Hello,
I am working on a SNP data set (generated via STACKS de novo) and, after testing for linkage disequilibrium using the R package genetics, I have tried 2 different methods for outliers detection with the idea of selecting the loci detected by at least 2 of the softwares. I have used Bayesca, PCAdapt and OutFLANK.
Bayescan has not detected any outlier, PCAdapt detected 129 outliers and OutFLANK 44. However, the outliers from PCAdapt and OutFLANK do not overlap at all. Is this normal?
What could I be doing wrong?
Can anyone help me with this?
Thanks
Giulia
The problem is that you are trying to detect outliers using two methods that are not compatible. Outliers are not simply data anomalies but are different because different estimation methods are based on different data properties for example different multivariate methods show different outliers. I wouldn't expect factor analysis outliers to be the same as manova outliers using the same data. The short answer then is determine your outliers using a method that reflects the goals of your analysis best. Don't use different algorithms. There is an entire literature on this. To enter that literature do a Google search on multivariate methods outlier detection. Best wishes, David Booth
Dear Giulia Trauzzi
Basically, Bayescan did not detect any outlier loci. The number of outlier SNPs identified by Lositan ranged from 14 to 28 among runs. Akpatok showed the highest number of outliers when compared to Prince Leopold (28 SNPs) and Minarets (24 SNPs), respectively.
This paper might be resourceful, have a look
https://www.nature.com/articles/srep13758
Kind Regards
Qamar Ul Islam
See these links:
Article Outlier analyses to test for local adaptation to breeding gr...
https://www.researchgate.net/post/How-should-I-choose-outlier-in-Bayescans-results
- Badges
- Science topic