BAT_correlating

This is has the potential to be very exciting or a huge letdown.

With BAT, you can provide methylation and gene expression data to BAT_correlating to get correlation information for the mean methylation rate of the DMR and expression of the associated gene. Even though this is part of the BAT DMR module, I decided to set up a new set of scripts for the analysis, since I had to run a couple of analyses in between to get to this point:

• singularity code
• envfile
• BAT_correlating code

I needed the following information:

-b BED file containing coordinates of region (for overlapping with methylation bigWig file) and name of associated gene identifier (to connect to expression file). Format: chr start end identifier. Identifier need to match gene identifiers of the expression files (option -e). -e File containing a list of expression files (one per sample). The content of each file must follow this format identifier something expression_value. -m File containing a list of methylation bigWig files (one per sample). -g File containing the association of sample identifiers to groups. Format: sample group. The list of samples in the file must be ordered analogously as in the list of files given with options -e and -m. -i Comma-separated list of group 1 and group2 as well as an optional additional third group for plotting. Groups need to match the group names in the file provide with option -g. -o Path/prefix for output (path for correlation plots and path/prefix.txt for statistics).

All of them seemed pretty straightforward to make. I got the gene expression data and a list of file mappings from Neel and proceeded to follow the example code to make input files:

#Create BAT_correlating input

mkdir ${CORR}/20_OC_N

echo "Create BEDfile with DMR and gene coordinates"

cat ${ANNOT}/20_OC_N_DMR.closestGene.bed \
| cut -f1-3,14 \
| tr ";" "\t" \
| cut -f1-4 \
| sed 's/ID=gene-//g' \
> ${CORR}/20_OC_N/20_OC_N_DMR_gene.bed

echo "Create methylation bigWig file list"

ls ${BIGWIG}/20_OC_N/*bw \
| grep -v "mean" \
| grep -v "diff" \
| sed 's/\t/\n/' \
> ${CORR}/20_OC_N/20_OC_N_methylation_files.list

echo "Create gene expression file list"

ls ${RNA}/*threeCol.txt \
| grep -e "20-N" -e "OC-N" \
| sed 's/\t/\n/' \
> ${CORR}/20_OC_N/20_OC_N_expression_files.list

echo "Create sample identifier list"

echo -e '20-N1\tNO\n20-N2\tNO\n20-N4\tNO\nOC-N1\tOC\nOC-N2\tOC\nOC-N3\tOC\nOC-N4\tOC\nOC-N5\tOC' > ${CORR}/20_OC_N/20_OC_N_sample_to_group.txt

A couple of notes about how I made the gene expression files. I got this new metadata table from Neel that maps the RNA-Seq and RRBS data files to eachother. The RNA-Seq and RRBS files have to have the same sample identifier. I renamed files and saved them in this folder.

Initially I didn’t modify the RNA-Seq data. Then when I was running my code I got the following error:

DMR Correlating Module Oxygen within NB: Outside Control vs. Hypoxia Create BEDfile with DMR and gene coordinates Create methylation bigWig file list Create gene expression file list Create sample identifier list [INFO] Fri Aug 16, 15:8:24, 2024 BAT_correlating v0.1 started [INFO] Fri Aug 16, 15:8:24, 2024 Checking input [INFO] Fri Aug 16, 15:8:24, 2024 Calculating average methylation /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory /bin/bash: /usr/bin/lesspipe.sh: No such file or directory [INFO] Fri Aug 16, 15:8:25, 2024 Getting expression values [INFO] Fri Aug 16, 15:8:25, 2024 Started plotting correlation plots Error in scan(file = file, what = what, sep = sep, quote = quote, dec = dec, : line 1 did not have 4 elements Calls: main -> read.table -> scan Execution halted AN ERROR has occurred: problem with calling R Done with DMR correlating

I reviewed the input file information and realized the RNA-Seq data needed three columns (identifier <tab> something <tab> expression_value), and that the second column didn’t need to be anything informative! I used awk to modify the columns of each RNA-Seq file:

#Navigate to bedGraph directory
cd /vortexfs1/home/yaamini.venkataraman/killifish-hypoxia-RRBS/data/RNA-Seq-renamed

#Sort bedGraphs
for f in *txt
do
awk -F "\t" '{ print $1, $1, $2 }' ${f} \
> $(basename ${f%.txt}).threeCol.txt
done

head *threeCol.txt

Modifying the files got rid of that error!

Once I had my input files, I ran the code:

#Create output directory
mkdir ${CORR}/20_OC_N/correlation_output

#Run BAT_DMRcalling
#-b: BED file with coordinates of methylation region and name of identifier: chr <tab> start <tab> end <tab> identifier
#-e: List of expression files, one per sample. Each sample's file: identifier <tab> something <tab> expression_value
#-m: File containing a list of methylation bigWig files (one per sample).
#-g: File containing the association of sample identifiers to groups. The list of samples in the file must be ordered analogously as in the list of files given with options -e and -m.: sample <tab> group
#-i: Comma-separated list of group1 and group2
#-o Path/prefix for output

BAT_correlating \
-b ${CORR}/20_OC_N/20_OC_N_DMR_gene.bed \
-e ${CORR}/20_OC_N/20_OC_N_expression_files.list \
-m ${CORR}/20_OC_N/20_OC_N_methylation_files.list \
-g ${CORR}/20_OC_N/20_OC_N_sample_to_group.txt \
-i NO,OC \
-o ${CORR}/20_OC_N/correlation_output/

The input and output files are saved here. In both statistical output files, I noticed that NA was used for correlation coefficients and p-values where the DMR did not overlap with the gene. When the DMR and gene overlapped, there were numerical values.

The normoxia vs. outside control comparison had no significant DMR-gene expression correlations. However, the hypoxia vs. outside control comparison did! None of the genes significantly correlated with DMR are differentially expressed. Given that non-mammalian systems don’t have clear correlations between methylation and gene expression (even in vertebrates), I think it’s still worth exploring what genes had significant correlations.

Table 1. Genes where expression was significantly correlated with methylation

None of these genes are annotated! Classic. Another letdown with this workflow is that it creates pdfs……but I can’t access them. The code creates pdfs plotting gene expression against region methylation to show the strength of the correlation. While I can see these plots were created, I can’t open or move them! Interestingly, I got an error when I tried to move the files with rsync saying they disappeared. These plots are probably easy enough to create myself, so I’ll do that when I make some DMR plots anyways.

Going forward

1. Create DMR figures
2. Update methods and results
3. Identify known SNP/DMR overlaps
4. Update OSF repository for all intermediate files