MGPfactR

A model-based, unsupervised manifold learning method that factors complex cellular trajectories into interpretable bifurcating Gaussian processes of transcription. The complete functionality of MGPfact is accessible in MGPfactR, enabling the discovery of specific biological determinants of cell fate.

Factorized Trajectory

Consensus Trajectory

installation

(1) Install julia environment

JULIA_VERSION=1.6.6
sudo wget https://julialang-s3.julialang.org/bin/linux/x64/$(echo $JULIA_VERSION | cut -d. -f 1-2)/julia-$JULIA_VERSION-linux-x86_64.tar.gz \
    && tar -xvzf julia-$JULIA_VERSION-linux-x86_64.tar.gz -C ./ 
echo 'export PATH=$PATH:'"$(pwd)"/julia-$JULIA_VERSION/bin >> ~/.bashrc
rm julia-$JULIA_VERSION-linux-x86_64.tar.gz

(2) Install MGPfact.jl and its dependencies

julia -e 'import Pkg; Pkg.add(url="https://github.com/renjun0324/MGPfact.jl")'
julia -e 'import Pkg; Pkg.add(["Mamba", "RData", "JLD2", "Distributions", "KernelFunctions"])'
julia -e 'import Pkg; Pkg.add(Pkg.PackageSpec(name="RCall", version="0.13.15"))'
julia -e 'import Pkg; Pkg.add(Pkg.PackageSpec(name="Suppressor", version="0.2.6"))'

# Test whether MGPfact.jl can be loaded
using MGPfact
using Mamba, RData, JLD2

(3) Install MGPfactR packages

R -e 'devtools::install_version("JuliaCall","0.16")'
R -e 'devtools::install_github("renjun0324/MURP@v0.6.5")'
R -e 'devtools::install_cran(c("dplyr", "purrr", "stringr","JuliaCall", "pbmcapply", "doParallel", "reshape", "reshape2", "igraph", "graphlayouts","oaqc","parallelDist"))'

# Test whether MGPfactR can be loaded
library(MGPfactR)

# Test if the R environment can be linked with the Julia environment
library(JuliaCall)
julia_home = gsub("/julia$","",system("which julia", intern = T))
julia_setup(JULIA_HOME=julia_home)

quick start

For the detailed usage process of MGPfact, please click here

data(fibroblast_reprogramming_treutlein)
data = fibroblast_reprogramming_treutlein
counts = data$counts
cell_info = data$cell_info
rownames(cell_info) = cell_info$cell_id
expression = LogNormalize(t(counts)) %>% t

# create object
ct <- CreateMGPfactObject(data_matrix = expression, MetaData = cell_info)

# downsampling
ct = MURPDownsampling(ct, omega = 0.9, iter = 10, seed = 723, fast = T, cores = 1,
                      pca.center = FALSE, pca.scale = FALSE, plot = T, max_murp = 20)
ct = GetMURPMapLabel(ct, labels = "time_point")

# initialize parameters
SaveMURPDatToJulia(ct, murp_pc_number = 3)
ct = SetSettings(ct, murp_pc_number = 3, trajectory_number = 3, pse_optim_iterations = 100, start_murp = 999)

# forcast pseudotime
ct = RunningmodMGPpseudoT(ct, julia_home = julia_home, cores = 1)

# trahectory construction
ct <- GetIterCor(ct, iteration_list = list(c(1, getParams(ct, "pse_optim_iterations"))))
ct <- GetPredT(object = ct, chains = 1:getParams(ct, "chains_number"))
ct <- GetPseSdf(ct)
ct <- GetBinTree(object = ct)
ct <- GetTbTree(object = ct)
ct <- GetTbTreeAllpoint(object = ct, save = T, labels = getParams(ct,"label"))
PlotPieBinLabel(ct, labels = getParams(ct,"label"))
PlotPieTbLabel(ct, labels = getParams(ct,"label"))
PlotPieConsensusMainLabel(ct, labels = getParams(ct,"label"))
PlotPieConsensusAllLabel(ct, labels = getParams(ct,"label"),size = 0.005)