scRNA-seq: oh, the joys – Nature Methods

Especially because of plummeting sequencing costs, the overall costs of scRNA-seq experiments have dropped, says Trapnell. And whereas what used to be a typical experiment involved a few hundred cells from tissue from which cells could be readily disassociated, these days, experiments can be run with millions of cells from many specimens. Most early applications focused on regulation of individual genes. Such work continues, but scRNA-seq scale-up and throughput have opened up new experimental possibilities.

Instead of perturbing cells and asking “how does my favorite gene change?” or how individual genes are regulated, one can perturb model organism embryos and ask how a favorite cell type changes in proportion to other cell types. One gets a sense of the overall program because one can study how cell types depend on one another. One might be studying cancer cells to study residual disease. “It’s like using single-cell RNA-seq like one would use flow cytometry,” he says. “But on a much larger scale.” These changes make it easier to design experiments.

The Trapnell and the Shendure lab have been applying scRNA-seq to developmental biology questions10,11. This work will be scaled up in their work at SeaHub, the new Seattle Hub for Synthetic Biology. Shendure is SeaHub’s scientific director and Trapnell will co-lead.

When you can sequence millions of cells from different specimens — in their cases, embryos of model organisms that have been perturbed in various ways — one can study how a change affects all cell types across the development of the embryo, says Trapnell, and begin addressing problems in genetics and developmental genetics one could not address with more conventional tools. Along with new tools to address computational and statistical problems, such as for inferring which genes are required for which cell types, how the cell types depend on one another, or how the genes regulate one another, “I think it’s going to provide us with a means of dissecting the genetic program that controls development.”

Plenty of technical issues remain to be solved in scRNA-seq measurement and analysis, says Hao. “We need the scRNA data at the population level with genetic and curated clinical information,” he says. Last year, the Chan Zuckerberg Initiative consolidated publicly available scRNA-seq data to build CZ CellxGene Discover Census, with which one can access, query and analyze scRNA-seq data. These data are invaluable for training AI models to learn the unified representation of all those cells. It would be useful, he says, to have data about the donors of these cells while also maintaining privacy.

Much exciting work is ongoing with scRNA-seq, says Hansen. Sequencing assays are tough to validate since measuring cells destroys them. He is glad to see methods that record the history of the cell before measurement, such as Phylotime, a retrospective lineage barcoding and analysis tool developed by his Hopkins colleagues Reza Kalhor and Hongkai Ji and their groups.

It’s particularly useful to see the increase in spatial resolution that technologies are bringing, say Teichmann and her colleagues. This, along with new tools, will let scientists precisely map gene expression to individual cells at their exact location. For instance, understanding the exact cellular interactions between immune cells and their targets in cancer and autoimmune diseases holds much promise for treatment and drug discovery.

The research community, says Hao, has used scRNA-seq to identify and describe novel, rare and previously overlooked cell types. This opens a way to understand cell types and gene programs that lead to complex diseases and thus power therapeutics development. But one needs scRNA data not just from from tens to hundreds of individuals but thousands of individuals, says Hao. Lowering costs of scRNA increases its accessibility, and among the next technical challenges is determining how to collect information about the individuals who donated their cells and maintain privacy.

The next frontier, in Satija’s view, is to move beyond observation and leverage these technologies “to understand not just what cells are doing, but why they are doing it.” This is a new direction his lab is taking. One technique from this new direction is the team’s Phospho-seq, to simultaneously profile proteins, quantify intracellular protein dynamics, use scATAC-seq in whole cells, and then integrate these data with scRNA-seq datasets using the bridge integration method.

One can track cell signaling during development and reconstruct gene-regulatory relationships this way. The lab has also begun large-scale experiments12 to identify the regulators and targets of diverse cellular responses. This work involves pooled genetic screens, single-cell sequencing such as use of Perturb-seq combined with combinatorial indexing, and high-throughput sequencing to find targets of signaling regulators in different biological contexts. More than 1,500 individual perturbations are performed across six cell lines and five different biological signaling contexts.

With CaRPool-seq, the lab has combined use of CRISPR and single-cell genomics technologies to massively parallelize the measurement of cellular responses under high-throughput genetic perturbations, and those perturbations can involve either single genes or multiple genes.

Scale will keep rising and cost is likely to keep dropping. It helps that scRNA-seq can be performed with much less material than used to be needed and on a wider range of tissues, says Trapnell. One can do things, he says, that were “off limits before.”

What trips people up the most in scRNA-seq work, says Trapnell, is study design. It has been too expensive for scientists to do the study they want, so they do a different experiment. “I think that what’s really going to change in the next couple years is that now people will be able to do the study they want,” he says. “And that’s going to be really enabling for a lot of labs.”