README.md

In duohongrui/simutils: Common functions used in data simulation and evaluation

Simutils implements essential utils functions for estimating parameters from real data, simulating and evaluating the simulated single-cell RNA-seq datastets.

Simutils contains many useful utils functions to establish the standard pipeline for defining a simulation method, preprocessing reference datasets with specific requirements, preparing important files for special simulators, evaluating the performance of different simulation methods and so on.

Installation

You can install the development version of simutils from GitHub with:

# install.packages("devtools")
devtools::install_github("duohongrui/simutils")

library(simutils)

Here we demonstrate some useful functions in simutils package:

• Check Python Installation
• Match Cells From Real And Simulated Data
• Format Conversion of Single-Cell Data
• Make Lineage Trees for Single-Cell Data

Check Python Installation

Simutils contains a function for checking the Python installation and environmrnt configuration for PROSSTT method.

simutils::check_python_installation()

## ✔ Python is already installed.

## ✔ Your python version is satisfied.

## ✔ numpy module is installed.

## ✔ scipy module is installed.

## ✔ pandas module is installed.

## ✔ newick module is installed.

## ✔ prosstt module is installed.

Match Cells From Real And Simulated Data

We adopted the Hungarian algorithm to match the cells from reference and simulated datasets. In addition, we also provide an improved Hungarian algorithm.

set.seed(666)
ref_data <- matrix(rpois(10 ^ 6, 2),
                   ncol = 1000,
                   nrow = 1000,
                   dimnames = list(paste0("ref_gene", 1:1000),
                                   paste0("ref_cell", 1:1000)))
set.seed(666)
sim_data <- matrix(rpois(10 ^ 6, 2.5),
                   ncol = 1000,
                   nrow = 1000,
                   dimnames = list(paste0("sim_gene", 1:1000),
                                   paste0("sim_cell", 1:1000)))

match_result <- simutils::match_cells(ref_data = ref_data,
                                      sim_data = sim_data,
                                      t = TRUE,
                                      algorithm = "Hungarian")

## Performing PCA...
## Performing Harmony...
## Calculate correlation matrix...

##

## Match simulated and real cells using Hungarian...
##     reference simulation match_value
## 1  ref_cell77  sim_cell1   0.3809364
## 2 ref_cell777  sim_cell2   0.4301080
## 3  ref_cell45  sim_cell3   0.5085714
## 4 ref_cell732  sim_cell4   0.4225210
## 5 ref_cell353  sim_cell5   0.4255942
## 6 ref_cell818  sim_cell6   0.4712125

head(match_result[["cell_pair"]][order(match_result[["cell_pair"]]$match_value, decreasing = TRUE), ])

##       reference  simulation match_value
## 921 ref_cell487 sim_cell921   0.6337095
## 20  ref_cell213  sim_cell20   0.6115246
## 186 ref_cell775 sim_cell186   0.5933733
## 650 ref_cell495 sim_cell650   0.5933733
## 70   ref_cell62  sim_cell70   0.5922209
## 840 ref_cell660 sim_cell840   0.5893397

We can also use improved Hungarian algorithm:

match_result2 <- simutils::match_cells(ref_data = ref_data,
                                       sim_data = sim_data,
                                       t = TRUE,
                                       algorithm = "Improved_Hungarian")

## Performing PCA...
## Performing Harmony...
## Calculate correlation matrix...
## Match simulated and real cells using improved Hungarian...
##   reference  simulation match_value
## 1 ref_cell1 sim_cell427   0.4751501
## 2 ref_cell2 sim_cell851   0.5117407
## 3 ref_cell3 sim_cell125   0.5028091
## 4 ref_cell4 sim_cell935   0.4354862
## 5 ref_cell5 sim_cell111   0.4327971
## 6 ref_cell6 sim_cell671   0.4223289

head(match_result2[["cell_pair"]][order(match_result2[["cell_pair"]]$match_value, decreasing = TRUE), ])

##       reference  simulation match_value
## 487 ref_cell487 sim_cell921   0.6337095
## 213 ref_cell213  sim_cell20   0.6115246
## 495 ref_cell495 sim_cell650   0.5933733
## 775 ref_cell775 sim_cell186   0.5933733
## 62   ref_cell62  sim_cell70   0.5922209
## 660 ref_cell660 sim_cell840   0.5893397

Format Conversion of Single-Cell Data

Simsite provides the function of converting single-cell data formats from SingleCellExperimental to Seurat, list and h5ad.

Here we construct a data with SingleCellExperimental format:

library(SingleCellExperiment)

set.seed(111)
data <- matrix(rpois(10 ^ 6, 2),
               ncol = 1000,
               nrow = 1000,
               dimnames = list(paste0("ref_gene", 1:1000),
                               paste0("ref_cell", 1:1000)))
SCE <- SingleCellExperiment::SingleCellExperiment(list(counts = data),
                                                  colData = data.frame("cell_name" = colnames(data)),
                                                  rowData = data.frame("gene_name" = rownames(data)))
SCE

## class: SingleCellExperiment 
## dim: 1000 1000 
## metadata(0):
## assays(1): counts
## rownames(1000): ref_gene1 ref_gene2 ... ref_gene999 ref_gene1000
## rowData names(1): gene_name
## colnames(1000): ref_cell1 ref_cell2 ... ref_cell999 ref_cell1000
## colData names(1): cell_name
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):

To Seurat object:

Seurat <- simutils::data_conversion(SCE_object = SCE, return_format = "Seurat")
Seurat

## An object of class Seurat 
## 1000 features across 1000 samples within 1 assay 
## Active assay: originalexp (1000 features, 0 variable features)

To h5ad file (and you will get a path at which the file locates):

h5ad <- simutils::data_conversion(SCE_object = SCE, return_format = "h5ad")

## Creating h5Seurat file for version 3.1.5.9900

## Adding counts for originalexp

## Adding data for originalexp

## No variable features found for originalexp

## Adding feature-level metadata for originalexp

## Validating h5Seurat file

## Adding data from originalexp as X

## Transfering meta.features to var

## Adding counts from originalexp as raw

## Transfering meta.features to raw/var

## Transfering meta.data to obs

## Your data has been save to /var/folders/1l/xmc98tgx0m37wxtbtwnl6h7c0000gn/T//Rtmp9kqkJv/20230529193826.h5ad

h5ad

## $file_type
## [1] "h5ad"
## 
## $save_path
## [1] "/var/folders/1l/xmc98tgx0m37wxtbtwnl6h7c0000gn/T//Rtmp9kqkJv/20230529193826.h5ad"

Make Lineage Trees for Single-Cell Data

The tree lineage of cell clusters within single-cell data can be traced by make_trees function in simutils.

set.seed(555)
data <- cbind(matrix(rpois(2.5e5, 2), ncol = 250, nrow = 1000),
              matrix(rpois(2.5e5, 4), ncol = 250, nrow = 1000),
              matrix(rpois(2.5e5, 8), ncol = 250, nrow = 1000),
              matrix(rpois(2.5e5, 12), ncol = 250, nrow = 1000))
colnames(data) <- paste0("Cell", 1:ncol(data))
rownames(data) <- paste0("Gene", 1:nrow(data))

Newick format:

newick <- simutils::make_trees(ref_data = data,
                               is_Newick = TRUE,
                               is_parenthetic = FALSE,
                               return_group = TRUE)

## Loading required package: amap

## Computing nearest neighbor graph

## Computing SNN

## Your data has 3 groups

newick$phyla

## [1] "(group2:3.5470224281145,(group1:5.70650677442804,group3:5.70650677442804):3.5470224281145);"

Phylo format:

phylo <- simutils::make_trees(ref_data = data,
                              is_Newick = FALSE,
                              is_parenthetic = TRUE,
                              return_group = TRUE)

## Computing nearest neighbor graph

## Computing SNN

## Your data has 3 groups

str(phylo$phyla)

## List of 1
##  $ :List of 4
##   ..$ edge       : int [1:4, 1:2] 4 4 5 5 1 5 2 3
##   ..$ edge.length: num [1:4] 3.55 3.55 5.71 5.71
##   ..$ Nnode      : int 2
##   ..$ tip.label  : chr [1:3] "group2" "group1" "group3"
##   ..- attr(*, "class")= chr "phylo"
##   ..- attr(*, "order")= chr "cladewise"

duohongrui/simutils documentation built on March 12, 2024, 8:40 p.m.