We then examined how variation in the proportions of basic cell types contributes to variation in bulk expression. To this end, we estimated the relative proportions of various cell types (“cell fractions”) for each tissue sample. In particular, we deconvolved the bulk tissue-level expression matrix by using the single-cell signatures to estimate cell fractions across individuals (matrix W), solving B ≈ CW (Fig. 2B) (21). As a validation, our estimated fractions of NEU+/− cells matched the experimentally determined fractions from reference brain samples (median difference = 0.04) (fig. S10). Overall, our analyses demonstrated that variation in cell types contributed substantially to bulk variation. That is, weighted combinations of single-cell signatures could account for most of the population-level expression variation, with an accuracy of >88% (Fig. 2D) (1 − ||B − CW||2/||B||2 > 88%), and when calculated on a per-person basis, this quantity varies ±4% over the 1866 individuals in our cohort (figs. S11 and S12). Also, our results explained more variation than previous deconvolution approaches (fig. S13) (21).
