Quality check on RNA-seq data from aging.brain-map.org

I am comparing the normalised and unnormalised fpkm values between people with Alzheimer’s/Dementia diagnosis and those with no Dementia. However, the difference in fpkm expression of APOE was not found to be significant between the two groups for any brain region, which I found odd as APOE expression is strongly correlated with Alzheimer’s. Does anyone know why this discrepancy is there and is the data still valid for Alzheimer’s research?


Hi @Chris. We spent a lot of time trying to understand how our results match with prior knowledge about Alzheimer’s disease. Our results are published in eLife, but I’ll summarize the relevant parts here. We find that these RNA-seq data are generally consistent with previous studies, but noisier. This noise is likely a combination of biological variability (e.g., old donors have a lot of pathology, regardless of cognitive status) and technical variability (e.g., this RNA-seq method turned out to be quite sensitive to RNA quality), and the end result was that very few genes showed significant associated with disease. We also find a relationship between RNA quality and dementia status, so when we normalized the RNA-seq data to remove variation due to RNA quality, this also removes some dementia-associated variation. Overall, these data are still valid for Alzheimer’s disease research, especially in combination with the extensive donor and pathology profiling, but I would use caution with making strong conclusions without validation from other sources.

1 Like