1. Cloning
2. Environment
3. Snakemake usage of this repo
4. Perform the processing of a single sample
5. Guppy Basecall
6. TailfindR
7. Flair
8. Albacore
9. EpiNano

Make sure to clone the submodules that comprises dependancies not manageable by conda (or pip).

git clone --recursive https://github.com/sunsetyerin/nanopore_dRNA_m6A_pA_tail_isoform.git

or

git clone https://github.com/sunsetyerin/nanopore_dRNA_m6A_pA_tail_isoform.git
cd direct_rna
git submodule update --init --recursive

This repository has been developped using python3.7.3 through python3.7.6 in a miniconda environment.

Though conda is not required, it is highly recommended in order to manage the dependencies. Users can choose to manage their dependencies manually and should consult the envs/environment.yaml file as reference.

Using snakemake to manage the environment

From a bash terminal:

snakemake --use-conda --conda-create-envs-only --cores [cores available]

Using conda to install the environment manually.

conda env create --file envs/environment.yaml

conda activate direct_rna

snakemake [args]

All scripts, steps and jobs are managed using snakemake

Input files, workflow parameters and output path are provided through config files (configs/*.yaml).

The workflow is designed to compare direct-RNA samples.

# To display the jobs and commands including subworkflows
snakemake -np

# to visualize the acyclic rulegraph
snakemake --rulegraph | dot | display -

# To launch the jobs
snakemake --use-conda

An example of snakemake profile for "numbers" cluster is available at configs/smk_profiles/numbers/config.yaml and a second one for "slurmgpu" will be available eventually.

So far the processing of a sample includes basecalling, polyA tail estimation, gene/isoform expression and match to Illumina gene expression (if available in Flair configs, see config file).

Config files contains the values of variables needed to perform analyses.

configs/sample_config.yaml is under development and is provided as an example. It's meant to be used to apply the analyses on those first data available from the PromethION platform.

Most of the data directories hosting results will be created ad-hoc by the snakemake manager, although it is possible to create the folders in advance. Any of these folders can be replaced by a symlink in order to redirect voluminous files to appropriate spaces. The entire data/ folder can be written to a project space with a simple:

ln -s /PATH/TO/SCRATCH/data data

It is recommended to use symlinks to redirect the data/ folder to a scrath space since the outputs can get particularly large.

Oxford Nanopore Technologies apparels uses MinKNOW software to basecall reads in realtime, at the expense of the production of the event table (formerly produced by Albacore). Albacore is discontinued and Guppy is currently favored. It can produce a move table during basecall that is similar to the deprecated event table (make sure to use --fast5_out). TailfindR requires one of those tables.

Download available here.

Current configs are using an installation available to dlhost01 & dlhost03. /opt/ont-guppy_3.2.2/bin/guppy_basecaller

TailfindR is designed to be supplied with a single directory containing all the .fast5 files. It supports parallelization by specifying the number of cores to attribute to the TailfindR job. However, it was designed and tested with MinIONs output files and I noticed that it doesn't scale well to a full PromethION flowcell. The first large usage of TailfindR on gphost04 using the sequencing run of the COLO829 cell line with 1563 .fast5 files of ~4000 reads (~6.25M reads) took 1.5 weeks to process all files and failed to produce the compilation of the results after another 2 weeks.

The chosen workaround to address the issue is to generate temporary directories each containing a softlink towards a single (or few) .fast5 file(s) (see rule tailfindr_butler). A TailfindR job is then generated for each temporary directories. Numerous TailfindR jobs should be submitted to a cluster such as numbers through snakemake.

Flair is a workflow in itself and requires its own environment to run. Both the scripts used for direct_rna analysis and environment .yaml files are provided in the flair submodule.

Unlike TailfindR the input is a single .fast5 which prompts the need for catenation of the multiple 4000 reads .fast5 produced by Guppy.

Albacore is a deprecated basecaller that was discontinued by Oxford Nanopore Technologies to the benefit of the currently maintained Guppy. It is included to this workflow as a requirement for EpiNano m6A caller which relies on the basecall errors of Albacore to assign methylated sites.

EpiNano states that it requires Albacore v2.1.7, to provide accurate prediction. This version is unavailable. The community only have access to Albacore v2.3.4. link to Nanoporetech community

NOTE Albacore is used along with the multiread .fast5 wrapper available via the seamlessf5 package.

The current version of EpiNano allows m6A sites calling at the genomic position level. The calling at read level is under development.