A python-based miRNA sequencing pipeline for isomiR quantification and analysis.
Please see https://github.com/Gu-Lab-RBL-NCI/QuagmiR for the newest version of this tool.
- Make sure that you have Python 3.4+ installed (type
python --versionin the console) - Make sure you have the latest version of pip:
pip3 install -U pip - Make sure you have Miniconda installed
- Download repository:
git clone https://github.com/kevchn/quagmir - Go into local quagmir folder:
cd quagmir - Install Python dependencies:
conda env create -f environment.yml
- Add your .fastq samples into the data folder (a sample has been provided for testing):
├── LICENSE ├── README.md ├── config.yaml ├── Snakefile ├── environment.yml ├── data/ │ ├── sample.fastq_ready │ └── YOUR_FILE_HERE.fastq_ready | ├── collapsed/ ├── motif-consensus.fa └── results/ │ └── tabular/
- Edit the motif-consensus.fa file to insert your miRNA information with the following format:
>miRNA_name miRNA_motif
miRNA_consensus_sequence
>passenger-shRNA-mir21-ORF59-5p-1 ACACCCTGGCCGGGT
CCGACACCCTGGCCGGGTTGT
- Edit the config.yaml file to change configuration options if needed (default values fine in most use cases):
# DISPLAY
min_ratio: .001
min_read: 9
display_summary: True
display_sequence_info: True
display_nucleotide_dist: True
# FUNCTION
destructive_motif_pull: False
# INPUT
motif_consensus_file: 'motif-consensus.fa'
- Run pipeline:
bash run.shor activate conda env and runsnakemake
Run the commands git reset --hard and git pull.
Quagmir skips any reads that it considers not to be an isomir of the miRNA, even if the reads are pulled in by the motif of the miRNA. These skipped reads are reported in the run log in the logs/ folder.
Currently, Quagmir by default skips reads that:
(1) Have none-templating trimming and tailing combinations on the 5' end
Biologically, 5' modifications are very rare/non-existent. This removes reads that supposedly have these modifications. Keeps isomirs that are a result of imprecise cleavage. False positive: Also keeps potentially 'fake' isomirs that have non-templating modifications beyond the extent of the provided consensus mirna.
(2) Have a mutation 3nt from the consensus end (shares rest of end)
**Most likely to be sequencing errors rather than trimming/tailing if 3nt or more are tailed that match the canonical sequence (1/4^3 = 1.5%).
Quagmir also has an optional flag (off by default) to skip reads that:
(3) Were reported to be an isomir of a previously reported miRNA
Biologically, a read can only belong to one miRNA
The step of collapsing sample files takes the longest time, but once the samples are collapsed, and you need to re-run the pipeline, the pipeline will automatically start from the collapsed files and take a far shorter amount of time.
To save time you may consider to skip distance metric (weighted-levenstein distance) which is calculated on group results.
Output will be a sample.fastq.results.txt file for each sample in the results/ folder and grouped results if multiple files are provided. All resulting files are in tabular (TSV) format.
- Have a provided primary miRNA transcript for each miRNA either in motif-consensus as well as gen-uniq-subseq, or have it pre-provided in local files. Then implement features to smartly filter out reads that are not isomiRs of a miRNA because they have statistically improbable behavior (e.g 5p non-templating addition) or sequencing errors.