What do we need computationally to make the most of single-cell RNA-seq data?

Invited talk at Genome Informatics 2016, held at the Wellcome Genome Campus, 19-22 September 2016.

The talk focuses primarily on the R/Bioconductor package "scater" for pre-processing, quality control, normalisation and visualisation of single-cell RNA-seq data, but also addresses open questions in the field - such as the necessity of dealing with batch effects in our analyses and areas for future development of computational methods and tools.

Transcript

1. What do we need computationally so that we can make the most of single-cell RNA-seq data? Davis McCarthy NHMRC Early Career Fellow Stegle Group, EMBL-EBI @davisjmcc www.ebi.ac.uk
2. Solid Tissue Dissociation Single Cell Isolation RNA RT& Second-strand Synthesis Amplified RNA cDNA Amplified cDNA Sequencing Library Sequencing Single-cell Expression Profiles Cell Types Identification IVT RT PCR OR Clustering Single Cell RNA Sequencing Workflow CAAGTTCCTACAGCTA AGTCCATGCCCATCCG AATCGGACTTCAGCCT GACCTAAGCCATCAGA AATCCTAGCATCCAGC ACCGTTACATCAACAG ATTCGATAACGACCAT CATGCCATTGACGATT Single-cell RNA sequencing https://en.wikipedia.org/wiki/File:RNA-Seq_workflow-5.pdf
3. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
4. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
5. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package E6.5 Pou5f1 (Oct4) Expression
6. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
7. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
8. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
9. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Flk1+ Pou5f1 (Oct4) Expression
10. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
11. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
12. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
13. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package CD41+ Pou5f1 (Oct4) Expression
14. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package CD41+ Pou5f1 (Oct4) Expression
15. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
16. Differentiation to mouse mesoderm (Scialdone et al, Nature, 2016) Visualisation with scater package Pou5f1 (Oct4) Expression
17. scRNA-seq data is high-resolution and high- dimensional. What can we do with it? Mouse mesoderm data (Scialdone et al, 2016) ENSMUSG00000024406_Pou5f1 0.0 2.5 5.0 7.5 0 300 600 900 DPT_rank Expression [log2(cpm + 1)] cellCategory E6.5 PS-Flk1+ NP-Flk1+ NP-CD41+Flk1+ NP-CD41+Flk1- HF-Flk1+ HF-CD41+Flk1+ HF-CD41+Flk1- Pou5f1 expression against diffusion pseudotime
18. So if we want general methods that make the most of single-cell RNA-seq data to infer gene regulation relationships, differential expression, sub-populations, etc… https://giphy.com/gifs/thebachelorette-yeah-um-uh-3o7TKVaVwWwRJjmJOg …we have some work to do.
19. So if we want general methods that make the most of single-cell RNA-seq data to infer gene regulation relationships, differential expression, sub-populations, etc… https://giphy.com/gifs/thebachelorette-yeah-um-uh-3o7TKVaVwWwRJjmJOg …we have some work to do.
20. What do we need?
21. What do we need? Well-posed question(s)
22. What do we need? Well-posed question(s) Answers with understanding of uncertainty
23. What do we need? Well-posed question(s) Answers with understanding of uncertainty
24. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data
25. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data
26. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data
27. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data good statistical model
28. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data software tools good statistical model
29. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data software tools good statistical model +
30. What do we need? Well-posed question(s) Answers with understanding of uncertainty the right data software tools good statistical model +
31. Let’s imagine that we carry out our well- designed study… …and get good-quality sequence data to analyse.
32. So we’ll just assume that all the cells worked well?
33. So we’ll just assume that all the cells worked well?
34. So we’ll just assume that all the cells worked well?
35. We need to conduct careful pre-processing, QC and normalisation before getting to any downstream analyses Raw data Pre-processing Quality control Downstream analysis e.g. expression quantification Normalisation
36. How do we QC our data?
37. How do we QC our data? • This depends on context, so is hands on
38. How do we QC our data? • This depends on context, so is hands on • Visualise it
39. How do we QC our data? • This depends on context, so is hands on • Visualise it • Calculate gene-level and cell-level QC metrics
40. How do we QC our data? • This depends on context, so is hands on • Visualise it • Calculate gene-level and cell-level QC metrics • Use QC metrics to filter out potentially problematic genes and cells
41. scater: R/Bioconductor package for pre- processing, QC, normalisation and visualisation • http://bioconductor.org/packages/scater/
 • bioRxiv pre-print: http://dx.doi.org/10.1101/069633
 • “A step-by-step workflow for low-level analysis of single-cell RNA-seq data": 
 http://f1000research.com/articles/5-2122/v1

42. scater workflow: powerful and flexible scater pre-processing and quality control workflow From raw RNA-seq reads to a clean, tidy dataset ready for downstream analysis Raw RNA-seq Reads [Fastq format] Summarised feature expression values [e.g. produced by bioinformatics core] runKallisto/ readKallisto runSalmon/readSalmon newSCESet Plotting methods plot plotQC plotPCA plotTSNE plotMDS plotDiffusionMap plotReducedDim plotExpression plotPhenoData plotFeatureData plotMetadata Filtered SCESet Tidy filtered and normalised SCESet Downstream modelling and statistical analysis 1. Obtain RNA-seq expression data 2. QC and filter features 3. QC and filter cells 4. Simple normalisation 5. QC of explanatory variables SCESet [Container: S4 class inheriting Bioconductor’s ExpressionSet] Object that contains assay data, phenotype data, feature data, and more, for single-cell analysis (6. Further normalisation) QC methods calculateQCMetrics Miscellaneous methods getBMFeatureAnnos summariseExprsAcross Features Normalisation methods normaliseExprs normalise • Integrated with Salmon + kallisto • Automatic calculation of QC metrics • QC diagnostic plots • Cell and gene filtering • Simple normalisation • Sophisticated data structure for single-cell data • Beautiful plots • Shiny GUI
43. scater QC’d SCESet object contains expression assay data, phenotype data, feature data, and more, for single-cell analysis Expression data from e.g. kallisto, Salmon, RSEM, featureCounts, HTSeq, etc. Cell and gene metadata from study design, expression quantification tool, etc. Normalisation BASiCS GRM scran Differential Expression BASiCS BPSC M3Drop MAST monocle scDD scde Heterogeneous Expression BASiCS scran Clustering PAGODA RaceID SC3 SINCERA Latent Variable Analysis cellCODE PEER RUVSeq svaseq Cell Cycle cyclone Pseudotime DeLorean destiny dpt embeddr monocle ouija SINCELL TSCAN scater ecosystem: take advantage of many other R/Bioconductor packages
44. Little example # cells C1 Machine 1 C1 Machine 2 Patient A 11 13 Patient B 25 24 For detailed case study see scater preprint on bioRxiv doi: http://dx.doi.org/10.1101/069633