Deconvolution of gene expression from cell populations across the C. elegans lineage

Knowledge of when and in which cells each gene is expressed across multicellular organisms is critical in understanding both gene function and regulation of cell type diversity. However, methods for measuring expression typically involve a trade-off between imaging-based methods, which give the precise location of a limited number of genes, and higher throughput methods such as RNA-seq, which include all genes, but are more limited in their resolution to apply to many tissues. We propose an intermediate method, which estimates expression in individual cells, based on high-throughput measurements of expression from multiple overlapping groups of cells. This approach has particular benefits in organisms such as C. elegans where invariant developmental patterns make it possible to define these overlapping populations of cells at single-cell resolution.

We implement several methods to deconvolve the gene expression in individual cells from population-level data and determine the accuracy of these estimates on simulated data from the C. elegans embryo.

These simulations suggest that a high-resolution map of expression in the C. elegans embryo may be possible with expression data from as few as 30 cell populations.