Read on Twitter
Our manuscript on MedRvix describes a large GWAS in Alzheimer: 33 new loci and 42 in a next update https://www.medrxiv.org/content/10.1101/2020.10.01.20200659v1.
This week, Wightman et al. described that they called the largest AD GWAS but only reported 7 new loci.
Why our GWAS discovered so many genes? A thread
This week, Wightman et al. described that they called the largest AD GWAS but only reported 7 new loci.
Why our GWAS discovered so many genes? A thread
Following our first IGAP publication (2013) and the use of the UK biobank (2018), most of the GWAS meta-analyses shared the same main GWAS datasets, making these studies not independent. In addition, the number of controls inflated but not the number of cases.
However, at the statistical level, it becomes useless to have only more and more controls. We analyzed almost 30,000 fully new AD cases (discovery/replication) whereas Wightman et al. included mainly only new controls through 23&me and FinnGen.
In addition methodologies are different between the studies. For instance, we developed a discovery/replication design conversely to Schwartzentruber and Wightman. We used the TopMed imputation panel allowing us to double the number of SNPs analysed with a high imputation quality
The new loci described in the De rojas, Schwartzentruber and Wightman papers, are not in common (whereas sharing main GWAS datasets). A large part of them are detected in the Bellenguez paper. This may indicate lack of statistical power and variability due to different designs.
Wightman et al. included the 23&me dataset. This approach was powerful in Parkinson helping to report dozen of new loci. However, this still needs to be evaluated in Alzheimer. The diagnosis is declarative and no demographic information seems to be reported in the current version
Take home message:
Clinically-diagnosed Cases add power … and potentially uncharacterized controls add noise?
The difference in the number cases defines the potential for novel discoveries.
Clinically-diagnosed Cases add power … and potentially uncharacterized controls add noise?
The difference in the number cases defines the potential for novel discoveries.
