Skip to main content

Bisulfite sequencing data processing and differential methylation analysis

Project description

methylpy

Welcome to the home page of methylpy, a pyhton-based analysis pipeline for

  • (single-cell) (whole-genome) bisulfite sequencing data
  • (single-cell) NOMe-seq data
  • differential methylation analysis

methylpy is available at github and PyPI.

Note

  • methylpy was initiated by and built on the work of Mattew D. Schultz
  • tutorial is being written
  • In new version, allc files for one sample are no longer split and the current allc format does not support header. Command cat allc_SAMPLENAME_*.tsv |grep chr -v > allc_SAMPLENAME.tsv can be used to change the older version of allc to the new version.
  • Current version methylpy has major difference compared to previous version. Please checkout this document and tutorial for details.

What can methylpy do?

Processing bisulfite sequencing data and NOMe-seq data

  • fast and flexible pipeline for both single-end and paired-end data
  • all the way from raw reads (fastq) to methylation state and/or open chromatin readouts
  • also support getting readouts from alignment (BAM file)
  • including options for read trimming, quality filter and PCR duplicate removal
  • accept compressed input and generate compressed output
  • support post-bisulfite adaptor tagging (PBAT) data

Calling differentially methylated regions (DMRs)

  • DMR calling at single cytosine level
  • support comparison across 2 or more samples/groups
  • conservative and accurate
  • useful feature for dealing with low-coverage data by combining data of adjacent cytosines

What you want to do

run methylpy -h to get a list of functions.

Install methylpy

Step 1 - Download methylpy

Easiest way of installing methylpy will be through PyPI by running pip install methylpy. The command pip install --upgrade methylpy updates methylpy to latest version. Alternatively, methylpy can be installed through github: enter the directory where you would like to install methylpy and run

git clone https://github.com/yupenghe/methylpy.git
cd methylpy/
python setup.py install

If you would like to install methylpy in path of your choice, run python setup.py install --prefix=/USER/PATH/. Then, try methylpy and if it gives no error, the setup is likely successful. See Test methylpy for more rigorious test. Last, processing large dataset will require space for storing temporary files. However, the default directory is unlikely to fit the need. You may want to set the TMPDIR environmental variable to the absolute path of a directory on hard drive with sufficient space (e.g. /YOUR/TMP/DIR/). This can be done by adding the below command to ~/.bashrc file: export TMPDIR=/YOUR/TMP/DIR/.

Step 2 - Install dependencies

methylpy is written in python so obviously python2/3 needs to be installed. methylpy also depends on two python modules, numpy and scipy. The easiest way to resolve these dependencies is to install anaconda

In addition, some features of methylpy depend on several publicly available tools (not all of them are required if you only use a subset of methylpy functions). * cutadapt (>=1.9) for raw read trimming * bowtie and/or bowtie2 for alignment * samtools (>=1.3) for alignment results manipulation * Picard (>=2.10.8) for removal of PCR duplicates * java (its path included in PATH environment variable) for running Picard * wigToBigWig for converting methylpy output to bigwig format

Lastly, if paths to cutadapt, bowtie/bowtie2, samtools and wigToBigWig are included in PATH variable, methylpy can run these tools directly. Otherwise, the paths have to be passed to methylpy as augments. Path to Picard needs to be passed to methylpy as a parameter to run PCR duplicate removal.

Optional step - Compile rms.cpp

DMR finding requires an executable methylpy/methylpy/run_rms_tests.out, which was compiled from C++ code methylpy/methylpy/rms.cpp. In most cases, the precompiled file can be used directly. To test this, simply run execute methylpy/methylpy/run_rms_tests.out. If help page shows, recompiling is not required. If error turns up, the executable needs to be regenerated by compiling rms.cpp and this step requires GSL installed correctly. In most linux operating system, the below commands will do the job

cd methylpy/methylpy/
g++ -O3 -l gsl -l gslcblas -o run_rms_tests.out rms.cpp

In Ubuntu (>=16.04), please try the below commands first.

cd methylpy/methylpy/
g++ -o run_rms_tests.out rms.cpp `gsl-config --cflags —libs`

Test methylpy

To test whether methylpy and the dependencies are installed and set up correctly, run

cd methylpy/test
python run_test.py

Process data

Please see methylpy tutorial for more details.

Step 1 - Build converted genome reference

Build bowtie/bowtie2 index for converted genome. Run methylpy build-reference -h to get more information. An example of building mm10 mouse reference index:

methylpy build-reference \
    --input-files mm10_bt2/mm10.fa \
    --output-prefix mm10_bt2/mm10 \
    --bowtie2 True

Step 2 - Process bisulfite sequencing and NOMe-seq data

Function single-end-pipeline is For processing single-end data. Run methylpy single-end-pipeline -h to get help information. Below code is an example of using methylpy to process single-end bisulfite sequencing data. For processing NOMe-seq data, please use num_upstr_bases=1 to include one base upstream cytosine as part of cytosine sequence context, which can be used to tease out GC sites.

methylpy single-end-pipeline \
    --read-files raw/mESC_R1.fastq.gz \
    --sample mESC \
    --forward-ref mm10_bt2/mm10_f \
    --reverse-ref mm10_bt2/mm10_r \
    --ref-fasta mm10_bt2/mm10.fa \
    --num-procs 8 \
    --remove-clonal True \
    --path-to-picard="picard/"

An command example for processing paired-end data. Run methylpy paired-end-pipeline -h to get more information.

methylpy paired-end-pipeline \
    --read1-files raw/mESC_R1.fastq.gz \
    --read2-files raw/mESC_R2.fastq.gz \
    --sample mESC \
    --forward-ref mm10_bt2/mm10_f \
    --reverse-ref mm10_bt2/mm10_r \
    --ref-fasta mm10_bt2/mm10.fa \
    --num-procs 8 \
    --remove-clonal True \
    --path-to-picard="picard/"

Output format

Output file(s) are (compressed) tab-separated text file(s) in allc format. “allc” stands for all cytosine (C). Each row in an allc file corresponds to one cytosine in the genome. An allc file contain 7 columns and no header:

index column name example note
1 chromosome 12 with no “chr”
2 position 18283342 1-based
3 strand
either + or -
4 sequence context CGT can be more than 3 bases
5 mc 18 count of reads supporting methylation
6 cov 21 read coverage
7 methylated 1 indicator of significant methylation

Call DMRs

This function will take a list of compressed/uncompressed allc files (output files from methylpy pipeline) as input and look for DMRs. Help information of this function is available via running methylpy DMRfind -h.

Below is the code of an example of calling DMRs for CG methylation between two samples, AD_HT and AD_IT on chromosome 1 through 5 using 8 processors.

methylpy DMRfind \
    --allc-files allc/allc_AD_HT.tsv.gz allc/allc_AD_IT.tsv.gz \
    --samples AD_HT AD_IT \
    --mc-type "CGN" \
    --chroms 1 2 3 4 5 \
    --num-procs 8 \
    --output-prefix DMR_HT_IT

Please see methylpy tutorial for details.

Additional functions for data processing

Extract cytosine methylation state from BAM file

The call-methylation-state function allows users to get cytosine methylation state (allc file) from alignment file (BAM file). It is part of the data processing pipeline which is especially useful for getting the allc file from alignment file from other methylation data pipelines like bismark. Run methylpy call-methylation-state -h to get help information. Below is an example of running this function. Please make sure to remove --paired-end True or use --paired-end False for BAM file from single-end data.

methylpy call-methylation-state \
    --input-file mESC_processed_reads_no_clonal.bam \
    --paired-end True \
    --sample mESC \
    --ref-fasta mm10_bt2/mm10.fa \
    --num-procs 8

Get methylation level for genomic regions

Calculating methylation level of certain genomic regions can give an estimate of the methylation abundance of these loci. This can be achieved using the add-methylation-level function. See methylpy add-methylation-level -h for more details about the input format and available options.

methylpy add-methylation-level \
    --input-tsv-file DMR_AD_IT.tsv \
    --output-file DMR_AD_IT_with_level.tsv \
    --allc-files allc/allc_AD_HT_1.tsv.gz allc/allc_AD_HT_2.tsv.gz \
        allc/allc_AD_IT_1.tsv.gz allc/allc_AD_IT_2.tsv.gz \
    --samples AD_HT_1 AD_HT_2 AD_IT_1 AD_IT_2 \
    --mc-type CGN \
    --num-procs 4

Merge allc files (of replicates)

The merge-allc function can merge multiple allc files into a single allc file. It is useful when separate allc files are generated for replicates of a tissue or cell type, and one wants to get a single allc file for that tissue/cell type. See methylpy merge-allc -h for more information.

methylpy merge-allc \
    --allc-files allc/allc_AD_HT_1.tsv.gz allc/allc_AD_HT_2.tsv.gz \
    --output-file allc/allc_AD_HT.tsv.gz \
    --compress-output True

Convert allc file to bigwig format

The allc-to-bigwig function generates bigwig file from allc file. Methylation level will be calculated in equally divided non-overlapping genomic bins and the output will be stored in a bigwig file. See methylpy allc-to-bigwig -h for more information.

methylpy allc-to-bigwig \
    --input-allc-file results/allc_mESC.tsv.gz \
    --output-file results/allc_mESC.bw \
    --ref-fasta mm10_bt2/mm10.fa \
    --mc-type CGN \
    --bin-size 100

Quality filter for bisulfite sequencing reads

Sometimes, we want to filter out reads that cannot be mapped confidently or are likely from under-converted DNA fragments. This can be done using the bam-quality-filter function. See methylpy bam-quality-filter -h for parameter inforamtion.

For example, below command can be used to filter out reads with less than 30 MAPQ score (poor alignment) and with mCH level greater than 0.7 (under-conversion) if the reads contain enough (at least 3) CH sites.

methylpy bam-quality-filter \
    --input-file mESC_processed_reads_no_clonal.bam \
    --output-file mESC_processed_reads_no_clonal.filtered.bam \
    --ref-fasta mm10_bt2/mm10.fa \
    --quality-cutoff 30 \
    --min-num-ch 3 \
    --max-mch-level 0.7 \
    --buffer-line-number 100

Reidentify DMRs from existing result

methylpy is able to reidentify-DMR based on the result of previous DMRfind run. This function is especially useful in picking out DMRs across a subset of categories and/or with different filters. See methylpy reidentify-DMR -h for details about the options.

methylpy reidentify-DMR \
    --input-rms-file results/DMR_P0_FBvsHT_rms_results.tsv.gz \
    --output-file results/DMR_P0_FBvsHT_rms_results_recollapsed.tsv \
    --collapse-samples P0_FB_1 P0_FB_2 P0_HT_1 P0_HT_2 \
    --sample-category P0_FB P0_FB P0_HT P0_HT \
    --min-cluster 2

Cite methylpy

If you use methylpy, please cite >Matthew D. Schultz, Yupeng He, John W.Whitaker, Manoj Hariharan, Eran A. Mukamel, Danny Leung, Nisha Rajagopal, Joseph R. Nery, Mark A. Urich, Huaming Chen, Shin Lin, Yiing Lin, Bing Ren, Terrence J. Sejnowski, Wei Wang, Joseph R. Ecker. Human Body Epigenome Maps Reveal Noncanonical DNA Methylation Variation. Nature. 523(7559):212-216, 2015 Jul.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for methylpy, version 1.0.8
Filename, size File type Python version Upload date Hashes
Filename, size methylpy-1.0.8.tar.gz (88.3 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page