rather subset the panel using gtools, then use that subset as panel.

[liuyibin2005]

Hi，there! How about the problem? I am doing the same thing and I don't know how to filter the population off those non-Asian ... Can I have your e-mail and have a chat ?