scTenifoldNet

scTenifoldNet: A Machine Learning Workflow for Constructing and Comparing Transcriptome-wide Gene Regulatory Networks from Single-Cell Data

See Patterns paper for more information.

scTenifoldNet in R

Installation

In an R shell, type:

install.packages("scTenifoldNet")

Basic Use

library(scTenifoldNet)
library(Matrix)
X <- read.csv('X1.csv', header = FALSE)
colnames(X) <- paste0('X1_', seq_len(ncol(X)))
X <- as.matrix(X)
Y <- read.csv('X2.csv', header = FALSE)
colnames(Y) <- paste0('X2_', seq_len(ncol(Y)))
Y <- as.matrix(Y)
rownames(X) <- rownames(Y) <- readLines('genelist.csv')
set.seed(1)
DR <- scTenifoldNet(X = X, Y = Y)
save(DR, file = 'netOut.RData')
write.csv(DR$diffRegulation, row.names = FALSE, file = 'netResult.csv')

scTenifoldNet in MATLAB

Installation

Run the following code in MATLAB:

tic
disp('Installing scTenifoldNet...')
unzip('https://github.com/cailab-tamu/scTenifoldNet/archive/master.zip');
addpath('./scTenifoldNet-master/MATLAB');
toc
if exist('sctenifoldnet_m.m','file')
    disp('scTenifoldNet installed!')
end

Basic Use

Run ‘SCTENIFOLDNET` with an example data file clean_data_1Ctl_2FgF2.mat in MATLAB:

load clean_data_1Ctl_2FgF2.mat
sce=sce.selectgenes(1,0.15);
sce=sce.qcfilter;

%%
X0=sce.X(:,sce.c_batch_id==1);
X1=sce.X(:,sce.c_batch_id==2);
T=sctenifoldnet(X0,X1,sce.g,'savegrn',true);
writetable(T,'resT.txt');
Tr=e_fgsearun(T);
writetable(Tr,'resTr.txt');
tgenes=T.genelist(T.pAdjusted<0.1);
e_fgseanet(Tr);

load(ls('A0_*.mat'))
[y,i]=isemember(tgenes,genelist);
assert(all(y))
a0=A0(i,i);

load(ls('A1_*.mat'))
[y,i]=isemember(tgenes,genelist);
assert(all(y))
a1=A1(i,i);

g1=digraph(a1,tgenes,'omitselfloops');
g2=digraph(a2,tgenes,'omitselfloops');
gui.i_doublegraphs(g1,g2);

scTenifoldNet in Julia

Installation

Run the following code in Julia:

using Pkg
Pkg.add(PackageSpec(url="git://github.com/jamesjcai/ScTenifold.jl.git"))
Pkg.test("ScTenifold")
# or
# ] add https://github.com/jamesjcai/ScTenifoldNet.jl

Basic Use

Here is a simple example using randomly generated data.

using ScTenifold
using DelimitedFiles
# cd(dirname(@__FILE__))

X1=rand(100,1000);
X2=copy(X1);
X2[4,:].=0.0;
@time d,fc,p,adjp=ScTenifold.sctenifoldnet(X1,X2,donorm=false)

open("output_small.txt", "w") do io
    writedlm(io, [d fc p adjp])
end

Exported Functions

Code	Function
pcnet	Computes a gene regulatory network based on principal component regression
tensordecomp	Performs CANDECOMP/PARAFAC (CP) Tensor Decomposition
manialn	Performs non-linear manifold alignment of two gene regulatory networks
drgenes	Evaluates gene differential regulation based on manifold alignment distances
tenrnet	Subsamples cells, constructs single-cell gene regulatory networks (scGRNs) using principal component regression (pcnet), and denoises scGRNs using tensor decomposition (tensordecomp).

Loading ScTenifoldNet

Once installed, ScTenifoldNet.jl can be loaded typing:

using ScTenifoldNet

Simulating of a dataset

Here we simulate a dataset of 2000 cells (columns) and 100 genes (rows) following the negative binomial distribution with high sparsity (~67%).

d=NegativeBinomial(20,0.98)
X=rand(d,100,2000)

Generating a perturbed network

We generate a perturbed network modifying the expression of genes 10, 2, and 3 and replacing them with the expression of genes 50, 11, and 5.

Y=copy(X)
Y[10,:]=Y[50,:]
Y[2,:]=Y[11,:]
Y[3,:]=Y[5,:]

X=X[:,vec(sum(X,dims=1).>30)]
Y=Y[:,vec(sum(Y,dims=1).>30)]

Generating networks

Here we run ScTenifoldNet under the H0 (there is no change in the regulation of the gene) using the same matrix as input and under the HA (there is a change in the regulation of the genes) using the control and the perturbed network.

Z0=ScTenifoldNet.tenrnet(X, donorm=true)
Z1=ScTenifoldNet.tenrnet(Y, donorm=true)

Differential regulation based on manifold alignment distances

As is shown below, under the H0, none of the genes shown a significative difference in regulatory profiles using an FDR cut-off of 0.1, but under the HA, the 6 genes involved in the perturbation (50, 11, 2, 10, 5, and 3) are identified as perturbed.

d,aln0,aln1=ScTenifoldNet.manialn(Z0,Z1)
fc,p,adjp=ScTenifoldNet.drgenes(d)

Plotting the results

Results can be easily displayed using quantile-quantile plots.

using StatsPlots, Distributions
x=rand(Chisq(1), length(fc))
qqplot(x, fc)