CASi
CASi is a comprehensive framework for analyzing longitudinal
scRNA-seq data, which provides users with: (1) cross-time points cell
annotation, (2) detection of potentially novel cell types emerged over
time, (3) visualization of cell population evolution, and (4)
identification of temporal differentially expressed genes (tDEGs). This
vignette introduces the functions of the CASi package and shows how it
analyzes longitudinal scRNA-seq data. CASi was created by Yizhuo Wang,
Ziyi Li and Xuelin Huang, and is now maintained by Yizhuo Wang.
Install CASi using the code below.
# install.packages("CASi")
library(CASi)
Make sure that Python is installed to load all required
dependencies.
CASi has four functions in total. A simulated
longitudinal scRNA-seq dataset is used here to demonstrate the usages of
these two functions. We obtain a publicly available dataset of
peripheral blood mononuclear cells (PBMC)[1], containing more than
60,000 sorted cells from eight immune cell types. We randomly extract
cells from five cell types to build the simulation data. The demo pbmc
data, “pbmc_demo”, has been lazy-loaded with the package. It is a list
of four items: * 1. train_count: used as The initial time point’s gene
expression. * 2. train_label: used as The initial time point’s cell
label. * 3. test_count: used as following time points’ gene expression.
* 4. test_label: the following time points’ cell labels, which is not
known in reality. * 5. Meta data: contains Group and Time
information.
data(pbmc_demo)
names(pbmc_demo)
#> [1] "train_count" "train_label" "test_count" "test_label"
Create a CASi list
The first step is to create a CASi object using the data. The
longitudinal scRNA-seq should be seperated to two subsets regardless of
how many time points you have, one includes the initial time point’s
data (\(t_{0}\)) and one includes all
the following time points’ data (\(t_{1}\)). This is because we always want to
use only the initial time point’s data as the training data to prevent
any potential bias during the training process.
The labels/cell types of \(t_{0}\)
data is optional. If you do not provide CASi with the cell labels, CASi
will first perform an unsupervised clustering for unlabeled data and use
the clustering number (e.g., C1, C2, C3…) as the cell type labels. Note
that CASi will automatically select a smaller set of informative
features/genes for computational efficiency. Below are the input
parameters of the CreatCASiList function:
- t0_count: Gene expression matrix of the initial data point/t0.
- t1_count: Gene expression matrix of the following data
points/t1.
- t0_label: Cell labels of the initial data point.
- nfeature: The number of top highly variable genes. Default to
2000.
l1 <- CreateCASiList(t0_count = as.matrix(pbmc_demo[[1]]), t1_count = as.matrix(pbmc_demo[[3]]), t0_label = pbmc_demo[[2]])
The CASi list should contain three elements:
names(l1)
#> [1] "Data: t0 cells gene expression" "Data: t0 cell types"
#> [3] "Data: t1 cell gene expression"
Cross timepoints annotation
CASi uses an artificial neural network built on Keras to annotate
cells of \(t_{1}\) data. The Keras R
interface uses the TensorFlow backend engine by default. So please make
sure you have installed the Tensorflow package properly. Below are the
input parameters of the CrossTimeAnnotation function:
- CASiList: The CASi list.
- DrawUMAP: Draw UMAP plots per time points and cell labels. Default
to TRUE.
l1 <- CrossTimeAnnotation(l1)
The CASi list after this function should contain six elements:
names(l1)
#> [1] "Data: t0 cells gene expression" "Data: t0 cell types"
#> [3] "Data: t1 cell gene expression" "Neural network: setting"
#> [5] "Neural network: training history" "Data: t1 cell types"
Detect novel cells
The next step of CASi is to identify any novel cell types that have
emerged over time. When data of interest is collected from multiple time
points’ samples, it is highly possible that the some cells such as tumor
cells will differentiate and new cell types (e.g., a distinct subclone
of tumor cells) might appear. This step helps to distinguish those new
cell types from existing cell types. For the DetectNovelCells function,
we have the following parameters:
- CASiList: The CASi list.
- corr.cutoff: The cutoff correlation value to determine novel cells.
Default to 0.4. Cells with the correlation larger than 0.4 will not be
considered novel cells.
- p.cutoff: The cutoff p value of t-tests to determine novel cells.
Default to 0.01. Cells with the p value smaller than 0.01 will be
considered novel cells.
- DrawUMAP: Draw UMAP plots per cell type-specific correlation and
cell labels.
l1 <- DetectNovelCells(l1)
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 362
#> Number of edges: 18353
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 362
#> Number of edges: 18353
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9913
#> Number of communities: 2
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.5990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.5890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.5790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 307
#> Number of edges: 17256
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.5700
#> Number of communities: 2
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.8090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7390
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7290
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7190
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.7090
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6490
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 433
#> Number of edges: 23692
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.6416
#> Number of communities: 2
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9990
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9890
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9790
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9690
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9590
#> Number of communities: 1
#> Elapsed time: 0 seconds
#> Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
#>
#> Number of nodes: 396
#> Number of edges: 23383
#>
#> Running Louvain algorithm...
#> Maximum modularity in 10 random starts: 0.9495
#> Number of communities: 2
#> Elapsed time: 0 seconds
The CASi
list after this function should contain seven elements:
names(l1)
#> [1] "Data: t0 cells gene expression"
#> [2] "Data: t0 cell types"
#> [3] "Data: t1 cell gene expression"
#> [4] "Neural network: setting"
#> [5] "Neural network: training history"
#> [6] "Data: t1 cell types"
#> [7] "Data: t1 cell types (potential novel cells)"
Find temporal differentially expressed genes (tDEGs)
This step performs differential analysis tailored to longitudinal
scRNA-seq data. We combine a generalized linear model with iterative
feature selection to select genes that have apparent
increasing/decreasing behavior over time and genes that behave
differently along time in different groups. Users of this function must
provide the metadata of two variables: ‘Group’, which should be a binary
factor variable, and ‘Time’, which can be either numerical or
categorical variable. Below are the input parameters of the
FindTemporalDEGs function:
- CASiList: The CASi list.
- metadata The dataframe containing the Group variable and the Time
variable.
- pct.cutoff The percentage of cells that the gene is expressed.
Default to 0.3.
- p.cutoff The cutoff p value of t-tests to determine tDEGs. Default
to 0.05. Genes with the p value smaller than 0.05 will be considered
tDEGs.
To demonstrate this function, we manually adjusted the previous
simulation dataset by adding cell-type-specific time/group effect. The
demonstration list, “CASiList_tDEG_demo” and its corresponding metadata,
“metadata_tDEG_demo” have been been lazy-loaded with the package. As
shown below, the metadata should contain a Group variable and a Time
variable, and the row number should match the total cell number of your
data.
data("CASiList_tDEG_demo")
data("metadata_tDEG_demo")
summary(metadata_tDEG_demo)
#> Group Time
#> Length:3097 Length:3097
#> Class :character Class :character
#> Mode :character Mode :character
nCell_t0 <- length(CASiList_tDEG_demo[[2]])
nCell_t1 <- length(CASiList_tDEG_demo[[6]])
nCell_t0 + nCell_t1
#> [1] 3097
l2 <- FindTemporalDEGs(CASiList_tDEG_demo, metadata_tDEG_demo)
The CASi list after this function should contain eight elements:
names(l2)
#> [1] "Data: t0 cells gene expression"
#> [2] "Data: t0 cell types"
#> [3] "Data: t1 cell gene expression"
#> [4] "Neural network: setting"
#> [5] "Neural network: training history"
#> [6] "Data: t1 cell types"
#> [7] "Data: t1 cell types (potential novel cells)"
#> [8] "Cell-type-specific p-value of temporal DEGs"