Multi-tool QC for data from Oxford Nanopore's Minion Sequencer

What?

A single script that is adapted to perform QC with a range of popular tools on the output of a single flowcell of Oxford Nanopore's Minion sequencer.

Why?

Becuase there are lots of good tools, and if you really want to know whether your data are any good, it's useful to use many of them at once.

Dependencies / tools used

Get a coffee - there are a few things to install first...

• Albacore: https://community.nanoporetech.com/protocols/albacore-offline-basecalli/v/abec_2003_v1_revs_29nov2016/run-albacore-on-linux
• minion_QC.R: https://github.com/roblanf/minion_qc
• NGMLR: https://github.com/philres/ngmlr
• qualimap: http://qualimap.bioinfo.cipf.es/
• samtools: https://github.com/lh3/samtools
• miniasm: https://github.com/lh3/miniasm
• Quast: http://quast.sourceforge.net/download.html

Install all of these, and then make sure everything is somewhere in your path so that you can run each tool succesfully as follows:

read_fast5_basecaller.py # albacore
minion_QC.R
ngmlr
qualimap 
samtools
minimap
miniasm
quast.py

Once you have all of those up and running, you can run the script.

Settings

I like to hard-code the settings in the script. For each flowcell I just copy the script into the folder with the data, change the settings below, and then run it (see below for how to run it). I do it this way so that the script itself provides the record of exactly what I did to QC each flowcell.

• inputf: an input folder of raw fast5 files from the minion. Fine if it has lots of subfolders.
• outputbase: the output file where you want all the processed data, QC reports, and summary stats to go. Output will be stored in subfolders of this. I usually just set it to be the parent folder of the input folder, as below, so that the QC and the raw data are in the same place.
• ref: The reference genome. Fine if it's a distant reference - we just use the mapping stats to get a rough idea of the what's in the reads, and to compare flowcells.
• gff: a gff of genes in your reference genome
• threads: how many cores to use for all the analyses
• mem_size: how much memory can you give to Qualimap. You may need to reduce this if you're running this on a desktop with <50G. WE have this here becuase qualimap sometimes needs a lot of memory.
• flowcellID: the ID of your flowcell. Albacore needs this
• kitID: the ID of your library prep kit. Albacore needs this too.

Here's an example

inputf="/data/nanopore/RB7_A2/20170609_0424_Epauciflora_A2/"
outputbase="/data/nanopore/RB7_A2/"
ref="/data/active_refs/Epau.fa.gz"
gff="/data/active_refs/Egrandis_genes_chr1_to_chr11.gff3"
threads=55 # number of threads to use
mem_size='50G'
flowcellID="FLO-MIN107"
kitID="SQK-LSK108"

Running it

Easy:

sh minion_multiQC.sh

Overview of what it does

1. Basecall with Albacore
2. QC plots and stats on the called bases using minion_QC.R
3. Map to the reference with NGMLR
4. Run qualimap on the whole reference, and just the genes (we do the latter because you often expect better mapping to genes with a distant reference, because genes tend to be more conserved)
5. Assess the mapping of reads at various lengths with samtools
6. Assemble with minimap/miniasm
7. Assess assembly with Quast

Output

A lot. Dig around in the folders and files, and look at the websites for each tool to get a handle on the information.

How long does it take?

On a server with 30 cores assigned, the whole thing runs in a few hours on a flowcell with ~4GB of data.