README.md

In johannesnicolaus/ASURAT_source: Assisting Interpretable Classifications in Single-cell Transcriptomics

ASURAT

ASURAT is a single-cell RNA sequencing (scRNA-seq) data analysis pipeline, developed for simultaneously clustering cells and biological interpretation.

Introduction, documentation, and tutorial can be found at

https://keita-iida.github.io/ASURAT/index.html

Installation

Use the devtools::install_github() command

devtools::install_github('johannesnicolaus/ASURAT_source')

ASURAT's workflow

1. Preprocessing (data quality control, normalization, etc.)
2. Collecting databases (DBs), but the following DBs were downloaded in December 2020: Disease Ontology (DO), Cell Ontology (CO), and Gene Ontology (GO) DB, Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome.
3. Creating signs (sign is a general biological term representing cell type and various biological functions)
4. Creating sign-by-sample matrices (SSMs)
5. Unsupervised clustering of cells
6. Finding significant signs
7. Multiple sing analysis (analyzing SSMs from the viewpoints of cell type, biological process, signaling pathway, etc.)

Quick start inputting a Seurat object

Although the above URL does not assume Seurat-based analyses, it is beneficial to begin with a Seurat object obj including obj@assays[["RNA"]]@counts data.

Load a Seurat object for human scRNA-seq data (below is an example).

cerv_seurat <- readRDS(file = "backup/cerv_small_seurat.rds")  # Seurat object

Below are stopgap installations. See Chapter 1 for all the requirements.

Note that users need to replace org.Hs.eg.db with other packages when analyzing other animal's scRNA-seq data.

library(tidyverse)                # For efficient handling of data.frame
library(org.Hs.eg.db)             # For using human genome annotation package
library(Seurat)                   # For using Seurat
source("R/function_general.R")      # ASURAT's function

Create an ASURAT object.

cerv <- make_asurat_obj(mat = cerv_seurat@assays[["RNA"]]@counts,
                        obj_name = "cerv_small")

Convert gene symbols into Entrez IDs by using org.Hs.eg.db package.

dictionary <- AnnotationDbi::select(org.Hs.eg.db,
                                    key = cerv[["variable"]][["symbol"]],
                                    columns = c("ENTREZID"), keytype = "SYMBOL")
dictionary <- dictionary[!duplicated(dictionary$SYMBOL), ]
names(dictionary) <- c("symbol", "entrez")
cerv[["variable"]] <- dictionary

The following function log1p_data() performs log transform of the input data with a pseudo count eps.

log1p_data <- function(obj, eps){
  obj[["history"]][["log1p_data"]][["eps"]] <- eps
  mat <- as.matrix(obj[["data"]][["raw"]])
  lmat <- log(mat + eps)
  obj[["data"]][["log1p"]] <- as.data.frame(lmat)
  return(obj)
}

cerv <- log1p_data(obj = cerv, eps = 1)

The following function centralize_data() centralizes the input data on a gene-by-gene basis.

centralize_data <- function(obj){
  mat <- as.matrix(obj[["data"]][["log1p"]])
  cmat <- sweep(mat, 1, apply(mat, 1, mean), FUN = "-")
  obj[["data"]][["centered"]] <- as.data.frame(cmat)
  return(obj)
}

cerv <- centralize_data(obj = cerv)

The following function do_cor_variables() computes a correlation matrix from the input data. Users can choose a measure of correlation coefficient by setting method (vector form is also accepted but not recommended due to the file size) such as pearson, spearman, and kendall.

do_cor_variables <- function(obj, method){
  res <- list()
  tmat <- t(obj[["data"]][["log1p"]])
  for(m in method){
    res <- c(res, list(cor(tmat, method = m)))
  }
  names(res) <- method
  return(res)
}

cerv_cor <- do_cor_variables(obj = cerv, method = c("spearman"))

Save the objects. Please note that the suffixes of the following filenames, such as 09 and 005, are only for identifying the computational steps (there is no special significance).

saveRDS(cerv, file = "backup/09_005_cerv_correlation.rds")
saveRDS(cerv_cor, file = "backup/09_006_cerv_correlation.rds")

ASURAT using several databases

Go to Chapter 8 for analyses using Disease Ontology database.

Go to Chapter 9 for analyses using Cell Ontology database.

Go to Chapter 10 for analyses using Gene Ontology database.

Go to Chapter 11 for analyses using Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

Go to Chapter 12 for analyses using Reactome database.

johannesnicolaus/ASURAT_source documentation built on Dec. 21, 2021, 2:11 a.m.