Overview
Nextflow is a pipeline engine that can take advantage of the batch nature of the HPC environment to efficiently and quickly run Bioinformatic workflows.
For more information about Nextflow, please visit Nextflow - A DSL for parallel and scalable computational pipelines
The ampliseq pipeline is a bioinformatics analysis pipeline used for 16S rRNA amplicon sequencing data. It combines multiple 16S analysis tools in a single pipeline, to produce a variety of statistical and quantitative outputs, that can be used in further analysis or in publications.
https://github.com/nf-core/ampliseq
Installing NextFlow on your HPC account
NextFlow needs to be set up locally for each user account on the HPC. Instructions for installing and setting up NextFlow for your account are here:
Follow the instructions in the above link, then when you have successfully run the NextFlow test (nextflow run hello generates Hello world!, etc), then run a test of the ampliseq pipeline (next section)
Running your NextFlow pipelines
NextFlow should never be run on the head node on the HPC (i.e. the node you are automatically logged on to) but should instead be run on a different node using the PBS job scheduler. For instructions on submitting PBS jobs, see here:
Running PBS jobs on the HPC Confluence page
Directory structure
When a NextFlow pipeline is run, it generates multiple directories and output files. We therefore recommend you create a directory where you run all your NextFlow pipelines, so that you don’t have output directories and files scattered across your home directory.
This creates a ‘nextflow’ subdirectory in your home directory. You can then create individual subdirectories for each pipeline you run (e.g. cd nextflow, mkdir apliseq_test).
cd ~ mkdir nextflow
Alternative to submitting PBS job: interactive session.
Run tmux first, so job keeps running when you log off.
tmux
Interactive PBS session:
qsub -I -S /bin/bash -l walltime=168:00:00 -l select=1:ncpus=4:mem=8gb
Then you can run all the following code.
Run an ampliseq test
This test is to see if NextFlow is installed correctly on your account and if you can run ampliseq. It uses a small built-in dataset to run ampliseq, running a full analysis and producing all the output directories and files.
If this test run fails, contact us at eResearch - eresearch@qut.edu.au - and we can examine the issue.
Note that running the test run on the HPC requires some additional steps to those listed on the ampliseq website. Primarily there are issues with ampliseq automatically downloading external files, so we need to download these locally, then change the ampliseq config file to point to these downloaded files.
Downloading test files
There are 3 files, or sets of files, that need to be downloaded first. The command to download each of these is included below.
First, go to your nextflow directory. We are assuming you called it ‘nextflow’ (see above ‘Running your NextFlow pipelines’ section for directory structure suggestions).
Go to your nextflow directory and create a subdirectory called ‘ampliseq_test’
cd ~/nextflow mkdir ampliseq_test cd ampliseq_test
Download the 3 required datafile sets:
1. The metadata file.
wget https://github.com/nf-core/test-datasets/raw/ampliseq/testdata/Metadata.tsv
2. The taxonomic classifier file.
wget https://github.com/nf-core/test-datasets/raw/ampliseq/testdata/GTGYCAGCMGCCGCGGTAA-GGACTACNVGGGTWTCTAAT-gg_13_8-85-qiime2_2019.7-classifier.qza
3. The test datafiles (fastq files). Note that the first line below is a very long one - it just adds all the datafile download paths to a test file, which can then be used in the wget command below it.
printf 'https://github.com/nf-core/test-datasets/raw/ampliseq/testdata/1_S103_L001_R1_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/1_S103_L001_R2_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/1a_S103_L001_R1_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/1a_S103_L001_R2_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/2_S115_L001_R1_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/2a_S115_L001_R2_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/2_S115_L001_R2_001.fastq.gz\nhttps://github.com/nf-core/test-datasets/raw/ampliseq/testdata/2a_S115_L001_R1_001.fastq.gz' > datafiles.txt wget -i datafiles.txt
Ampliseq test config file
NextFlow pipelines have a series of default settings, which can be overridden by modifying a config file. The default config file for ampliseq points to downloadable datafile locations. As we’ve downloaded the datafiles locally, we need to modify the config file to point to these local files instead.
Make sure you are in the directory you created for this test (containing the downloaded test datafiles) - cd ~/nextflow/ampliseq_test
Load a text editor first. Nano will do.
module load nano
Create and edit a ‘nextflow.config’ file
nano nextflow.config
This will create and open an empty text file in nano. Into this file copy and paste the following:
params {
classifier = "GTGYCAGCMGCCGCGGTAA-GGACTACNVGGGTWTCTAAT-gg_13_8-85-qiime2_2019.7-classifier.qza"
metadata = "Metadata.tsv"
readPaths = [
['1_S103', ['1_S103_L001_R1_001.fastq.gz', '1_S103_L001_R2_001.fastq.gz']],
['1a_S103', ['1a_S103_L001_R1_001.fastq.gz', '1a_S103_L001_R2_001.fastq.gz']],
['2_S115', ['2_S115_L001_R1_001.fastq.gz', '2_S115_L001_R2_001.fastq.gz']],
['2a_S115', ['2a_S115_L001_R1_001.fastq.gz', '2a_S115_L001_R2_001.fastq.gz']]
]
}
In nano you can then save the file by ‘ctrl o’ and then exit nano with ‘ctrl x’
Ampliseq test command
Run the following command to test the ampliseq pipeline.
nextflow run nf-core/ampliseq -profile test,singularity --metadata "Metadata.tsv"
This will submit multiple PBS jobs, for each test file and analysis step. How fast this will run depends on how large the queue is for using the HPC. If there is no queue (rare) the test run will finish in approx 30 minutes. regardless, this does not need to be actively monitored. You can check later to see if the test run succeeded or failed. If it succeeded (you will see ‘Pipeline completed successfully’ message), continue with the analysis of your dataset. if it failed (multiple potential error messages), contact us at eResearch and we will work through the issue: eresearch@qut.edu.au.
You should see several output directories and files have been created in your ‘ampliseq_test’ directory. These contain the test analysis results. Have a look through these, as they are similar to the output from a full ampliseq run (i.e. on your dataset).
Ampliseq output
As can be seen in the test results (see above section), ampliseq produces a ‘results’ directory with several subdirectories, which contain various analyses outputs. These are outlined here:
https://nf-co.re/ampliseq/1.1.3/output
Briefly, these include quality control and a variety of taxonomic diversity and abundance analyses, using QIIME2 and DADA2 as core analysis tools.
These directories contain various tables and figures that can be used in either downstream analysis or directly in publications.
Running ampliseq on your dataset
In this section we will focus primarily on the commands and files you need to run the pipeline on your data. A complete description of the ampliseq pipeline is on the ampliseq websites. To properly understand the ampliseq processes and analysis outputs, it is advisable that you thoroughly read through these.
https://github.com/nf-core/ampliseq
Ampliseq requires the creation of some additional files and modification of parameters files in order to run on your dataset. Instructions below.
Directory structure and files
Make sure you have created a subdirectory in your ‘nextflow’ directory. Give it a meaningful name (e.g. mkdir <yourprojectname>_nextflow. Make sure you are in that directory (cd ~/nextflow/<yourprojectname>_nextflow).
As with the test run, you will need to download some datafiles and create some new files (manifest file, metadata file, nextflow.config file) to get ampliseq running on the HPC.
NOTE: be very careful about the naming and structure of these files. Sample IDs in the manifest and metadata files must match exactly and the file paths need to be correct. Column names must be named exactly as in the examples below (including case). Spelling errors, a stray comma or other character in these files is one of the more common reasons for ampliseq to fail
Taxonomic database
Download the silva database. This is the main database ampliseq uses for taxonomic classification.
wget https://www.arb-silva.de/fileadmin/silva_databases/qiime/Silva_132_release.zip
Manifest file
In the test run, a list of your filenames and the associated sample ID was included in the nextflow.config file. With many sample files it’s much easier to include this information in a separate manifest file.
This is a tab delimited file that contains 3 columns:
‘sampleID’, with the sample IDs. You can call these whatever you like, but it should be meaningful (e.g. groupA_1, groupA_2, groupb_1, groupB_2, etc)
‘forwardReads’. The full path for the forward reads. e.g. /home/myproject/fastq/sample1_S22_L001_R1.fastq.gz
‘reverseReads’. The full path for the forward reads. e.g. /home/myproject/fastq/sample1_S22_L001_R2.fastq.gz
This file can be created with Excel and then saved as a tab-delimited file (File → Save as → ‘manifest.txt’ → Text (Tab delimited)), then copied across to your NextFlow project directory (using WinSCP, Cyberduck, etc). Example:
sampleID | forwardReads | reverseReads |
|---|---|---|
groupA_1 | /home/myproject/fastq/sample1_S22_L001_R1.fastq.gz | /home/myproject/fastq/sample1_S22_L001_R2.fastq.gz |
groupA_2 | /home/myproject/fastq/sample2_S23_L001_R1.fastq.gz | /home/myproject/fastq/sample2_S23_L001_R2.fastq.gz |
etc… |
Creating a manifest file at the command line
As mentioned above, spelling mistakes or extra characters in the file paths will cause ampliseq to fail. One way to avoid this is to generate the manifest file on the command line using the tools awk and sed.
Below is an example of how to generate the manifest file. You may need to modify this, depending on how your files are named.
To create the manifest using awk, paste, sed:
List all the fastq files in the directory (both read pairs)
ls *_R1*.fastq.gz -lh | awk '{print $9}' > read1
ls *_R2*.fastq.gz -lh | awk '{print $9}' > read2
2. List the sample IDs. If the sample names are in the sample files, they can be extracted using sed. For example:
cat read1 | sed 's/_S.*//' > ID
The sample file names in this case are like such: ‘Raw8h_S10_L001_R1_001.fastq.gz’
The sample ID is ‘Raw8h’. The above sed command strips the characters after ‘_S’, leaving just the ID name. Depending on how your sample files are named, you can create a list of your sample IDs by modifying the above sed command.
3. Paste these together with the sample file directory prepended and tab delimiters. Output as ‘manifest.txt’.
paste ID read1 read2 | awk '{print $1 "\t" "/path/to/your/nextflow/myproject/fastq/" $2 "\t" "/path/to/your/nextflow/myproject/fastq/" $3}' > manifest.txt
Make sure you them manually add the 3 column names at the top of each column: ‘sampleID’, ‘forwardReads’ and ‘reverseReads’ (e.g. use a text editor like nano, or download the file and modify it in Excel, then re-upload it).
Finally, copy the created manifest.txt to the directory where you will be running ampliseq from.
Metadata file
This is a tab separated values file (.tsv) that is required by QIIME2 to compare taxonomic diversity with phenotype (e.g. how diversity varies per experimental treatment). It contains the same sample IDs found in the manifest file and a column for each category of metadata you have for the samples. This may include sequence barcodes, experimental treatment group (e.g. high fat vs low fat) and any other measurements taken, such as age, date collected, tissue type, sex, collection location, weight, length, etc, etc, etc). QIIME2 will compare every metadata column with taxonomic results, then calculate and plot correlations and diversity indices. See here for more details:
https://docs.qiime2.org/2019.10/tutorials/metadata/
To get a better idea of the structure and format of the metadata file, you can download an example file from here:
https://data.qiime2.org/2019.10/tutorials/moving-pictures/sample_metadata.tsv
This file can also be created in Excel and saved as a tab-delimited file. It can be any format, but .tsv is the default (File → Save as → ‘metadata.tsv’ → Text (Tab delimited))
nextflow.config file
Create a custom nextflow.config file
module load nano nano nextflow.config
Edit the newly created ‘nextflow.config’ file to contain something like the following:
params {
max_cpus=32
max_memory=512.GB
max_time = 48.h
FW_primer = "CCTACGGGNGGCWGCAG"
RV_primer = "GACTACHVGGGTATCTAATCC"
metadata = "metadata.txt"
manifest = "manifest.txt"
reference_database = "Silva_132_release.zip"
retain_untrimmed = true
}
NOTE: This is an example nextflow.config file. Don’t simply copy and paste the above. You’ll need to modify it to reflect the primers you used to generate your sequences. FW_primer = and RV_primer =
The remaining lines can stay the same, presuming that you called your metadata file 'metadata.txt' and you have all the files in the directory where you will be running ampliseq from.
Running NextFlow’s ampliseq pipeline
Make sure Java is loaded (should be already loaded if you are continuing from the above steps, otherwise ‘module load java’) and that you have started an interactive PBS session (again, you should be in this if continuing from above)
nextflow run nf-core/ampliseq -profile singularity
Running the full pipeline may take a day or so, depending on how many people are using the HPC. If ampliseq fails for whatever reason (you’ll see red script and the error message in your console) it’s a good idea to resume the ampliseq run a couple of times, just to make sure it’s not a transitory error, which occurs sometimes. To do this, add --resume to the end of the command.
nextflow run nf-core/ampliseq -profile singularity --resume
If ampliseq continues to fail after you have run --resume a couple of times, contact us at eResearch support: eresearch@qut.edu.au
Results
Ampliseq generates multiple output directories in a main ‘Results’ directory, including tables, figures, analysis results, etc. See here for details:
https://nf-co.re/ampliseq/1.1.3/output
These results can be used for further downstream analysis - for example plotting the table of taxon abundance - in Excel, R or other programs.
An example of how to analyse the results in R is here:
Downstream analysis of NextFlow ampliseq output (16S amplicon analysis)