Single Cell Genomics

Single cell genomics is the branch of genomics that reads the DNA, RNA, or chromatin of individual cells rather than of bulk tissue.

11 frameworks, from 1990 to today · Genomics · Natural Sciences

still a working view ended · each leads to its entry below

Overview

Single cell genomics is the branch of genomics that reads the DNA, RNA, or chromatin of individual cells rather than of bulk tissue12. Its premise is that a tissue is not one genome or one transcriptome but thousands of slightly different ones, and that averaging over them hides the biology that matters. Bulk sequencing of a tumor returns one consensus sequence in which a mutation carried by a small fraction of cells may be invisible; bulk sequencing of a brain region returns an average expression profile that no actual neuron has13. Single cell genomics exists to recover that lost variation: which cells carry which mutations, which cells express which genes, and how a single fertilized egg gives rise to hundreds of distinct cell types. What is at stake is a cell-by-cell account of how genomes build and corrupt bodies.

Amplifying a genome from a picogram

The founding problem was material. A mammalian cell contains roughly six picograms of DNA, far below what sequencing or even gel electrophoresis could detect in the 1980s and 1990s. The first single cell genomic work therefore came from clinical genetics, where the target was not discovery but diagnosis: preimplantation genetic diagnosis of embryos and analysis of single sperm or fetal cells. In the early 1990s, researchers including Alan Handyside in London demonstrated that a single cell biopsied from an eight-cell embryo could be genotyped by polymerase chain reaction (PCR), the enzymatic copying of a DNA segment. The obstacle was that PCR from one or two template molecules is unreliable: alleles drop out, and the whole genome must first be copied without bias.

Whole-genome amplification (WGA) became the enabling technology. Degenerate oligonucleotide-primed PCR (DOP-PCR), introduced in the early 1990s, amplified a genome using primers that bind at many scattered sites; it was robust but uneven. Multiple displacement amplification (MDA), developed in the early 2000s using the phi29 polymerase, copied DNA with high fidelity and long products but introduced strong sequence-dependent bias. In 2012, Zong and colleagues at Harvard introduced MALBAC (multiple annealing and looping-based amplification cycles), which reduced that bias enough to call copy-number variants and even single-nucleotide variants from one cell. These methods made single-cell PCR genotyping routine in reproductive medicine and cancer research, and they set the pattern for the field: a chemical trick to amplify scarce material, followed by a computational correction for its distortions.

Reading the transcriptome of one cell

Genomes are static; the more informative signal is which genes a cell is transcribing. The first single-cell transcriptome work came in the early 1990s from molecular biologists studying neurons, who used reverse transcription and PCR to read a handful of genes from individual cells. The decisive step was the demonstration, by Iscove and colleagues in 1990 and later by Brady and Iscove, that the entire expressed repertoire of a single cell could be amplified. In the 2000s, James Eberwine's laboratory and others developed linear amplification of mRNA from single neurons, and in 2009 Fuchou Tang, then in Azim Surani's group in Cambridge, published the first whole-transcriptome sequencing of single mouse cells, combining reverse transcription, poly-A tailing, and PCR amplification.

The dominant chemistry that followed was plate-based full-length transcriptomics: cells sorted one per well into 96- or 384-well plates, lysed, reverse-transcribed with a template-switching mechanism that adds a known sequence to the cDNA end, then amplified and sequenced. The Smart-seq family, introduced by Picelli and colleagues in 2013, became the standard. Its virtue was full-length coverage, which allowed detection of alternative splicing and of low-expressed genes; its cost was throughput, since each cell required its own well and its own library. For a decade this approach carried most single-cell biology done on hundreds or low thousands of cells.

Droplets and barcodes: the throughput revolution

The limitation of plate-based methods was scale. Cell atlases and developmental trajectories need tens of thousands to millions of cells, and pipetting cannot get there. The solution came from microfluidics and from a conceptual shift: instead of keeping each cell in a separate tube, pool everything and use a short DNA barcode to remember which cell each molecule came from.

In 2015, two groups independently published droplet-based methods. Macosko and colleagues at Harvard, working with the McCarroll laboratory, introduced Drop-seq, and Klein and colleagues at Harvard Medical School introduced inDrop. Both encapsulated single cells in oil droplets together with a bead carrying millions of barcoded primers; each transcript was tagged with a cell barcode and a unique molecular identifier (UMI), a random sequence that lets software count original molecules rather than PCR duplicates. The result was thousands of cells per run at a fraction of the cost per cell. Commercial platforms, notably 10x Genomics' Chromium system launched in 2016, made the approach standard. The trade-off was that droplet methods capture only the 3′ or 5′ end of each transcript and detect fewer genes per cell than Smart-seq, so the two families coexisted: plate-based full-length methods for depth and splicing, droplet methods for scale.

Beyond RNA: chromatin, DNA, and lineage

Once cells could be counted in millions, the next question was what else could be measured in them. Single-cell chromatin accessibility profiling adapted the assay for transposase-accessible chromatin (ATAC-seq), in which an enzyme inserts sequencing tags into open, regulatory DNA, to single cells14. Buenrostro and colleagues published the first single-cell ATAC-seq in 2015, and droplet-based versions followed, allowing researchers to ask which regulatory elements are open in which cell type—a layer of information invisible to RNA sequencing.

In parallel, single-cell DNA sequencing matured. Whole-genome amplification combined with sequencing allowed reconstruction of copy-number evolution in tumors, and in 2014 Navin and colleagues used single-cell sequencing to show that breast tumors contain distinct clonal subpopulations. Single-cell clonal lineage tracing took a different route: rather than reading a cell's genome, it reads a heritable mark written into it. Methods using CRISPR to introduce accumulating mutations into a genomic "barcode" region, and earlier approaches using naturally occurring somatic mutations or viral integration sites, let researchers reconstruct the family tree of cells in a developing tissue. The lineage-tracing work of Weissman and colleagues, and later CRISPR-based recorders, made it possible to ask not just what a cell is but where it came from.

Computation as a discipline

The data these methods produce are sparse and noisy: most genes read as zero in any given cell, partly because they are not expressed and partly because the chemistry missed them. Single-cell computational analysis grew into a field of its own to handle this. Normalization methods corrected for differing capture efficiency; dimensionality reduction (principal component analysis, then t-SNE and UMAP) projected thousands of genes onto two dimensions so that cell types appeared as clusters; clustering algorithms grouped cells by expression profile; and trajectory inference methods ordered cells along a developmental path. Tools such as Seurat, Scanpy, and the earlier Cell Ranger pipeline became standard, and the field developed its own statistical vocabulary—batch correction, doublet detection, imputation—for problems that bulk genomics never faced. The computational layer is not an afterthought: much of what single-cell genomics claims rests on choices made in software.

Atlases, consortia, and multi-omics

With throughput and computation in place, the field turned to cataloging. The Human Cell Atlas, launched in 2016 with the aim of describing every cell type in the human body, and the earlier Mouse Cell Atlas and Tabula Muris efforts, organized hundreds of laboratories around shared protocols and data standards. Similar consortia mapped the developing human and mouse embryo, the immune system, and individual organs. These atlases were as much social and infrastructural achievements as technical ones: they required agreed cell-type ontologies, data portals, and reproducibility standards.

The most recent turn is single-cell multi-omics: measuring more than one modality in the same cell. Methods such as CITE-seq (2017) attached antibodies to RNA reads so that surface proteins and transcripts were captured together; joint ATAC-RNA assays read chromatin and expression from one cell; and spatial transcriptomics added position. Spatial methods, which measure expression while preserving where each cell sat in the tissue, developed from early in situ sequencing and barcoded array approaches in the mid-2010s into a major branch of the field, because a cell's identity depends on its neighbors. The combination of single-cell resolution with spatial context is now the frontier, alongside single-cell proteomics and methods that preserve full-length isoform information at scale.

Where the field stands

Today single cell genomics runs on a small set of dominant platforms—droplet-based 3′ capture for scale, plate-based full-length methods for depth, combinatorial barcoding for very large experiments—with spatial methods and multi-omic assays layered on top. The computational stack has consolidated around a few open-source ecosystems. The open questions are substantive rather than merely technical: how to define a cell type when expression varies continuously; how to distinguish technical noise from biological signal in sparse data; how to scale spatial resolution to the level of individual transcripts; and how to move from cataloging cells to predicting what they will do. The field began by amplifying a picogram of DNA so that a clinician could genotype an embryo, and it now aims at a complete, spatially resolved, multi-modal map of every cell in the body.

References

  1. Single-cell sequencing (introduction, Transposase-accessible chromatin sequencing (scATAC-seq)), Wikipedia
  2. Single-cell analysis, Wikipedia
  3. Single-cell transcriptomics (Background), Wikipedia
  4. ATAC-seq, Wikipedia

The frameworks of Single Cell Genomics

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1990-2010

    Single-cell PCR genotyping

    Can a single cell's genotype be read reliably?

    Fly to its tower in Nousfarer

  2. 1992-Present

    Whole-genome amplification

    How can a picogram of DNA be copied without bias?

    Fly to its tower in Nousfarer

  3. 2009-Present
  4. 2011-Present

    Single-cell DNA sequencing and clonal evolution

    Which mutations distinguish one tumor cell from another?

    Fly to its tower in Nousfarer

  5. 2013-Present

    Single-cell computational analysis

    How can sparse single-cell data be made sense of?

    Fly to its tower in Nousfarer

  6. 2014-Present

    Single-cell lineage tracing

    Where did each cell in a tissue come from?

    Fly to its tower in Nousfarer

  7. 2015-Present

    Droplet barcoding and UMI counting

    How can millions of cells be read at once?

    Fly to its tower in Nousfarer

  8. 2015-Present
  9. 2015-Present

    Spatial transcriptomics

    Which genes are expressed where in a tissue?

    Fly to its tower in Nousfarer

  10. 2016-Present

    Cell atlas consortia

    What cell types make up a whole body?

    Fly to its tower in Nousfarer

  11. 2017-Present

    Single-cell multi-omics

    How do several molecular layers combine in one cell?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Single-cell PCR genotypingWhole-genome amplificationPlate-based full-length single-cell t…Single-cell DNA sequencing and clonal…Single-cell computational analysisSingle-cell lineage tracingDroplet barcoding and UMI countingSingle-cell chromatin accessibility p…Spatial transcriptomicsCell atlas consortiaSingle-cell multi-omics

Beyond the field

Where Single Cell Genomics's ideas came from in other fields, and what they led to there: the strongest 12 of 40 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Single Cell Genomics is a planet in Nousfarer

Its city has a tower for each of these 11 frameworks. Land, walk in, and climb one concept at a time.

Land on Single Cell Genomics

Related fields

Fields its ideas are linked to

More in Genomics

See where Single Cell Genomics sits on the Map of Human Knowledge