Read on Twitter
Another paper illustrating the limitations of the TWAS method which I think should lead to greater caution in using TWAS results to assign causality in genetic studies
https://www.biorxiv.org/content/10.1101/2020.08.31.273458v2
@MarylynRitchie https://twitter.com/biorxivpreprint/status/1300867559701520388
https://www.biorxiv.org/content/10.1101/2020.08.31.273458v2
@MarylynRitchie https://twitter.com/biorxivpreprint/status/1300867559701520388
As this paper demonstrates, TWAS has a very high false positive rate, under a variety of scenarios:
The strongest metabolite hit ever reported is for UGT1A1 and bilirubin (close to 1e-20000 in https://www.biorxiv.org/content/10.1101/660506v1.)
In this TWAS paper, all the genes in the region are tagged, even genes with no relation to bilirubin.
In this TWAS paper, all the genes in the region are tagged, even genes with no relation to bilirubin.
As they note in the TWAS paper this probably reflects a shared regulatory mechanism.
Not stated in this paper is this is likely the source of the high false positive rate in TWAS and other eQTL-related analyses.
Not stated in this paper is this is likely the source of the high false positive rate in TWAS and other eQTL-related analyses.
The authors miss an opportunity to highlight another example of this. They find only 2 "hits" for HDL levels. One is the well-known CETP gene.
Out of the entire genome the only other hit is...
The gene right next door to CETP:
Out of the entire genome the only other hit is...
The gene right next door to CETP:
Given all this I think the authors' final conclusion is completely unjustified.
I think the paper illustrates the pleiotropy of gene expression but I don't think it reflects any novel biology of HIV infection.
I think the paper illustrates the pleiotropy of gene expression but I don't think it reflects any novel biology of HIV infection.
