Fig. 1From: A Poisson reduced-rank regression model for association mapping in sequencing dataIllustration of PRRR. We fit PRRR to a toy dataset containing two cell-specific covariates and two genes. The two covariates showed negative correlation, and the two genes were jointly associated with the covariates (panel A). PRRR identifies the low-rank structure of these multivariate relationships by decomposing the full coefficient matrix into two low-rank matrices, U and V (panels B and C)Back to article page