$30 off During Our Annual Pro Sale. View Details »

R/Medicine 2022

R/Medicine 2022

Leonardo Collado-Torres

August 25, 2022
Tweet

More Decks by Leonardo Collado-Torres

Other Decks in Science

Transcript

  1. 29
    Spatially-resolved Transcriptomics Analysis
    with R/Bioconductor and Beyond
    Leonardo Collado-Torres, Ph.D.
    Lieber Institute for Brain Development
    R/Medicine
    August 25, 2022
    Keri Martinowich Stephanie C Hicks
    Lieber Institute Johns Hopkins
    @lcolladotor
    #spatialLIBD
    Kristen R Maynard
    Lieber Institute
    https://speakerdeck.com/
    lcolladotor/medicine-2022

    View Slide

  2. The spatial architecture of the brain is
    fundamentally connected to its function
    2
    chartdiagram.com slideshare.net

    View Slide

  3. 3
    Image Credit: Bo Xia, https://twitter.com/boxia7/status/1261464021322137600?s=12
    Studying gene expression in human brain
    Bulk RNA-seq Single cell/nucleus RNA-seq Spatial transcriptomics

    View Slide

  4. Laminar position of a cell influences its gene
    expression, morphology, physiology, and function
    4
    Kwan et al., 2012, Development

    View Slide

  5. Visium & Single nucleus RNA-sequencing technologies
    (Commercial platform 10x Genomics)
    5
    Single Cell Gene Expression
    Spatial Gene Expression

    View Slide

  6. Study design for Visium experiments in
    dorsolateral prefrontal cortex (DLPFC)
    6
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  7. Visualizing gene expression in a histological context
    7
    logcounts logcounts logcounts
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  8. 2 pairs spatial adjacent replicates x subject = 12 sections
    8
    Subject 1
    Subject 2
    Subject 3
    Adjacent spatial replicates (0µm) Adjacent spatial replicates (300µm)
    PCP4
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  9. “Pseudo-bulking” collapses data: spot to layer level
    9
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  10. Three statistical models to assess laminar enrichment
    “ANOVA”
    model
    10
    “Enrichment”
    model
    “Pairwise”
    model
    Is any layer different? Is one layer > the rest? Is layer X > layer Y?
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  11. 11
    Identification of laminar enriched genes
    “Enrichment” model
    Is one layer > the rest?
    Group FDR<0.05
    Layer1 3033
    Layer2 1562
    Layer3 183
    Layer4 740
    Layer5 643
    Layer6 379
    WM 9124
    Only a subset of previous layer
    marker genes in mouse and human
    showed laminar association
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  12. bioconductor.org/packages/spatialLIBD
    Pardo et al, BMC Genomics, 2022,
    https://doi.org/10.1186/s12864-022-08601-w
    Maynard, Collado-Torres, Nat Neuro, 2021
    Brenda Pardo Abby Spangler
    @PardoBree @abspangler

    View Slide

  13. 13
    SpatialExperiment: infrastructure for spatially resolved
    transcriptomics data in R using Bioconductor
    Righelli, Weber, Crowell, et al, Bioinformatics, 2022
    DOI https://doi.org/10.1093/bioinformatics/btac299
    Dario Righellli Helena L Crowell
    @drighelli @CrowellHL
    Lukas M Weber
    @lmwebr

    View Slide

  14. 14
    Madhavi Tippani
    @MadhaviTippani
    bioRxiv, doi: https://doi.org/10.1101/2021.08.04.452489

    View Slide

  15. 15
    We provided a framework for
    comparing clustering results vs the
    manual annotation (aka, ground
    truth)

    View Slide

  16. 16 Zhao et al, Nature Biotechnology, 2021

    View Slide

  17. 17 Zhao et al, Nature Biotechnology, 2021

    View Slide

  18. Openly sharing data accelerates science:
    share and you will reap the benefits too!
    18
    Us: 346 days Them: 271 days
    Total sequential (fictional): 617 days
    Reality (preprint to BayesSpace pub): 461 days
    Difference saved: 156 days
    Preprints: 190 days

    View Slide

  19. 19
    What helps also: provide a ground truth and a path
    towards benchmarking
    • Fully unsupervised was initially very far from the
    ground truth
    • Truth has caveats and should be considered a
    guideline
    • Ultimately, the goal is not to fully reproduce the
    ground truth, but learn what helps and what doesn’t
    • Ground truth will evolve ;)

    View Slide

  20. 20
    High accessions, citations,
    AltMetric, …
    This data is way more
    challenging than the mouse:
    mouse you are looking at
    different brain regions

    View Slide

  21. The Development Process
    - Making a module
    - New, experimental software can change dramatically (function
    and syntax) between versions
    - Promotes collaboration by allowing two researchers to share
    exact code and instantly run software without special set-up
    SpatialExperiment release 3.14
    SpatialExperiment devel 3.15
    module load tangram/1.0.2
    module load cell2location/0.8a0
    module load spagcn/1.2.0
    @Nick-Eagles (GH)
    Nicholas J Eagles
    https://github.com/LieberInstitute/jhpce_mod_source
    https://github.com/LieberInstitute/jhpce_module_config

    View Slide

  22. The Development Process
    - Regular interaction with software
    authors to clarify functionality and
    report bugs
    - Documentation for code and author
    responsiveness on GitHub can be
    critical in successfully applying
    software to our data
    @Nick-Eagles (GH)
    Nicholas J Eagles

    View Slide

  23. Documentation + wrapper functions + tests (GitHub Actions +
    Bioconductor)
    23
    http://bioconductor.org/packages/spatialLIBD
    http://bioconductor.org/packages/release/data/experiment/vignettes/spatialLIBD/
    inst/doc/TenX_data_download.html

    View Slide

  24. Gandal et al, Science, 2018
    SFARI GENE; 2.0 by Abrahams et al, Mol Autism, 2013
    Jaffe et al, Nature Neuroscience, 2020
    - Curated lists
    - GWAS/TWAS
    hits
    - Differential
    expression
    - …
    Layer-enriched gene expression profiling

    View Slide

  25. 0
    2
    4
    6
    8
    10
    12
    WM
    L6
    L5
    L4
    L3
    L2
    L1
    SFAR
    I
    ASC
    102
    ASD
    53
    D
    D
    ID
    49
    D
    E.U
    p
    D
    E.D
    ow
    n
    2.7
    2.1
    2.7
    4
    3.6
    4.9
    4.5
    2.5
    5
    2.8
    5
    6.4
    2.8
    ASD
    WM
    L6
    L5
    L4
    L3
    L2
    L1
    PE.U
    p
    PE.D
    ow
    n
    BS2.U
    p
    BS2.D
    ow
    n
    BS2.U
    p
    BS2.D
    ow
    n
    PE.U
    p
    PE.D
    2.1
    2
    3.1
    1.8
    2.2
    1.8
    8.8
    5
    2.7
    2.6
    4.6
    SCZD−DE SCZD−TW
    (A) (B)
    DIY at
    http://spatial.libd.org/spatialLIBD/
    Laminar-enrichment of clinical gene sets
    Autism Spectrum Disorder (ASD)
    • SFARI: Abrahams et al, Mol Autism,
    2013
    • ASC102: Satterstrom et al, Cell,
    2020
    Break up into:
    • ASD53: ASD dominant traits
    • DDID49: neurodevelopmental
    delay
    COLOR is significance (-log10[p])
    NUMBER is enrichment (odds ratio)
    25
    Maynard, Collado-Torres, et al, Nat Neuro, 2021

    View Slide

  26. 26
    Adopted and modified from B Wang (2018) and the Brain from the Top to Bottom in McGill University
    Progressive neurodegeneration
    in Alzheimer’s disease

    View Slide

  27. Integration of proteomic and transcriptomic data
    with Visium-Immunofluorescence (Visium-IF)
    27
    Can we define pathology-associated changes in gene
    expression in Alzheimer’s Disease in human brain?

    View Slide

  28. 28
    Visium-IF AD Study Design (Inferior Temporal Cortex)
    Sang Ho Kwon
    @sanghokwon17
    (Kwon et al., in preparation)

    View Slide

  29. Visium
    * ~5k spots in honeycomb
    * gene expression per spot
    * tissue (H&E staining)
    Immunofluorescence (IF)
    * multi-channel (6) images
    * identifies morphological features of interest
    * large: might be broken in tiles
    Channel 1
    * triangle feature
    Channel 2
    * cloud feature
    Channel 6
    * xyz feature
    Tissue (bright field image)
    Visium spot
    Channel 1 feature
    Channel 2 feature
    +
    Visium-IF raw data: 2 types

    View Slide

  30. Spot ID # Triangle # Cloud % triangle % cloud
    spot0001 0 12 0 17
    spot0002 4 0 27 0
    Merge Visium & IF
    IF
    Spot ID Gene 1 Gene 2 Gene X In
    Tissue
    # cells
    spot0001 0 12 39 true 3
    spot0002 4 0 27 false 0
    Visium
    downstream
    * QC
    * analyses

    View Slide

  31. 31
    Registering pathology maps with gene expression spots
    Madhavi Tippani
    @MadhaviTippani
    (Kwon et al., in preparation)
    Prop IF/Spot
    VistoSeg now supports Visium-IF

    View Slide

  32. Annotating and pseudo-bulking spots by pathology
    for differential expression analyses
    32 Sowmya Parthiban
    @sowmyapartybun (Kwon et al., in preparation)

    View Slide

  33. Identification of genes associated with AD pathology
    33
    (Kwon et al., in preparation)
    p<0.05 in
    targeted
    sequencing
    panel

    View Slide

  34. Working with Visium
    • It’s very powerful
    • Open source friendly
    • 6.5 mm2 too restrictive? Opportunity for creativity
    • Visium and Visium-IF have required the development of software
    • It’s fun to work on something where there are no answers on Google =)
    but also a challenge
    34

    View Slide

  35. Future Directions
    • Integration of proteomic and transcriptomic data
    • Visium-IF AD proof-of-concept
    • Integration of snRNA-seq and Visium data
    • Visium + snRNA-seq on LC
    • Increasing resolution (# spots) and area (array size)
    • Visium HD
    • Leveraging rich histology/imaging data
    • Clustering (SpaGCN), spot deconvolution, etc.
    • Building educational resources
    • Completing Orchestrating Spatially Resolved
    Transcriptomics Analysis with Bioconductor (OSTA)
    35

    View Slide

  36. Acknowledgements
    Lieber Institute
    Sang Ho-Kwon
    MadhaviTippani
    Abby Spanger
    Brenda Pardo
    Joseph L. Catallini II
    Matthew N. Tran
    Vijay Sadashivaiah
    Heena Divecha
    Kelsey Montgomery
    Nick Eagles
    Josh Stolz
    Louise Huuki
    Rahul Bharadwaj
    Stephanie Page
    Leonardo Collado-Torres
    Keri Martinowich
    Andrew Jaffe
    Joel E. Kleinman
    Thomas M. Hyde
    Daniel R. Weinberger
    JHU Biostatistics Dept
    Stephanie Hicks
    Lukas Weber
    Sowmya Parthiban
    10x Genomics
    Courtney Anderson
    Cedric Uytingco
    Stephen R. Williams
    Charles Bruce
    Jennifer Chew
    YifengYin
    Nikhil Rao
    Michelle Mak
    Guixia Yu
    Julianna Avalos-Gracia
    JHU Oncology Tissue Services (Kristen Lecksell)
    JHU SKCCC Flow Core (Jessica Gucwa)
    JHU Transcriptomics & Deep Sequencing Core (Linda Orzolek)
    JHU Tumor Microenvironment Core (Liz Engle)
    We are hiring! https://www.libd.org/careers/
    @lcolladotor
    #spatialLIBD team

    View Slide

  37. #spatialLIBD is a supportive LIBD & JHU team
    37
    Check for your yourself at
    https://twitter.com/lcolladotor/status/1516587531369811971
    https://lcolladotor.github.io/team_surveys/

    View Slide

  38. We are hiring! https://www.libd.org/careers/
    38

    View Slide