Thread by @thatdnaguy

#ASHG21 BJS: Benjamin J Strober.
Uncovering context-specific genetic regulation of gene expression from single-cell RNA-sequencing using latent-factor models.

#ASHG21 BJS: Many contexts modulate gene expression. Such as cell type, developmental stage, environmental stimulus.
One way to study is to look at eQTLs. Especially context-specific eQTLs. Example might be an eQTL in B cells that isn't an eQTL in T cells.

#ASHG21 BJS: Developed SURGE method. Single cell Unsupervised Regulation of Gene Expression. Want to find context-specific eQTLs without guidance. Mainly intended for single-cell eQTL data since you don't have to specify context.

#ASHG21 BJS: Interaction eQTL calling between genotype and context in a number of samples. Context required.

#ASHG21 BJS: SURGE things of context as an unobserved latent variable and try to find the most likely value for this unobserved variable.

#ASHG21 BJS: Instead of one latent context assume there are K latent context. [ if you want the equations ask someone else. I think algorithmically and not letters in matrix formulas ]

#ASHG21 BJS: SURGE applied to GTEx (n=10) tissues. Identifies 7 latent contexts. Examining the contexts looks like they are capturing particular cell types in the tissues.

#ASHG21 BJS: Some latent contexts cluster samples by ancestry rather than tissue [ hmm ]

#ASHG21 BJS: Can apply UMAP to the latent space. What are the contexts it identifies other than cell type? One example maturation gradient in T cells.

#ASHG21 BJS: SURGE identifies more trait colocalizations than the standard eQTL method.

#ASHG21 BJS: Allows for unsupervised exploration of the single-cell data. Increases ability to detect context-specific eQTLs.