This repo contains the code for running the Nanocompore analysis workflow using Nextflow.

The pipeline runs the following steps:

• Basecalls the raw fast5 files using Albacore
• Run pycoQC on Albacore's output
• Prepare a fasta file for the reference transcriptome
• Map the basecalled data to the reference using minimap2
• Realign the raw signal-level data to the kmers of the reference with Nanopolish
• Collapse Nanopolish output by kmer using NanopolishComp
• Run Nanocompore (to be implemented)

All steps are executed in a Conda environment that includes all dependencies.

Prepare a tab-separated file that describes the samples:

SampleName  Condition DataPath
Ctrl1       Ctrl      /path/to/fast5dir
Ctrl2       Ctrl      /path/to/fast5dir
Treated1    Trt       /path/to/fast5dir
Treated2    Trt       /path/to/fast5dir

Configure the pipeline by editing the nextflow.config file.

All parameters are described in the comments and should be self explanatory.

The only two options that deserve an explanation are target_trancripts and input_is_basecalled. The former allows you to provide the path to a text file that lists Ensembl transcript IDs of interest. Any transcript not present in this list will be discarded from the reference. input_is_basecalled allows you to start the pipeline after the Albacore step. In order for this to work, the paths in the sample annotation file must point to basecalled fast5 files.

To run the pipeline just execute: nextflow run pipeline.nf

The pipeline is shipped with 2 profiles: local (default) and yoda, which respectively run the processing on the local machine or on the Yoda HPC cluster at the EBI-EMBL. The local profile limits the number of cuncurrent tasks to 10, which is reduced to 2 and 3 for Albacore and Nanopolish respectively. The Yoda profile if configured to use the LSF executor with dynamic memory management.