troygarry/l1em

Estimate locus specific human LINE-1 expression.

Installation

conda way

You will need

1. git (https://git-scm.com/book/en/v2/Getting-Started-Installing-Git)
2. anaconda (https://docs.anaconda.com/anaconda/install/)

Download from github

git clone https://github.com/wmckerrow/L1EM

Create conda environment

cd L1EM
conda env create -f L1EM.yml

Before running L1EM, activate the environment:

source activate L1EM

When finished, deactivate the environment:

source deactivate L1EM

old way

Alternatively you can install the following dependencies yourself:

• python version 2.7
• bwa (version 0.7.17 tested)
• samtools (version 1.9 tested)
• numpy (version 1.14.3 tested)
• scipy (version 1.1.0 tested)
• pysam (version 0.15.0 tested)
• bedtools (version 2.27.1 tested)

No compiling of L1EM is necessary. Python scripts will be called from inside the L1EM directory.

If necessary, you can specify the path for bwa and samtools in the run_L1EM.sh script. You must use samtools >=1.0. Early version of pysam will not work. I highly recommend that you use bwa 0.7.17. Earlier versions may differ in how they write the XA tag. This will lead to inaccurate results without throwing an error.

Quick guide

First time: build L1EM reference

You will need the hg38 reference genome in fasta format, with bwa index. Downloaded from UCSC genome browser:

wget http://hgdownload.cse.ucsc.edu/goldenPath/hg38/bigZips/hg38.fa.gz
zcat hg38.fa.gz > hg38.fa
bwa index hg38.fa

Note: this will take some time.

Then you can build the L1EM reference using the provided shell script:

bash generate_L1EM_fasta_and_index.sh /fullpathto/hg38.fa

This should be done inside the L1EM directory

Executing the L1-EM pipeline

You will need a bam file with strand specific paired end read alignments to hg38. You can use any aligner, but make sure that all reads from the original fastq files are present trimming should be okay, but is tested. Filtering reads will potentially break the pipeline.

First move to an empty directory and then execute the shell script:

bash /fullpathto/run_L1EM.sh /fullpathto/alignments.bam /fullpathto/L1EM /fullpathto/hg38.fa

L1EM will write files with specific names, so do NOT run two instances of L1EM in the same directory.

At the end of the run_L1EM.sh script are a commented set of commands to delete all the intermediate files. If you wish to automatically delete intermediate files, you can delete these comments.

Output

At completion, three tab delimited tables will be written.

1. full_counts.txt: raw count estimates for each L1HS/L1PA* element with any aligned read pairs
2. l1hs_transcript_counts.txt: expression estimates for L1HS elements, reported as raw counts
3. filter_L1HS_FPM.txt: L1HS whose expression is supported by at least 100 read pairs, reported as FPM (read pairs per million properly aligned)

The rows of all files are L1 loci.

For full_counts.txt each of the five transcript types: only, runon, passive (sense), passive (antisense), antisense are reported.

For l1hs_transcript_counts.txt and filter_L1HS_FPM.txt only proper transcription from L1HS elements start at the 5' UTR is reported.

The results are also written as pickle files to facilitate further analysis in python. To generate a python dictionary with keys being the transcript names and values being the relative expression:

X_est = dict(zip(pickle.load(open('names_final.pkl')),pickle.load(open('X_final.pkl'))))

Additional details

• Our Bioinformatics paper introducing L1EM: https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btz724/5581349
• More details can be found in manual.md

Mouse Version

Scripts and annotation to measure the expression of LINE-1 loci in mm39 has been added. The mouse version uses all the same methodology as the human version, but has not been as rigorously tested.

1. Download and index the mm39 reference genome (UCSC genome browser version)

wget http://hgdownload.cse.ucsc.edu/goldenPath/mm39/bigZips/mm39.fa.gz
zcat mm39.fa.gz > mm39.fa
bwa index mm39.fa

2. Build the mm39 L1EM reference.

bash generate_mm39_L1EM_fasta_and_index.sh /fullpathto/mm39.fa

3. Run L1EM.

bash /fullpathto/run_L1EM_mm39.sh /fullpathto/alignments.bam /fullpathto/L1EM /fullpathto/mm39.fa

All L1Md loci are quantified in full_counts.txt. Normalized expression of 5' UTR intact young (L1Md_Tf I/II/II, L1Md_Gf I/II, L1Md_A I/II/III) LINE-1 loci supported by at least 100 reads can be found in filter_active_L1Md_FPM.txt.