johnlees
commented
1 year ago In my opinion, initially, I attempted to conduct GWAS without adjusting for population structure, which led to inflated p-values. Subsequently, to address this issue, I performed GWAS with the similarity option. Although this approach may remove some lineage signals (not all population-specific unitigs may be detected), I was able to identify lineage-specific one genome region. However, it is possible that other specific genome regions still remain undetected. Is this understanding correct?
Yes this is right. You may find it helpful to read the following: https://figshare.com/articles/thesis/The_background_of_bacterial_GWAS/5550037 https://www.nature.com/articles/ncomms12797 https://www.nature.com/articles/nmicrobiol201641 https://www.nature.com/articles/nrg.2016.132
Hello, thank you for your advice in #238 .
In my case, the phenotype was population defined by population structure analysis(like lineageA vs lineageB). In phenotype file, binary (0,1) values were used(lineageA samples->1, others->0).
As mentioned #238 , at first, p-values were inflated with --no-distances option.
Then I adjusted for population structure like above, were the lineage signals removed after adjusting for population structure? As a result of this analysis, one of interested gene was detected(this gene include a lot of significant unitigs).
Or is there other way to analyze?
In my opinion, initially, I attempted to conduct GWAS without adjusting for population structure, which led to inflated p-values. Subsequently, to address this issue, I performed GWAS with the similarity option. Although this approach may remove some lineage signals (not all population-specific unitigs may be detected), I was able to identify lineage-specific one genome region. However, it is possible that other specific genome regions still remain undetected. Is this understanding correct?
Thank you in advance.