/cseaR

Cell type enrichment analysis borrowing functions from pSI, EWCE, and phenoTest R packages.

cseaR

Cell type enrichment analysis borrowing functions from pSI, EWCE, and phenoTest R packages.

About this package

cseaR includes code/workflow (Rmd) and pre-processed cell data for human analysis

The CSEA methods I use are:

  • pSI R package (a "competitive" enrichment) - with list of significant (e.g. FDR < 0.05) genes
  • EWCE R package (a "summed" enrichment) - with list of significant (e.g. FDR < 0.05) genes
  • GSEA function from phenoTest R package modified for CSEA - with full ranked t-statistics (gam error, bypassed by doing 10 genesets at a time; SUB-ideal!)

The cell datasets I use are:

  • Darmanis Human SCT (based on data from Andrew Jaffe)
  • Lake Human SCT (Neuron subtypes)
  • Barres Human RNA-Seq from sorted cells (pSI only since not SCT)
  • Linnarsson Mouse SCT (Celltypes and extensive subtypes -- Neuron human~mouse agreement is pretty good but glia not so much)
  • Barres Mouse RNA-Seq from sorted cells (for mouse experiments) (pSI only since not SCT)

Details about cell datasets

Darmanis, et al PNAS 2015 - Human SCT

Linnarsson: Zeisel, et al. Science 2015 - Mouse SCT

Ben Barres - Mouse and Human (not SCT)

Downloaded 7/4/2016
Average taken for replicates

Human RNA-Seq

Mouse RNA-Seq

Lake et al Science 2016 - Human SCT

  • Part of SCAP-T
  • README
  • Data
    • Summarized data available publically online
    • Full raw data available at dbGaP (accession phs000833.v4.p1)
  • Tissue = 6 cortical regions: FC (BA8, BA10), TC (BA21, BA22, BA41), and VC (BA17)
    • BA8 = frontal cortex (FC)
    • BA10 = anterior pre-frontal cortex (PFC)
    • BA17 = visual cortex in occipital lobe (VC)
    • BA21 = middle temportal cortex; auditory (mTC)
    • BA22 = superior temporal cortex; auditory (sTC)
    • BA41 = anterior transverse temporal gyrus (aTC)
  • Selection = NeuN+ nuclei from normal postmortem tissue (PMI = 22hrs)

  • Individuals = 1 ("Patient 1568")

  • Source = NICHD Brain and Tissue Bank for Developmental Disorders (Univ Maryland)
  • Age = 51
  • Sex = F
  • Hemisphere = both (these 6 regions are known to have low inter-hemispheric differences)

  • Cells --> see corrected Suppl. Methods online!

    • From paper = 4,488 (total) --> 3,227 (filtered)
    • From README & summary data = 4,039 (total) --> 3,083 (filtered)
  • exprsData (TPM):
    • Lake-2016_Gene_TPM.dat = both exon and intron reads
    • Lake-2016_Exon_TPM.dat = exon-only reads
  • phenoData:
    • For 3,083 annotated single nuclei (4,039 - low mapping outliers and potential doublets)
    • For both Gene and Exon .dat files
    • Neuronal subtypes: excitatory (Ex1-Ex8) and inhibitory (In1-In8)
    • BA origin
    • dbGaP sample name (accession# phs000833.v4.p1; raw data and additional phenoData avail for each sample is available from dbGaP.
  • Recommendations:
    • Use phenoData to subset for only 3,083 quality-filtered data sets
    • Exclude "MT-" genes (mitochondrial) that may have randomly associated with nuclear membrane
  • Publically-availbable online files:
    • Lake-2016_Gene_TPM.dat (exprsData!)
    • Lake-2016_Gene_TPM_Sample-annotation.txt (phenoData!)
    • Lake-2016_Exon_TPM.dat
    • Lake-2016_Exon_TPM_Sample-annotation.txt

Cell breakdown

Lake | FC | PFC | VC | mTC | sTC | aTC | tot. | Linn1 | Markers (not complete)2

Ex1 | 70 | 153 | 62 | 218 | 131 | 424 | 1058 | | CNR+, CUX2+, THSD7A+, CBLN2+
Ex2 | 35 | 32 | 0 | 19 | 6 | 4 | 96 | | CUX2+, RORB+, SYNPR+, THSD7A+, BHLHE22+, CBLN2+
Ex3 | 4 | 15 | 181 | 5 | 10 | 84 | 299 | | CUX2+, RORB+, SYNPR+, THSD7A+, BHLHE22+, SV2C~
Ex4 | 17 | 56 | 0 | 35 | 25 | 41 | 174 | | SLC17A7++, CNR1+, RORB+, SYNPR+, FOXP2+, GABRG1+
Ex5 | 65 | 64 | 11 | 49 | 27 | 34 | 250 | | SLC17A7++, RORB+, KCNK2+, FOXP2+, PCP4+
Ex6 | 42 | 37 | 10 | 31 | 11 | 8 | 139 | | CDH11++, SULF1+, PDE9A+, HTR2C+, SYT6+, FOXP2+, PCP4+, CBLN2+, GRM4+
Ex7 | 21 | 37 | 1 | 29 | 17 | 10 | 115 | | CNR1+, CBLN2+
Ex8 | 8 | 17 | 3 | 7 | 6 | 6 | 47 | | SYNPR++, NR4A2++, OPRK+, PDE9A+, CBLN2+, SLC6A8+, ADRA2A+, NPY1R+

In1 | 34 | 41 | 6 | 17 | 18 | 44 | 160 | | CNR1+, VIP+, RELN+, CCK+ In2 | 2 | 12 | 7 | 7 | 9 | 18 | 55 | Int6 | GAD1++, CNR1+, VIP+, OPRK1+, CCK+
In3 | 10 | 13 | 1 | 14 | 6 | 27 | 71 | | PDE9A+, HTR2C+, VIP+, RELN+, CCK+, CNR1
In4 | 37 | 25 | 9 | 9 | 14 | 27 | 121 | | GAD1++, SV2C++, RELN+, CCK+, KCNH1+, GABRG1+, KCNAB1+, SST~, CNR~
In5 | 22 | 10 | 2 | 11 | 9 | 8 | 62 | | SV2C+, LHX6+, CCK+, KCNAB1+
In6 | 77 | 18 | 16 | 37 | 18 | 73 | 239 | Int3 | GAD1++, PVALB+, SULF1+, PDE9A+, LHX6+, ASIC2+, GRIK3+
In7 | 4 | 6 | 3 | 5 | 10 | 34 | 62 | Int1 | SST+, LHX6+, NPY+, GRIK3+
In8 | 21 | 28 | 16 | 20 | 12 | 38 | 135 | Int2 | SST+, LHX6+, CACNA1G+, RELN~

N | 564 | 328 | 513 | 329 | 880 | 469 | 3083 | |

  • SLC17A7 aka VGLUT1
  • GAD1 aka GAD67
  • 1: Source: FigS12; Linnarsson mouse data had an additional 7 RELN+ subtypes, not shown in FigS12
  • 2: Sources: FigS11, FigS12, FigS16 (red)