Goldfinch Bio is an exciting startup working on precision therapies for chronic kidney disease — a condition that affects 26 million people but hasn’t had a major advance in treatment in over 20 years. I recently had the pleasure of working with Goldfinch to help characterize its high-throughput organoid platform, which it developed for preclinical validation of potential therapies for kidney disease. The Goldfinch team had already performed single-cell RNA sequencing (scRNAseq) on organoids from eight different time points in vitro and six other time points after transplantation in rats. This produced a massive dataset on >300,000 cells from 42 different organoids. I worked with the team at Goldfinch to analyze this data, with a particular focus on understanding the maturation state of the organoids and the reproducibility of the model.
scRNAseq is well-suited to studying heterogeneous populations of cells such as those from organoids. We used the expression of cell-type marker genes to identify more than a dozen cell types in the organoids without prior sorting. Having identified the cells, we were then able to compare the proportions of cell types present in the organoids to assess their maturation as well as variability between replicates. We found that the organoids differed significantly between time points but showed little variation between replicates. We also observed that organoids that had been transplanted in rats had a greater proportion of mature cell types than those cultured in vitro. Finally, we compared the transcriptional profiles of two types of podocytes (epithelial cells that help the kidneys filter blood) — cells from each organoid, and cells from adult and fetal kidney samples. We found that organoids cultured in vivo were closely correlated with the tissue samples.
All told, our findings demonstrated that transplantation of organoids in rats consistently promoted organoid maturation and vascularization. As a computational biologist, I also gained some field-specific insights from working with the Goldfinch team. Although I have analyzed scRNAseq data from a variety of tissues, the complexity of the kidney organoids was unlike any other dataset I have worked with — this was a large dataset with numerous samples and many different cell types. Our ability to identify and characterize so many distinct cell populations within these samples speaks to the inherent strength of scRNAseq in deconvoluting complex populations.
I would like to thank the Goldfinch Bio team for the opportunity to work on this project — in particular, Amy Westerling-Bui, Tom Soare, Eva Fast, and Srinivasan Venkatachalan, for plenty of great discussions and teamwork. This work was recently published on biorxiv.org.