#ASHG21 AN: Aparna Nathan.
Modeling eQTLs at single-cell resolution identifies 2000 eQTLs with differential effects across continuous T cell states.
Modeling eQTLs at single-cell resolution identifies 2000 eQTLs with differential effects across continuous T cell states.
#ASHG21 AN: cell-state dependent eQTLs means that the same genetic variant may have an effect or no effect depending on cell type and activation. How do you find state-dependent QTLs?
#ASHG21 AN: Think about cell states as continuum rather than a fixed, binarized cell type. Use a poisson mixed effect model in individual cells. Thought to be more robust of single cell modeling than things like linear mixed models of normalized counts.
#ASHG21 AN: Random effects by donor and batch. Shown in their preprint. Modeled in a data from ~500k T cells (Nathan 2021 Nat Imm).
#ASHG21 AN: Use canonical correlation analysis (CCA). Cell given a continuous score across 7 dimensions from the 7 CVs they found. That's how you test for state dependence.
#ASHG21 AN: ~6500 memory t cell eQTLs from pseudo-bulk analysis. 2/3rds had a cell-state dependent analysis.
#ASHG21 AN: Can have opposing state effects at the same locus, i.e. degree of cytotoxicity function can have opposite effects for individual variants.
#ASHG21 AN: The eQTL effect in each cell is a sum of individual CV effects. Can calculate a sum of the betas to get the overall effects.
#ASHG21 AN: Find some new eQTLs that aren't found in pseudo-bulk analysis. Particularly apparent for rare cell types.
#ASHG21 AN: Where are these effects? Enriched in promoters. The non-promoter t-cell regions you see cell-state interacting eQTLs over non state interacting.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
