ARMET-tc infers rates of changes in tissue composition across a covariate of interest (e.g., treatment status, time or disease grade).
Input (see below):
- Data frame of tissue gene counts
- Design matrix
- Covariate of interest
$mix
# A tibble: 24,507 x 16
gene `TCGA-CC-A8HT` `TCGA-DD-AACZ` `TCGA-DD-A73G` `TCGA-DD-A3A5`
<fct> <dbl> <dbl> <dbl> <dbl>
1 TSPAN6 2984 8280 3508 2783
2 TNMD 0 0 1.00 3.00
3 DPM1 1336 1751 521 1548
4 SCYL3 565 650 307 301
5 C1orf112 550 348 73.0 186
6 FGR 538 830 62.0 90.0
7 CFH 15926 38744 70886 42886
8 FUCA2 7318 9106 4212 2153
9 GCLC 3799 5108 582 2776
10 NFYA 1593 1633 307 333
# ... with 24,497 more rows, and 11 more variables: `TCGA-DD-A4ND` <dbl>,
# `TCGA-DD-AACX` <dbl>, `TCGA-DD-AAVY` <dbl>, `TCGA-DD-AAVV` <dbl>,
# `TCGA-DD-AAVR` <dbl>, `TCGA-DD-AAW1` <dbl>, `TCGA-DD-AAVZ` <dbl>,
# `TCGA-DD-A4NV` <dbl>, `TCGA-UB-A7MB` <dbl>, `TCGA-DD-A4NK` <dbl>,
# `TCGA-DD-AAD6` <dbl>
$design_matrix
# A tibble: 15 x 4
sample `(Intercept)` relapse age
<fct> <dbl> <dbl> <dbl>
1 TCGA-CC-A8HT 1.00 1.00 27334
2 TCGA-DD-AACZ 1.00 1.00 23230
3 TCGA-DD-A73G 1.00 0 26949
4 TCGA-DD-A3A5 1.00 1.00 24288
5 TCGA-DD-A4ND 1.00 1.00 20782
6 TCGA-DD-AACX 1.00 1.00 24329
7 TCGA-DD-AAVY 1.00 0 20709
8 TCGA-DD-AAVV 1.00 0 20751
9 TCGA-DD-AAVR 1.00 0 16157
10 TCGA-DD-AAW1 1.00 0 20288
11 TCGA-DD-AAVZ 1.00 0 14005
12 TCGA-DD-A4NV 1.00 0 22438
13 TCGA-UB-A7MB 1.00 1.00 8951
14 TCGA-DD-A4NK 1.00 1.00 29244
15 TCGA-DD-AAD6 1.00 1.00 24185
$cov_to_test
[1] "relapse"
Output:
- Matrix of proportions of cell type for each provided sample
- Coefficients of change (slope)
ARMET-tc is tipically used with a design matrix as you would use a linear model of differential expression estimator (e.g., Limma or edgeR)
ARMET-tc can be also uses without any design matrix, in this case the only output would be the matrix of proportions of cell type for each provided sample.
IMPORTANT: in the latter case the algorithm assumes that the query experiment includes homogeneous samples. That is, coming from similar tissue. For example, querying samples of blood together is a good choice, whereas querying samples of pure macrophages, together with samples of pure t-cells is not a good design. In the latter case those two groups should be queried separately.
# Installation
# For a fresh Ubuntu system
###############################################
# Install R/3.5.1
###############################################
sudo apt install apt-transport-https software-properties-common
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
sudo add-apt-repository 'deb https://cloud.r-project.org/bin/linux/ubuntu bionic-cran35/'
sudo apt update
sudo apt-get install r-base
sudo apt-get install libssl-dev libcurl4-openssl-dev libxml2-dev
# For an initialised machine
###############################################
# Dependencies
###############################################
For non C++14 native machines
1) create ~/.R/Makevars
2) write in it
CXX14 = g++ # or clang++ if you have that
CXX14FLAGS = -O3 -Wno-ignored-attributes -fPIC
###############################################
# Dependencies
###############################################
library(devtools)
source("https://bioconductor.org/biocLite.R")
biocLite("limma")
biocLite("edgeR")
install.packages("data.tree")
install.packages("abind")
install_github("mjskay/tidybayes")
install.packages("future")
install.packages("reshape")
install.packages("StanHeaders")
install.packages("rstan")
###############################################
install_github("stemangiola/ARMET", args = "--preclean", build_vignettes = FALSE)
if("package:ARMET" %in% search()) detach("package:ARMET", unload=TRUE, force=TRUE)
library(ARMET)
# Inference
results =
ARMET_tc(
mix = test_data$mix,
my_design = test_data$design_matrix,
cov_to_test = test_data$cov_to_test,
do_debug = F
)
# Report
ARMET_getFit(results)
ARMET_plotFit(results, "immune_cell")
ARMET_plotPolar(results)
# Input data shape
test_data
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent