A pipeline to process m6A-seq data and down stream analysis for he lab members.
Data analysis would benefit greatly from careful design of your experiment. Be sure each batch of m6A-Seq contains a normal control. Use the same analysis pipeline for all your datasets when you need to compare them.
You can use filezilla to download the data from che@osrfftp.uchicago.edu (port: 21) and upload it to youraccount@128.135.225.178 (port: 22).
Alternatively, you can ssh log into youraccount@128.135.225.81 and go to the directory by cd /directory of your favorite where you want to analyze your data. Then use the
sftp -r che@osrfftp.uchicago.edu:/data_file_or_directory_name ./
## for example
sftp -r che@osrfftp.uchicago.edu:/170823_NS500373_0154_AHG3MTBGX3-CHe-JN-63S-pl1 ./
to directly download the data from sequencing facility server to the lab analysis server.
In order to be easily processed, rename the files to short names without space and postfix .IN.fastq.gz or .m6A.fastq.gz for input and IP respectively.
For example,
Ctl1.IN.fastq.gz
Ctl1.m6A.fastq.gz
...
FTO_KO5.IN.fastq.gz
FTO_KO5.m6A.fastq.gz
Preparing a detailed descriptive metadata table is highly recommended. Then backup your datasets using the following command line:
## single file
smbclient -U username%passward //helabdata2.uchicago.edu/your_directory -c 'put local_file remote_file'
## directory
smbclient -U username%passward //helabdata2.uchicago.edu/your_directory -c 'recurse; prompt; mput local_directory*'
## for example
smbclient -U username%passward //helabdata1.uchicago.edu/xcui -c 'put m6A_IP.fq.gz m6A_IP.fq.gz'
Please find Scott if you don't have an account in our databackup server at <helabdata2.uchicago.edu>. If you have question about using smbclient to upload and download data, please consult Xiaolong.
We are going to use fast and sensitive alignment program Hisat2 to align the reads to the mycoplasma genome. You can go to official page of Hisat2 to learn more about the software and parameters setting.
Mapping can be done at command line by excuting the Hisat2 command one by one for each sample. Alternatively, we use a bash file to wrap the Hisat2 command to loop through all our samples to save us some effort to keep excuting hisat2 command.
You can use any text editor to make the bash file and save it to something.sh. For example, in unix command line, use vim to creat and edit vi run_alignment.sh and type i to insert content. After entering the content, hit Esc to quit insertion mode and Shift + : + wq + Enter to save and quit.
An example bash file is shown below. You use this template and repalce the sample names by yours.
INDEX="$HOME/Database/genome/contamination/mycoplasma"
Data="$HOME/path_to/your_favorite_directory"
mySample="sample1 sample2 sample3 sample4 treated1 treated2 treated3 treated4 treated5"
for s in $mySample
do
hisat2 -x $INDEX -k 7 --un-gz $s.IN.noMyco.fastq.gz --summary-file $s.IN.myco_summary -p 4 -U $Data/$s.IN.fastq.gz > $s.myco.input.sam
hisat2 -x $INDEX -k 7 --un-gz $s.m6A.noMyco.fastq.gz --summary-file $s.m6A.myco_summary -p 4 -U $Data/$s.m6A.fastq.gz > $s.myco.m6A.sam
## remove alignment result files because we don't really need alignment to mycoplasma, we only want to see alignment summary
rm $s.myco.input.sam
rm $s.myco.m6A.sam
wait
done
Then excute the bash file in the command line by
bash run_alignment.sh
The code above will align reads to mycoplasma genome and output alignment summary. You can check the alignment summary to assess the potential contamination by Mycoplasma.
It will also output reads that don't align to mycoplasma nameed as samplename.IN.noMyco.fastq.gz. This will be unmmaped reads filtering out mycoplasma reads, which will be input for next round of alignment to human/mouse/whatever_organism reference genome.
Similar to the code above, we use the Hisat2 to align the reads to the reference genome of the organism you are studying. The commonly used reference genome for Human and Mouse are downloaded and available at $HOME/Database/genome directory. I have also prepared the hisat2 index for hg38/19 mm10 so that you can directly use. If you need to use reference genome of other organism, please refer to the hisat2 mannual to download and build the index.
An example of bash file to map to the human hg38 reference genome is provided below:
INDEX="$HOME/Database/genome/hg38/hg38_UCSC"
SPLICE="$HOME/Database/genome/hg38/hisat2_splice_sites.txt"
Data="$HOME/path_to/the_directory_where_sample.IN/m6A.noMyco.fastq.gz_locate"
Output="/path_to_output_directory"
mySample="sample1 sample2 sample3 sample4 treated1 treated2 treated3 treated4 treated5"
for s in $mySample
do
hisat2 -x $INDEX --known-splicesite-infile $SPLICE -k 1 --no-unal --summary-file $s.IN.align_summary -p 4 -U $Data/$s.IN.noMyco.fastq.gz |samtools view -bS |samtools sort -o $Output/$s.input.bam
hisat2 -x $INDEX --known-splicesite-infile $SPLICE -k 1 --no-unal --summary-file $s.m6A.align_summary -p 4 -U $Data/$s.m6A.noMyco.fastq.gz | samtools view -bS |samtools sort -o $Output/$s.m6A.bam
wait
done
mkdir alignment_summary
mv *.align_summary alignment_summary/
Excuting the bash file above will align the reads to hg38 reference genome and out put bam (binary format of sam) files named as sample.input.bam and sample.m6A.bam in the output directory defined in the bash file.
With the aligned bam files, you are ready to proceed with varies down stream analysis such as Differential gene expression , Peak calling or Differential methylation analysis.
Gene level differential expression analysis can refer to tutorial here.
The following tutorial will assume that you store the mapped bam files in directory /home/xxx/project1/bam_files and the bam files are named by convention samplename.input.bam and samplename.m6A.bam.
First in a terminal (connected to our analysis server 128.135.225.178), enter R session by type R and Enter. You should see something like this when you entered R session.
R version 3.4.3 (2017-11-30) -- "Kite-Eating Tree"
Copyright (C) 2017 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
Natural language support but running in an English locale
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
Now you can load the R package m6Amonster for peak calling.
library(m6Amonster)
Then define the sample names of your analysis and the count the reads in consecutive bins on transcript.
samplenames = c("sample1","sample2","sample3")
monster <- countReads(samplenames = samplenames,
gtf = "~/Database/genome/hg38/hg38_UCSC.gtf",
bamFolder = "/home/xxx/project1/bam_files",
outputDir = "/home/xxx/project1/fisherPeak",
modification = "m6A",
binSize = 50,
threads = 6
)
Then countReads function will count reads on continuous bins of size defined by binSize = 50. The parameter gtf = "~/Database/genome/hg38/hg38_UCSC.gtf" defines the annotation file for gene model. If you are working on mouse, thie should be gtf = "~/Database/genome/mm10/mm10_UCSC.gtf" (you might need to download gtf files yourself for other organisms).
The parameter bamFolder = "/path" defines the directory where you put your bam files.
Note m6Amonster requires paired BAM files for both INPUT and IP. For each samplenames given to the countReads function, you will need to have samplename.input.bam and samplename.m6A.bam in your bamFolder directory.
Next we call peak by
monster <- callPeakFisher(monster, min_counts = 15, peak_cutoff_fdr = 0.05 , peak_cutoff_oddRatio = 1, threads = 6)
You can play with different parameter to tune the stringency of peak calling. The callPeakFisher() function call peak on each sample separately and add a matrix of logical value to list monster indicating for each samples (column), whether a bin (row) is enriched or not. You can check the logical values by monster$peakCallResult.
If you are interested in getting read counts in each peak (e.g. to estimate enrichment of IP experiment), you can use
monster <- JointPeakCount(monster, joint_threshold = 2)
The parameter joint_threshold = 2 defines the least number of sample to have peak called at each locus to call this locus as joint peak. If you require a peak to consistently be called in all samples, you should use joint_threshold = #of samples. If you want to include more locus where peak is call only in a subset of samples, this parameter difines the number of sample in the subset. Default is 2. The counts of Input is store at monster$jointPeak_input and the counts of IP is at monster$jointPeak_ip.
To report joint peak of all samples
Joint_peak <- reportJointPeak(monster, threads = 6)
unique.Joint_peak <- Joint_peak[which(!duplicated(paste(Joint_peak$chr,Joint_peak$start,Joint_peak$end,sep = ":"))),]
write.table(unique.Joint_peak, file = "/home/<your account name>/project1/fisherPeak/Joint_peak.bed", sep = "\t", row.names = F, col.names = F, quote = F)
library(MeTPeak)
samplenames = c("sample1","sample2","sample3")
Input_bamPath <- paste0("/home/xxx/project1/bam_files/",samplenames,".input.bam")
IP_bamPath <- paste0("/home/xxx/project1/bam_files/",samplenames,".m6A.bam")
metpeak(IP_BAM = IP_bamPath, INPUT_BAM = Input_bamPath,
GENE_ANNO_GTF= "~/Database/genome/hg38/hg38_UCSC.gtf",
OUTPUT_DIR= "/home/xxx/project1/fisherPeak",
EXPERIMENT_NAME = "You name it",
FRAGMENT_LENGTH = 150
)
In unix shell, cd to directory where your peak.bed file locate. First extract peak sequence from reference genome using bed file of peaks. -
bedtools getfasta -fi ~/Database/genome/hg38/hg38_UCSC.fa -bed Joint_peak.bed -fo Joint_peak.fa -split -s
Please read bedtools documentation to understand what this line of code is doing.
Then we can use motif search tools to perform motif enrichment analysis. Here I will demonstrate using Homer to search for motif.
findMotifs.pl jointPeak.fa fasta /home/xxx/project1/fisherPeak/homer_motif -fasta ~/Database/transcriptome/backgroud_peaks/hg38_200bp_randomPeak.fa -rna -p 10 -len 5,6,7
jointPeak.fa is the peak sequence you extracted In previous step in fasta format. fasta defined the format of input file. /home/xxx/project1/fisherPeak/homer_motif defined the output directory of homer motif analysis. -fasta difined the background sequence in fasta format. Here we provide a randomly sample 200bp peaks as background. -rna tell homer to output "U" instead of "T". -p 10 tell homer to use 10 threads. -len 5,6,7 tell homer to search for motif of length 5,6,7.
Here I use R package Guitar to plot the meta gene distribution of peaks. Guitar normalized the density of peaks on each feature (5UTR, CDS, 3UTR) by the length of the feature, which correlates to the expected occurence of peak by chance.
To get start, we need to have the peak in bed12 format read into R. In previous peak calling step, if you used m6Amonster to call joint peak:
Joint_peak <- reportJointPeak(monster, threads = 6)
You already have your peak in variable Joint_peak. You can proceed to next step
plotMetaGene(Joint_peak,gtf = "~/Database/genome/hg38/hg38_UCSC.gtf")
If you used other tools to call peak, you can first read peak file into R and then plot the meta gene:
peak <- read.table("path/to/peak.bed",sep = "\t", stringsAsFactors = F)
plotMetaGene(peak,gtf = "~/Database/genome/hg38/hg38_UCSC.gtf")