Gene co-expression network analysis: Example in Apple

Description

In this project, our aim is the creation of co-expression networks for gene correlation analysis using public apple RNA datasets. We will examine modular relationships among genes across various functional domains and tissue types.

Downloading a genome reference

First, we need to select and download the apple reference genome in which we will do the mapping of the raw reads.

I downloaded the apple genome published by Sun, et al.,2020.

We need the genome FASTA file, annotation GFF file, and blast2go file for functional analysis.

mkdir Genome
wget http://bioinfo.bti.cornell.edu/ftp/Apple_genome/genome/haploid/Gala/Gala_haploid_v2.chr.fa.gz
wget http://bioinfo.bti.cornell.edu/ftp/Apple_genome/genome/haploid/Gala/Gala_haploid_v2.blast2go.gz
wget http://bioinfo.bti.cornell.edu/ftp/Apple_genome/genome/haploid/Gala/Gala_haploid_v2.gff.gz

Use gunzip to unzip the files.

It is important to index the genome according the mapper that it will be used. Here, we use STAR.

#!/bin/sh
# Slurm directives:
#SBATCH --cpus-per-task 4
#SBATCH --mem-per-cpu 4G
#SBATCH --time 1-5:10
#SBATCH --output=logs/STAR.out
#SBATCH --error=logs/STAR.err
#SBATCH --job-name=STAR
# Generate the index genome. Make sure you have downloaded all the necessary files from the repository.
mkdir -p Genome/index
cd Genome || exit  # Change directory or exit if it fails
STAR --runThreadN 16 --runMode genomeGenerate --genomeDir ~/Genome/index --genomeFastaFiles Genome/GalaChrs.fasta --sjdbGTFfile Genome/Gala_haploid_v2.gff --sjdbOverhang 100 --genomeSAindexNbases 12
cd ..

Downloading RNAseq samples from NCBI

You can select which type of date is of your interest. For this example, I select samples from different apple tissues. I create an txt file with the number accession RNA_samples.txt. You can find the description of each sample in Samples_NCBI_info.xlsx.

Dowbloading the samples using a loop script:

#!/bin/sh
# Slurm directives:
#SBATCH --cpus-per-task 4
#SBATCH --mem-per-cpu 4G
#SBATCH --time 1-5:10
#SBATCH --output=logs/NCBIdownload.out
#SBATCH --error=logs/NCBIdownload.err
#SBATCH --job-name=NCBIdownloading
# Define the input file with the names of RNAseq samples
SAMPLES=~/RNA_samples_names.txt  # txt file with the names of the RNA sample accessions from NCBI

# Download and convert SRA data to FASTQ format
while read -r RNA_samples
do
    ${HOME}/sratoolkit.3.0.7-centos_linux64/bin/fastq-dump --split-files "$RNA_samples"
done < "$SAMPLES"

Create directory for FASTQ files if it doesn't exist

mkdir -p Fastq

Move downloaded files to Fastq directory

mv SRR* Fastq/

STAR alignments

Now we will align FASTQ files using STAR.

SAMPLES=~/RNA_samples_names.txt 
mkdir -p sorted_bam #create the output directory for the bam files
cd sorted_bam || exit  # Change directory or exit if it fails
while IFS= read -r RNA_samples
do
    STAR --genomeDir ~/Genome/index \
         --runThreadN 8 \
         --sjdbGTFfile ~/Genome/Gala_haploid_v2.gtf \
         --readFilesIn ~/Fastq/"$RNA_samples"_1.fastq ~/Fastq/"$RNA_samples"_2.fastq \
         --quantMode TranscriptomeSAM GeneCounts \
         --outSAMstrandField intronMotif \
         --outSAMtype BAM SortedByCoordinate \
         --outSAMattributes NH HI AS NM MD \
         --outFilterIntronMotifs RemoveNoncanonical \
         --outFileNamePrefix "${RNA_samples}noncanonical"
done < "$SAMPLES"

It is important to erase unnecessary files to save space.

rm *.sam

The output files that we will use for the transcript quantification are the sorted bam files.

e. g. SRR26729830noncanonicalAligned.toTranscriptome.out.bam

Salmon quantification

Now, we are going to quantify transcript abundance using Salmon.

#!/bin/sh
# Slurm directives:
#SBATCH --cpus-per-task 4
#SBATCH --mem-per-cpu 4G
#SBATCH --time 1-5:10
#SBATCH --output=logs/salmon.out
#SBATCH --error=logs/salmon.err
#SBATCH --job-name=salmon
SAMPLES=~/RNA_samples_names.txt 
mkdir -p salmon_quant
cd salmon_quant || exit  # Change directory or exit if it fails
while IFS= read -r RNA_samples
do
    salmon quant -t ~/Genome/transcripts.apple.fa \
                 -l A \
                 -a ~/sorted_bam/"${RNA_samples}noncanonical"Aligned.toTranscriptome.out.bam \
                 -o "${RNA_samples}noncanonical"
done < "$SAMPLES"

We need to merge all the quantification files per accession into one file salmon_merge.txt to be used as input for WGCNA in R.

salmon quantmerge --quants SRR*noncanonical -o salmon_merge.txt

WGCNA in R

To generate the modules and cluster of genes, we will work in R with the package WGCNA.

This example was created under the example of jashapiro Fix WGCNA url (#477). Here the link of the original script: https://github.com/AlexsLemonade/refinebio-examples/blob/staging/04-advanced-topics/network-analysis_rnaseq_01_wgcna.Rmd

First download all the libraries:

require(devtools)
library(dplyr)
library(tidyverse)     
library(magrittr)      
library(WGCNA)
library(ggplot2)
library(tidyverse)
library(cowplot)
library(patchwork)
library(DESeq2)
library(igraph)

Then we are using the based on an example of WGCNA. Run the script apple_WGCNA.R

Here is an example of the outputs of the WGCNA analysis.

From the main WGCNA results, we have a file apple_wgcna_gene_allgenes_to_module.tsv from where we can find all the modules that were created. We can choose a specific Module for a downstream analysis and plots.

Let’s make a plot of module 27.

Then, we can create a summary heatmap of a given module, here module 27. It is using the full gene expression matrix and by default is `normalized_counts'.

We can extract this information and the correlation of the important genes which where clasified in the "module 27"

LISTOFGENES <- gene_module_key[gene_module_key$module == "ME27", 1] #Select genes from gene_module_key where the module is "ME27".
SUBNET <- normalized_counts[, colnames(normalized_counts) %in% LISTOFGENES$gene] #Create a subset of normalized_counts based on the selected genes
CR <- cor((SUBNET)) #Calculate the correlation matrix
DF <- as.data.frame(as.table(CR)) #Convert the correlation matrix into a data frame
M27_Correlations <- DF[DF$Freq > 0.7 | DF$Freq < -0.7,] #Select rows from the data frame where the absolute correlation coefficient is greater than 0.7
write.table(M27_Correlations, file = "FM27corrleations", row.names = FALSE)

Here we use a strong filter of a correlation -0.7 ; 0.7.

Var1	Var2	Freq
Mdg_02g014990-mRNA1	Mdg_15g019970-mRNA1	-0.865905603
Mdg_03g005210-mRNA1	Mdg_02g014990-mRNA1	-0.83285841
Mdg_06g019020-mRNA1	Mdg_02g014990-mRNA1	-0.830533544
Mdg_02g014990-mRNA1	Mdg_02g014440-mRNA1	-0.819903242
Mdg_10g028000-mRNA1	Mdg_02g014990-mRNA1	-0.817650655
Mdg_02g014990-mRNA1	Mdg_10g011740-mRNA1	-0.811750223
Mdg_02g014990-mRNA1	Mdg_09g009450-mRNA1	-0.81066779
Mdg_02g014990-mRNA1	Mdg_13g007380-mRNA1	-0.804955294
Mdg_02g014990-mRNA1	Mdg_12g007560-mRNA1	-0.802932795

This new file, FM27corrleations.txt, we can be used as input for Cytoescape.

Cytoescape visualization

Here we have the visualization network of module 27 created in Cytoescape.

References

Langfelder, P., Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008). https://doi.org/10.1186/1471-2105-9-559

Patro, R., Duggal, G., Love, M. I., Irizarry, R. A., & Kingsford, C. (2017). Salmon provides fast and bias-aware quantification of transcript expression. Nature Methods.

Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T. Cytoscape: a software environment for integrated models of biomolecular interaction networks

Spliced Transcripts Alignment to a Reference © Alexander Dobin, 2009-2024 https://www.ncbi.nlm.nih.gov/pubmed/23104886. Alex Dobin, dobin@cshl.edu. https://github.com/alexdobin/STAR/issues

Sun, X., Jiao, C., Schwaninger, H. et al. Phased diploid genome assemblies and pan-genomes provide insights into the genetic history of apple domestication. Nat Genet 52, 1423–1432 (2020). https://doi.org/10.1038/s41588-020-00723-9

Alex's Lemonade Stand Foundation.Childhood Cancer Data Lab of ALSF.2020. Link: https://github.com/AlexsLemonade