Review

Single-cell epigenomics: Recording the past and predicting the future

See allHide authors and affiliations

Science  06 Oct 2017:
Vol. 358, Issue 6359, pp. 69-75
DOI: 10.1126/science.aan6826

Figures

  • Fig. 1 Single-cell methods and heterogeneity of different molecular layers.

    (Left) Overview of different molecular layers that can be assayed using single-cell protocols. (Right) A cell with different layers of multi-omics measurements, as defined on the left. Concordance or heterogeneity respectively may exist between the different layers, and this can be recorded by single-cell sequencing and computationally evaluated.

  • Fig. 2 Depth versus breadth: Multi-omics and cell-barcoding methods.

    Examples of different technical approaches are shown. (Top) Single-cell nucleosome, methylation, and transcription sequencing (scNMT-seq) (11) by which nucleosome accessibility, DNA methylation, and the transcriptome are read simultaneously at considerable depth in each cell; however, with individual cells processed in parallel but separately, cell numbers that can be currently analyzed in this way are limited to hundreds or thousands. (Middle) Barcoding chromatin in individual cells encapsulated in oil droplets, followed by pooling to bulk up material, enables thousands of cells to be processed while seeking to preserve signal-to-noise ratio (12). (Bottom) Combinatorial-cell barcoding (8, 64), where readouts can be identified as coming from individual cells by unique combinations of barcodes present in each cell. This approach can be carried out on large numbers of cells (millions), but the depth of information per cell is limited.

  • Fig. 3 Multi-omics and computational methods.

    Shown are typical trade-offs between single-cell RNA-seq, single-cell epigenome protocols, and multi-omics methods that provide readouts from multiple molecular layers in parallel. Consequently, it is commonly required to integrate data from different sequencing protocols. Raw sequence reads from these methods are deduplicated and aggregated into locus-specific readouts, with an optional imputation step to complete missing information. Associations between molecular layers can be used for completing missing data and allow for discovering regulatory associations.

  • Fig. 4 Time scales of epigenetic heterogeneity at different layers and lineage tracing.

    (A) Shown are different layers of information that can be recorded at least in principle by single-cell multi-omics, from transcription factor binding and transcriptional responses to long-term epigenetic memory such as is possible with DNA methylation. Rough time scales are indicated by colored bars—with shading indicating transitions in information—and may range from seconds to years. With aging, fidelity of epigenetic information such as DNA methylation may degrade, leading to increased cell-to-cell heterogeneity. (B) Lineage tracing using genetic or epigenetic memory. Cell lineage can be traced by CRISPR scarring approaches in which each cell and its descendants within a lineage are linked by unique mutations or barcodes. DNA modifications may also be used to track lineage based on their inheritance and on errors in their maintenance at DNA replication. Nonheritable modifications (5hmC, 5fC, and 5caC) have a short-term lineaging potential, whereas heritable modifications (5mC) have long-term noninvasive lineaging potential.

Navigate This Article