Upgrade to Pro — share decks privately, control downloads, hide ads and more …

RNA-quality-degradation

 RNA-quality-degradation

Applying Statistical Correction for Brain Tissue RNA Degradation to Gene Expression Differences in Schizophrenia

References:
BrainSeq: A Human Brain Genomics Consortium. Brainseq: neurogenomics to drive novel target discovery for neuropsychiatric disorders. Neuron 2015;88(6):1078-1083. doi:10.1016/j.neuron.2015.10.047.
Jaffe AE, Tao R, Norris AL, et al. qSVA framework for RNA quality correction in differential expression analysis. Proc Natl Acad Sci USA 2017;114(27):7130-7135. doi:10.1073/pnas.1617384114.
Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet. 2007;3(9):1724-1735. doi:10.1371/journal.pgen.0030161.

Amy Peterson

May 04, 2018
Tweet

More Decks by Amy Peterson

Other Decks in Science

Transcript

  1. 11
    Applying Statistical Correction for
    Brain Tissue RNA Degradation to
    Gene Expression Differences
    in Schizophrenia
    Amy Peterson
    Capstone Advisors:
    Andrew Jaffe, PhD
    Leonardo Collado-Torres, PhD

    View full-size slide

  2. 2018
    Elon Musk
    put a Tesla in
    space
    THE SCIENTIFIC FRONTIER
    65+ YEARS: WAITING FOR A BREAKTHROUGH
    Molecular targets of all current psychotherapeutic drugs are the same as their 1950’s prototypes.
    1957
    Sputnik I
    1952
    Discovery of
    Antipsychotic
    Chlorpromazine
    (DRD2 blockade)
    2018
    Antipsychotics
    for treatment
    of schizophrenia
    all work via
    DRD2 blockade
    ?
    2

    View full-size slide

  3. Animal Models Neuronal Cell Models
    Drug Discovery
    New Treatments
    2300+
    Human postmortem brains
    1000+
    Cell lines from individuals
    Genomics + Transcriptomics + Proteomics
    3
    Mechanisms of Illness
    Clinical Genetics
    BrainSeq: A Human Brain Genomics Consortium
    THE SCIENTIFIC FRONTIER

    View full-size slide

  4. Animal Models Neuronal Cell Models
    Drug Discovery
    New Treatments
    2300+
    Human postmortem brains
    1000+
    Cell lines from individuals
    Genomics + Transcriptomics + Proteomics
    4
    Mechanisms of Illness
    Clinical Genetics
    BrainSeq: A Human Brain Genomics Consortium
    DLPFC
    495 samples
    BrainSeq Phase I
    polyA
    Jaffe et al., bioRxiv, 2017
    THE SCIENTIFIC FRONTIER

    View full-size slide

  5. Animal Models Neuronal Cell Models
    Drug Discovery
    New Treatments
    2300+
    Human postmortem brains
    1000+
    Cell lines from individuals
    Genomics + Transcriptomics + Proteomics
    5
    Mechanisms of Illness
    Clinical Genetics
    BrainSeq: A Human Brain Genomics Consortium
    DLPFC
    495 samples
    BrainSeq Phase I
    polyA
    Jaffe et al., bioRxiv, 2017
    DLPFC
    453 samples
    HIPPO
    447 samples
    BrainSeq Phase II
    RiboZero
    THE SCIENTIFIC FRONTIER

    View full-size slide

  6. BACKGROUND
    6
    RESEARCH QUESTION
    Do patients with schizophrenia exhibit gene
    expression differences across various brain
    regions compared to non-psychiatric controls?

    View full-size slide

  7. BACKGROUND
    7
    RESEARCH QUESTION
    Do patients with schizophrenia exhibit gene
    expression differences across various brain
    regions compared to non-psychiatric controls?
    • Impact of RNA quality?
    • Functionality of genes identified as differentially
    expressed?

    View full-size slide

  8. RNA-seq reads
    Genome
    (DNA)
    RNA transcripts
    (many possible
    variants)
    Measuring gene expression: RNA-seq
    Adapted from @jtleek
    8

    View full-size slide

  9. SAMPLE SIZE
    9
    • 712 total RNA-seq samples
    • 379 DLPFC, 333 HIPPO
    • 447 Individuals
    • 177 Schizophrenia cases
    • 270 Non-psychiatric controls
    Dataset summary
    DLPFC HIPPO
    CASE 153 133
    CONTROL 226 200
    DLPFC
    453 samples
    HIPPO
    447 samples

    View full-size slide

  10. qSVA WORKFLOW
    10
    quality surrogate variable analysis (qSVA)
    Degradation confounds
    postmortem human
    brain gene expression
    PCA

    View full-size slide

  11. qSVA WORKFLOW
    11

    View full-size slide

  12. qSVA WORKFLOW
    12

    View full-size slide

  13. qSVA WORKFLOW
    13
    PCA

    View full-size slide

  14. 14
    REGION-SPECIFIC qSVs
    qSV1 associated with RIN and case-control status
    SCZD
    SCZD

    View full-size slide

  15. DEqual HIPPO
    15
    Model 1 (6429 genes)
    Model 1. Naïve model
    E = 0
    + 1

    DEqual plots demonstrate effectiveness of statistical correction
    HIPPO
    333 samples
    r = 0.412

    View full-size slide

  16. DEqual HIPPO
    16
    Model 1 (6429 genes) Model 2 (63 genes)
    Model 1. Naïve model
    E = 0
    + 1

    Model 2. Added RNA-quality and demographic covariates
    E = 0
    + 1
    + 2
    + 3
    + 4
    + 5
    te + 6

    + 7
    + ∑ γ

    ^
    _`a
    DEqual plots demonstrate effectiveness of statistical correction
    HIPPO
    333 samples
    r = 0.412 r = 0.0712

    View full-size slide

  17. DEqual HIPPO
    17
    Model 1 (6429 genes) Model 2 (63 genes) Model 3 (48 genes)
    Model 1. Naïve model
    E = 0
    + 1

    Model 2. Added RNA-quality and demographic covariates
    E = 0
    + 1
    + 2
    + 3
    + 4
    + 5
    te + 6

    + 7
    + ∑ γ

    ^
    _`a
    Model 3. Added qSVs
    E = 0
    + 1
    + 2
    + 3
    + 4
    + 5
    te + 6

    + 7
    + ∑ γ

    ^
    _`a
    + ∑ Ζ

    h
    _`a
    DEqual plots demonstrate effectiveness of statistical correction
    HIPPO
    333 samples
    r = 0.412 r = 0.0712 r = -0.00173

    View full-size slide

  18. DIFFERENTIAL EXPR
    18
    A B
    Comparing differentially expressed genes
    • A. T-statistics for top 400 differentially expressed genes for
    HIPPO compared to DLPFC (blue line: regression, red line: loess)
    • B. BrainSeq Phase 2 and BrainSeq Phase 1 (BSP1) DLPFC
    −6 −4 −2 0 2 4
    −4 −2 0 2 4
    t−statistic HIPPO
    t−statistic DLPFC
    r = 0.644
    −6 −4 −2 0 2 4 6
    −6 −4 −2 0 2 4 6
    t−statistic DLPFC
    t−statistic BSP1
    r = 0.809

    View full-size slide

  19. SUMMARY
    19
    • Statistical correction successfully removed
    confounding effect of RNA quality (DEqual plots)
    • Findings comparable with previous use of qSVA in
    BrainSeq Phase I (DLPFC)
    Final Model
    E = 0
    + 1
    + 2
    + 3
    +
    4
    + 5
    te + 6
    + 7
    +


    ^
    _`a
    + ∑ Ζ

    h
    _`a

    View full-size slide

  20. REFERENCES
    20
    References
    BrainSeq: A Human Brain Genomics Consortium. Brainseq: neurogenomics to drive novel target
    discovery for neuropsychiatric disorders. Neuron 2015;88(6):1078-1083.
    doi:10.1016/j.neuron.2015.10.047.
    Jaffe AE, Tao R, Norris AL, et al. qSVA framework for RNA quality correction in differential
    expression analysis. Proc Natl Acad Sci USA 2017;114(27):7130-7135. doi:10.1073/pnas.1617384114.
    Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by surrogate variable
    analysis. PLoS Genet. 2007;3(9):1724-1735. doi:10.1371/journal.pgen.0030161.
    Code: https://github.com/LieberInstitute/qsva_brain
    JHPCE Cluster. https://jhpce.jhu.edu
    >838,168 linear regressions
    All analyses completed in R 3.4
    Resources and Reproducibility

    View full-size slide

  21. THANK YOU!
    Questions?
    Amy Peterson
    MPH Candidate 2018
    [email protected]
    amy-peterson.github.io
    21
    Acknowledgements
    • Andrew Jaffe and lab members
    • BrainSeq Consortium
    • Leonardo Collado-Torres

    View full-size slide