Towards projectionFISH, a Method for Efficient Spatial Transcriptomics

Abstract

The modern revolution in single cell transcriptomics has revealed unprecedentedly detailed characterizations of cell states, but not without removing the cells from their spatial contexts. FISH-based spatial transcriptomic methods aim to bridge this gap by simultaneously reading out a cell's position, morphology, and gene-expression profile, with the latest methods approaching a whole-transcriptome readout. These methods, while powerful, are difficult and expensive, precluding their widespread use. We are developing a complementary method that leverages (non-spatially resolved) single cell RNA-sequencing data to design FISH probes that can be used to efficiently estimate a cell's transcriptome. We hypothesized that the low-dimensional structure of transcriptome-space (due to gene regulatory networks and redundancy in expression) meant that a small number of carefully chosen FISH measurements each corresponding to integer-coefficient linear combinations of genes' expression levels would suffice to determine the transcriptome. Using pre-existing single cell RNAseq data to design gene combinations and to decode gene expression level from the expression levels of those combinations, we found that three FISH measurements should be sufficient to reconstruct a hematopoietic stem cell's gene expression protocol almost perfectly. Here we describe the methods used to choose these FISH measurements, the methods used to decode these FISH measurements, and preliminary experimental methods for FISH in hematopoietic progenitor cells, with the goal of implementing our proposed method projectionFISH in primary hematopoietic progenitors.