nf-core/eager
A fully reproducible and state-of-the-art ancient DNA analysis pipeline
2.3.1
). The latest
stable release is
2.5.2
.
22.10.6
.
Learn more.
Introduction
Running the pipeline
Quick Start
Before you start you should change into the output directory you wish your results to go in. This will guarantee, that when you start the Nextflow job, it will place all the log files and ‘working’ folders in the corresponding output directory, (and not wherever else you may have executed the run from)
The typical command for running the pipeline is as follows:
where the reads are from FASTQ files of the same pairing.
This will launch the pipeline with the docker
configuration profile. See below
for more information about profiles.
Note that the pipeline will create the following files in your working directory:
To see the the nf-core/eager pipeline help message run: nextflow run nf-core/eager --help
If you want to configure your pipeline interactively using a graphical user
interface, please visit nf-co.re
launch. Select the eager
pipeline and
the version you intend to run, and follow the on-screen instructions to create a
config for your pipeline run.
Updating the pipeline
When you run the above command, Nextflow automatically pulls the pipeline code from GitHub and stores it as a cached version. When running the pipeline after this, it will always use the cached version if available - even if the pipeline has been updated since. To make sure that you’re running the latest version of the pipeline, make sure that you regularly update the cached version of the pipeline:
Reproducibility
It’s a good idea to specify a pipeline version when running the pipeline on your data. This ensures that a specific version of the pipeline code and software are used when you run your pipeline. If you keep using the same tag, you’ll be running the same version of the pipeline, even if there have been changes to the code since.
First, go to the nf-core/eager releases page and find the latest version number - numeric only (eg. 1.3.1
). Then specify this when running the pipeline with -r
(one hyphen) - eg. -r 1.3.1
.
This version number will be logged in reports when you run the pipeline, so that you’ll know what you used when you look back in the future.
Additionally, nf-core/eager pipeline releases are named after Swabian German Cities. The first release V2.0 is named “Kaufbeuren”. Future releases are named after cities named in the Swabian league of Cities.
Automatic Resubmission
By default, if a pipeline step fails, nf-core/eager will resubmit the job with twice the amount of CPU and memory. This will occur two times before failing.
Core Nextflow arguments
NB: These options are part of Nextflow and use a single hyphen (pipeline parameters use a double-hyphen).
-profile
Use this parameter to choose a configuration profile. Profiles can give configuration presets for different compute environments.
Several generic profiles are bundled with the pipeline which instruct the pipeline to use software packaged using different methods (Docker, Singularity, Podman, Conda) - see below.
We highly recommend the use of Docker or Singularity containers for full pipeline reproducibility, however when this is not possible, Conda is also supported.
The pipeline also dynamically loads configurations from https://github.com/nf-core/configs when it runs, making multiple config profiles for various institutional clusters available at run time. For more information and to see if your system is available in these configs please see the nf-core/configs documentation.
Note that multiple profiles can be loaded, for example: -profile test,docker
- the order of arguments is important!
They are loaded in sequence, so later profiles can overwrite earlier profiles.
If -profile
is not specified, the pipeline will run locally and expect all software to be installed and available on the PATH
. This is not recommended.
docker
- A generic configuration profile to be used with Docker
- Pulls software from Docker Hub:
nfcore/eager
singularity
- A generic configuration profile to be used with Singularity
- Pulls software from Docker Hub:
nfcore/eager
podman
- A generic configuration profile to be used with Podman
- Pulls software from Docker Hub:
nfcore/eager
conda
test_tsv
- A profile with a complete configuration for automated testing
- Includes links to test data so needs no other parameters
Important: If running nf-core/eager on a cluster - ask your system administrator what profile to use.
Institution Specific Profiles These are profiles specific to certain HPC clusters, and are centrally maintained at nf-core/configs. Those listed below are regular users of nf-core/eager, if you don’t see your own institution here check the nf-core/configs repository.
uzh
- A profile for the University of Zurich Research Cloud
- Loads Singularity and defines appropriate resources for running the pipeline.
binac
- A profile for the BinAC cluster at the University of Tuebingen 0 Loads Singularity and defines appropriate resources for running the pipeline
shh
- A profile for the S/CDAG cluster at the Department of Archaeogenetics of the Max Planck Institute for the Science of Human History
- Loads Singularity and defines appropriate resources for running the pipeline
Pipeline Specific Institution Profiles There are also pipeline-specific institution profiles. I.e., we can also offer a profile which sets special resource settings to specific steps of the pipeline, which may not apply to all pipelines. This can be seen at nf-core/configs under conf/pipelines/eager/.
We currently offer a nf-core/eager specific profile for
shh
- A profiler for the S/CDAG cluster at the Department of Archaeogenetics of the Max Planck Institute for the Science of Human History
- In addition to the nf-core wide profile, this also sets the MALT resources to match our commonly used databases
Further institutions can be added at nf-core/configs. Please ask the eager developers to add your institution to the list above, if you add one!
If you are likely to be running nf-core
pipelines regularly it may be a good idea to request that your custom config file is uploaded to the nf-core/configs
git repository. Before you do this please can you test that the config file works with your pipeline of choice using the -c
parameter (see definition above). You can then create a pull request to the nf-core/configs
repository with the addition of your config file, associated documentation file (see examples in nf-core/configs/docs
), and amending nfcore_custom.config
to include your custom profile.
If you have any questions or issues please send us a message on Slack on the #configs
channel.
-resume
Specify this when restarting a pipeline. Nextflow will used cached results from any pipeline steps where the inputs are the same, continuing from where it got to previously.
You can also supply a run name to resume a specific run: -resume [run-name]
. Use the nextflow log
command to show previous run names.
-c
Specify the path to a specific config file (this is a core Nextflow command). See the nf-core website documentation for more information.
Custom resource requests
Each step in the pipeline has a default set of requirements for number of CPUs, memory and time. For most of the steps in the pipeline, if the job exits with an error code of 143
(exceeded requested resources) it will automatically resubmit with higher requests (2 x original, then 3 x original). If it still fails after three times then the pipeline is stopped.
Whilst these default requirements will hopefully work for most people with most
data, you may find that you want to customise the compute resources that the
pipeline requests. You can do this by creating a custom config file. For
example, to give the workflow process star
32GB of memory, you could use the
following config:
See the main Nextflow documentation for more information.
If you are likely to be running nf-core
pipelines regularly it may be a good
idea to request that your custom config file is uploaded to the
nf-core/configs
git repository. Before you do this please can you test that
the config file works with your pipeline of choice using the -c
parameter (see
definition below). You can then create a pull request to the nf-core/configs
repository with the addition of your config file, associated documentation file
(see examples in
nf-core/configs/docs
),
and amending
nfcore_custom.config
to include your custom profile.
If you have any questions or issues please send us a message on
Slack on the #configs
channel.
-name
Name for the pipeline run. If not specified, Nextflow will automatically generate a random mnemonic.
This is used in the MultiQC report (if not default) and in the summary HTML / e-mail (always).
NB: Single hyphen (core Nextflow option)
Running in the background
Nextflow handles job submissions and supervises the running jobs. The Nextflow process must run until the pipeline is finished.
The Nextflow -bg
flag launches Nextflow in the background, detached from your terminal so that the workflow does not stop if you log out of your session. The logs are saved to a file.
Alternatively, you can use screen
/ tmux
or similar tool to create a detached session which you can log back into at a later time.
Some HPC setups also allow you to run nextflow within a cluster job submitted your job scheduler (from where it submits more jobs).
To create a screen session:
To disconnect, press ctrl+a
then d
.
To reconnect, type:
to end the screen session while in it type exit
.
Nextflow memory requirements
In some cases, the Nextflow Java virtual machines can start to request a large amount of memory.
We recommend adding the following line to your environment to limit this (typically in ~/.bashrc
or ~./bash_profile
):
Input Specifications
There are two possible ways of supplying input sequencing data to nf-core/eager. The most efficient but more simplistic is supplying direct paths (with wildcards) to your FASTQ or BAM files, with each file or pair being considered a single library and each one run independently. TSV input requires creation of an extra file by the user and extra metadata, but allows more powerful lane and library merging.
Direct Input Method
This method is where you specify with --input
, the path locations of FASTQ (optionally gzipped) or BAM file(s). This option is mutually exclusive to the TSV input method, which is used for more complex input configurations such as lane and library merging.
When using the direct method of --input
you can specify one or multiple samples in one or more directories files. File names must be unique, even if in different directories.
By default, the pipeline assumes you have paired-end data. If you want to run single-end data you must specify --single_end
For example, for a single set of FASTQs, or multiple paired-end FASTQ files in one directory, you can specify:
If you have multiple files in different directories, you can use additional wildcards (*
) e.g.:
⚠️ It is not possible to run a mixture of single-end and paired-end files in one run with the paths
--input
method! Please see the TSV input method for possibilities.
Please note the following requirements:
- Valid file extensions:
.fastq.gz
,.fastq
,.fq.gz
,.fq
,.bam
. - The path must be enclosed in quotes
- The path must have at least one
*
wildcard character - When using the pipeline with paired end data, the path must use
{1,2}
notation to specify read pairs. - Files names must be unique, having files with the same name, but in different directories is not sufficient
- This can happen when a library has been sequenced across two sequencers on the same lane. Either rename the file, try a symlink with a unique name, or merge the two FASTQ files prior input.
- Due to limitations of downstream tools (e.g. FastQC), sample IDs may be truncated after the first
.
in the name, Ensure file names are unique prior to this! - For input BAM files you should provide a small decoy reference genome with pre-made indices, e.g. the human mtDNA or phiX genome, for the mandatory parameter
--fasta
in order to avoid long computational time for generating the index files of the reference genome, even if you do not actually need a reference genome for any downstream analyses.
TSV Input Method
Alternatively to the direct input method, you can supply to --input
a path to a TSV file that contains paths to FASTQ/BAM files and additional metadata. This allows for more complex procedures such as merging of sequencing data across lanes, sequencing runs, sequencing configuration types, and samples.
Only different libraries from a single sample that have been BAM trimmed will be merged together. Rescaled or PMD filtered libraries will not be merged prior genotyping as each library may have a different model applied to it and have their own biases (i.e. users may need to play around with settings to get the damage-removal optimal).
The use of the TSV --input
method is recommended when performing more complex procedures such as lane or library merging. You do not need to specify --single_end
, --bam
, --colour_chemistry
, -udg_type
etc. when using TSV input - this is defined within the TSV file itself. You can only supply a single TSV per run (i.e. --input '*.tsv'
will not work).
This TSV should look like the following:
Sample_Name | Library_ID | Lane | Colour_Chemistry | SeqType | Organism | Strandedness | UDG_Treatment | R1 | R2 | BAM |
---|---|---|---|---|---|---|---|---|---|---|
JK2782 | JK2782 | 1 | 4 | PE | Mammoth | double | full | https://github.com/nf-core/test-datasets/raw/eager/testdata/Mammoth/fastq/JK2782_TGGCCGATCAACGA_L008_R1_001.fastq.gz.tengrand.fq.gz | https://github.com/nf-core/test-datasets/raw/eager/testdata/Mammoth/fastq/JK2782_TGGCCGATCAACGA_L008_R2_001.fastq.gz.tengrand.fq.gz | NA |
JK2802 | JK2802 | 2 | 2 | SE | Mammoth | double | full | https://github.com/nf-core/test-datasets/raw/eager/testdata/Mammoth/fastq/JK2802_AGAATAACCTACCA_L008_R1_001.fastq.gz.tengrand.fq.gz | https://github.com/nf-core/test-datasets/raw/eager/testdata/Mammoth/fastq/JK2802_AGAATAACCTACCA_L008_R2_001.fastq.gz.tengrand.fq.gz | NA |
A template can be taken from here.
⚠️ Cells must not contain spaces before or after strings, as this will make the TSV unreadable by nextflow. Strings containing spaces should be wrapped in quotes.
When using TSV_input, nf-core/eager will merge FASTQ files of libraries with the same Library_ID
but different Lanes
values after adapter clipping (and merging), assuming all other metadata columns are the same. If you have the same Library_ID
but with different SeqType
, this will be merged directly after mapping prior BAM filtering. Finally, it will also merge BAM files with the same Sample_ID
but different Library_ID
after duplicate removal, but prior to genotyping. Please see caveats to this below.
Column descriptions are as follows:
- Sample_Name: A text string containing the name of a given sample of which there can be multiple libraries. All libraries with the same sample name and same SeqType will be merged after deduplication.
- Library_ID: A text string containing a given library, which there can be multiple sequencing lanes (with the same SeqType).
- Lane: A number indicating which lane the library was sequenced on. Files from the libraries sequenced on different lanes (and different SeqType) will be concatenated after read clipping and merging.
- Colour Chemistry A number indicating whether the Illumina sequencer the library was sequenced on was a 2 (e.g. Next/NovaSeq) or 4 (Hi/MiSeq) colour chemistry machine. This informs whether poly-G trimming (if turned on) should be performed.
- SeqType: A text string of either ‘PE’ or ‘SE’, specifying paired end (with both an R1 [or forward] and R2 [or reverse]) and single end data (only R1 [forward], or BAM). This will affect lane merging if different per library.
- Organism: A text string of the organism name of the sample or ‘NA’. This currently has no functionality and can be set to ‘NA’, but will affect lane/library merging if different per library
- Strandedness: A text string indicating whether the library type is’single’ or ‘double’. This will affect lane/library merging if different per library.
- UDG_Treatment: A text string indicating whether the library was generated with UDG treatment - either ‘full’, ‘half’ or ‘none’. Will affect lane/library merging if different per library.
- R1: A text string of a file path pointing to a forward or R1 FASTQ file. This can be used with the R2 column. File names must be unique, even if they are in different directories.
- R2: A text string of a file path pointing to a reverse or R2 FASTQ file, or ‘NA’ when single end data. This can be used with the R1 column. File names must be unique, even if they are in different directories.
- BAM: A text string of a file path pointing to a BAM file, or ‘NA’. Cannot be specified at the same time as R1 or R2, both of which should be set to ‘NA’
For example, the following TSV table:
Sample_Name | Library_ID | Lane | Colour_Chemistry | SeqType | Organism | Strandedness | UDG_Treatment | R1 | R2 | BAM |
---|---|---|---|---|---|---|---|---|---|---|
JK2782 | JK2782 | 7 | 4 | PE | Mammoth | double | full | data/JK2782_TGGCCGATCAACGA_L007_R1_001.fastq.gz.tengrand.fq.gz | data/JK2782_TGGCCGATCAACGA_L007_R2_001.fastq.gz.tengrand.fq.gz | NA |
JK2782 | JK2782 | 8 | 4 | PE | Mammoth | double | full | data/JK2782_TGGCCGATCAACGA_L008_R1_001.fastq.gz.tengrand.fq.gz | data/JK2782_TGGCCGATCAACGA_L008_R2_001.fastq.gz.tengrand.fq.gz | NA |
JK2802 | JK2802 | 7 | 4 | PE | Mammoth | double | full | data/JK2802_AGAATAACCTACCA_L007_R1_001.fastq.gz.tengrand.fq.gz | data/JK2802_AGAATAACCTACCA_L007_R2_001.fastq.gz.tengrand.fq.gz | NA |
JK2802 | JK2802 | 8 | 4 | SE | Mammoth | double | full | data/JK2802_AGAATAACCTACCA_L008_R1_001.fastq.gz.tengrand.fq.gz | NA | NA |
will have the following effects:
- After AdapterRemoval, and prior to mapping, FASTQ files from lane 7 and lane 8 with the same
SeqType
(and all other metadata columns) will be concatenated together for each Library. - After mapping, and prior BAM filtering, BAM files with different
SeqType
(but with all other metadata columns the same) will be merged together for each Library. - After duplicate removal, BAM files with different
Library_ID
s but with the sameSample_Name
and the sameUDG_Treatment
will be merged together. - If BAM trimming is turned on, all post-trimming BAMs (i.e. non-UDG and half-UDG ) will be merged with UDG-treated (untreated) BAMs, if they have the same
Sample_Name
.
Note the following important points and limitations for setting up:
- The TSV must use actual tabs (not spaces) between cells.
- File names must be unique regardless of file path, due to risk of over-writing (see: https://github.com/nextflow-io/nextflow/issues/470).
- If it is ‘too late’ and you already have duplicate file names, a workaround is to concatenate the FASTQ files together and supply this to a nf-core/eager run. The only downside is that you will not get independent FASTQC results for each file.
- Lane IDs must be unique for each sequencing of each library.
- If you have a library sequenced e.g. on Lane 8 of two HiSeq runs, you can give a fake lane ID (e.g. 20) for one of the FASTQs, and the libraries will still be processed correctly.
- This also applies to the SeqType column, i.e. with the example above, if one run is PE and one run is SE, you need to give fake lane IDs to one of the runs as well.
- All BAM files must be specified as
SE
underSeqType
.- You should provide a small decoy reference genome with pre-made indices, e.g. the human mtDNA or phiX genome, for the mandatory parameter
--fasta
in order to avoid long computational time for generating the index files of the reference genome, even if you do not actually need a reference genome for any downstream analyses.
- You should provide a small decoy reference genome with pre-made indices, e.g. the human mtDNA or phiX genome, for the mandatory parameter
- nf-core/eager will only merge multiple lanes of sequencing runs with the same single-end or paired-end configuration
- Accordingly nf-core/eager will not merge lanes of FASTQs with BAM files (unless you use
--run_convertbam
), as only FASTQ files are lane-merged together. - Same libraries that are sequenced on different sequencing configurations (i.e single- and paired-end data), will be merged after mapping and will always be considered ‘paired-end’ during downstream processes
- Important running DeDup in this context is not recommended, as PE and SE data at the same position will not be evaluated as duplicates. Therefore not all duplicates will be removed.
- When you wish to run PE/SE data together
-dedupper markduplicates
is therefore preferred. - An error will be thrown if you try to merge both PE and SE and also supply
--skip_merging
. - If you truly want to mix SE data and PE data but using mate-pair info for PE mapping, please run FASTQ preprocessing mapping manually and supply BAM files for downstream processing by nf-core/eager
- If you regularly want to run the situation above, please leave a feature request on github.
- DamageProfiler, NuclearContamination, MTtoNucRatio and PreSeq are performed on each unique library separately after deduplication (but prior same-treated library merging).
- nf-core/eager functionality such as
--run_trim_bam
will be applied to only non-UDG (UDG_Treatment: none) or half-UDG (UDG_Treatment: half) libraries. - Qualimap is run on each sample, after merging of libraries (i.e. your values will reflect the values of all libraries combined - after being damage trimmed etc.). - Genotyping will be typically performed on each
sample
independently, as normally all libraries will have been merged together. However, if you have a mixture of single-stranded and double-stranded libraries, you will normally need to genotype separately. In this case you must give each the SS and DS libraries distinctSample_IDs
; otherwise you will receive afile collision
error in steps such assexdeterrmine
, and then you will need to merge these yourself. We will consider changing this behaviour in the future if there is enough interest.
Clean up
Once a run has completed, you will have lots of (some very large) intermediate
files in your output directory. These are stored within the directory named
work
.
After you have verified your run completed correctly and everything in the module output directories are present as you expect and need, you can perform a clean-up.
Important: Once clean-up is completed, you will not be able to re-rerun the pipeline from an earlier step and you’ll have to re-run from scratch.
While in your output directory, firstly verify you’re only deleting files stored
in work/
with the dry run command:
⚠️ some institutional profiles already have clean-up on successful run completion turned on by default.
If you’re ready, you can then remove the files with
This will make your system administrator very happy as you will halve the hard drive footprint of the run, so be sure to do this!
Troubleshooting and FAQs
I get a file name collision error during merging
When using TSV input, nf-core/eager will attempt to merge all Lanes
of a
Library_ID
, or all files with the same Library_ID
or Sample_ID
. However,
if you have specified the same Lane
or Library_ID
for two sets of FASTQ
files you will likely receive an error such as
In this case: for lane merging errors, you can give ‘fake’ lane IDs to ensure
they are unique (e.g. if one library was sequenced on Lane 8 of two HiSeq runs,
specify lanes as 8 and 16 for each FASTQ file respectively). For library merging
errors, you must modify your Library_ID
s accordingly, to make them unique.
Tutorials
Tutorial - How to investigate a failed run
As with most pipelines, nf-core/eager can sometimes fail either through a problem with the pipeline itself, but also sometimes through an issue of the program being run at the given step.
To help try and identify what has caused the error, you can perform the following steps before reporting the issue:
1a Nextflow reports an ‘error executing process’ with command error
Firstly, take a moment to read the terminal output that is printed by an nf-core/eager command.
When reading the following, you can see that the actual command failed. When
you get this error, this would suggest that an actual program used by the
pipeline has failed. This is identifiable when you get an exit status
and a
Command error:
, the latter of which is what is reported by the failed program
itself.
If you find it is a common error try and fix it yourself by changing your options in your nf-core/eager run - it could be a configuration error on your part. However in some cases it could be an error in the way we’ve set up the process in nf-core/eager.
To further investigate, go to step 2.
1b Nextflow reports an ‘error executing process’ with no command error
Alternatively, you may get an error with Nextflow itself. The most common one would be a ‘process fails’ and it looks like the following.
However in this case, there is no exit status
or Command error:
message. In
this case this is a Nextflow issue.
The example above is because a user has specified multiple sequencing runs of different libraries but with the same library name. In this case Nextflow could not identify which is the correct file to merge because they have the same name.
This again can also be a user or Nextflow error, but the errors are often more abstract and less clear how to solve (unless you are familiar with Nextflow).
Try to investigate a bit further and see if you can understand what the error refers to, but if you cannot - please ask on the #eager channel on the nf-core slack or leave a github issue.
2 Investigating an failed process’s work/
directory
If you haven’t found a clear solution to the failed process from the reported errors, you can next go into the directory where the process was working in, and investigate the log and error messages that are produced by each command of the process.
For example, in the error in 1a you can see the following line
A shortened version of the ‘hash’ directory ID can also be seen in your terminal while the pipeline is running in the square brackets at the beginning of each line.
If you change into this with cd
and run ls -la
you should see a collection
of normal files, symbolic links (symlinks) and hidden files (indicated with .
at the beginning of the file name).
- Symbolic links: are typically input files from previous processes.
- Normal files: are typically successfully completed output files from some of some of the commands in the process
- Hidden files are Nextflow generated files and include the submission commands as well as log files
When you have an error run, you can firstly check the contents of the output
files to see if they are empty or not (e.g. with cat
or zcat
),
interpretation of which will depend on the program thus dependent on the user
knowledge.
Next, you can investigate .command.err
and .command.out
, or .command.log
.
These represent the standard out or error (in the case of .log
, both combined)
of all the commands/programs in the process - i.e. what would be printed to
screen if you were running the command/program yourself. Again, view these with
e.g. cat
and see if you can identify the error of the program itself.
Finally, you can also try running the commands yourself. You can firstly try
to do this by loading your given nf-core/eager environment (e.g. singularity shell /\<path\>/\<to\>/nf-core-eager-X-X-X.img
or conda activate nf-core-eager-X.X.X
), then running bash .command.sh
.
If this doesn’t work, this suggests either there is something wrong with the
nf-core/eager environment configuration, or there is still a problem with the
program itself. To confirm the former, try running the command within the
.command.sh
file (viewable with cat
) but with locally installed versions of
programs you may already have on your system. If the command still doesn’t work,
it is a problem with the program or your specified configuration. If it does
work locally, please report as a github
issue.
If it does, please ask the developer of the tool (although we will endeavour to help as much as we can via the nf-core slack in the #eager channel).
Tutorial - What are Profiles and How To Use Them
Tutorial Profiles - Background
A useful feature of Nextflow is the ability to use configuration profiles that can specify many default parameters and other settings on how to run your pipeline.
For example, you can use it to set your preferred mapping parameters, or specify where to keep Docker, Singularity or Conda environments, and which cluster scheduling system (and queues) your pipeline runs should normally use.
This are defined in .config
files, and these in-turn can contain different
profiles that can define parameters for different contexts.
For example, a .config
file could contain two profiles, one for
shallow-sequenced samples that uses only a small number of CPUs and memory e.g.
small
, and another for deep sequencing data, deep
, that allows larger
numbers of CPUs and memory. As another example you could define one profile
called loose
that contains mapping parameters to allow reads with aDNA damage
to map, and then another called strict
that reduces the likelihood of damaged
DNA to map and cause false positive SNP calls.
Within nf-core, there are two main levels of configs
- Institutional-level profiles: these normally define things like paths to common storage, resource maximums, scheduling system
- Pipeline-level profiles: these normally define parameters specifically for a pipeline (such as mapping parameters, turning specific modules on or off)
As well as allowing more efficiency and control at cluster or Institutional levels in terms of memory usage, pipeline-level profiles can also assist in facilitating reproducible science by giving a way for researchers to ‘publish’ their exact pipeline parameters in way other users can automatically re-run the pipeline with the pipeline parameters used in the original publication but on their own cluster.
To illustrate this, lets say we analysed our data on a HPC called ‘blue’ for which an institutional profile already exists, and for our analysis we defined a profile called ‘old_dna’. We will have run our pipeline with the following command
Then our colleague wished to recreate your results. As long as the
old_dna_profile.config
was published alongside your results, they can run the
same pipeline settings but on their own cluster HPC ‘purple’.
(where the old_dna
profile is defined in old_dna_profile.config
, and
hpc_purple
is defined on nf-core/configs)
This tutorial will describe how to create and use profiles that can be used by or from other researchers.
Tutorial Profiles - Inheritance Rules
Tutorial Profiles - Profiles
An important thing to understand before you start writing your own profile is
understanding ‘inheritance’ of profiles when specifying multiple profiles, when
using nextflow run
.
When specifying multiple profiles, parameters defined in the profile in the first position will overwrite those in the second, and everything defined in the first and second will overwrite everything in a third.
This can be illustrated as follows.
This would be translated as follows.
If your parameters looked like the following
Parameter | Resolved Parameters | my_paper | cluster | institution |
---|---|---|---|---|
—executor | singularity | <none> | <none> | singularity |
—max_memory | 256GB | <none> | 256GB | 756GB |
—bwa_aln | 0.1 | 0.1 | 0.01 | <none> |
(where ‘<none>’ is a parameter not defined in a given profile.)
You can see that my_paper
inherited the 0.1
parameter over the 0.01
defined in the cluster
profile.
⚠️ You must always check if parameters are defined in any ‘upstream’ profiles that have been set by profile administrators that you may be unaware of. This is make sure there are no unwanted or unreported ‘defaults’ away from original nf-core/eager defaults.
Tutorial Profiles - Configuration Files
⚠️ This section is only needed for users that want to set up institutional-level profiles.
Expand to view
In actuality, a nf-core/eager run already contains many configs and profiles, and will normally use multiple configs profiles in a single run. Multiple configuration and profiles files can be used, and each new one selected will inherit all the previous one’s parameters, and the parameters in the new one will then overwrite any that have been changed from the original.
This can be visualised here
Using the example given in the background, if
the hpc_blue
profile has the following pipeline parameters set
However, the profile old_dna
has only the following parameter
Then running the pipeline with the profiles in the order of the following run command:
In the background, any parameters in the pipeline’s nextflow.config
(containing default parameters) will be overwritten by the
old_dna_profile.config
. In addition, the old_dna
profile will overwrite
any parameters set in the config but outside the profile definition of
old_dna_profile.config
.
Therefore, the final profile used by your given run would look like:
You can see here that markduplicates
has not changed as originally defined in
the hpc_blue
profile, but the mapper
parameter has been changed from
bwamem
to bwaaln
, as specified in the old_dna
profile.
The order of loading of different configuration files can be seen here:
Loading Order | Configuration File |
---|---|
1 | nextflow.config in your current directory, |
2 | (if using a script for nextflow run ) a nextflow.config in the directory the script is located |
3 | config stored in your human directory under ~/.nextflow/ |
4 | <your_file>.config if you specify in the nextflow run command with -c |
5 | general nf-core institutional configurations stored at nf-core/configs |
6 | pipeline-specific nf-core institutional configurations at nf-core/configs |
This loading order of these .config
files will not normally affect the
settings you use for the pipeline run itself; -profiles
are normally more
important. However this is good to keep in mind when you need to debug profiles
if your run does not use the parameters you expect.
⚠️ It is also possible to ignore every configuration file other when specifying a custom
.config
file by using-C
(capital C) instead of-c
(which inherits previously specify parameters)
Another thing that is important to note is that if a specific profile is
specified in nextflow run
, this replaces any ‘global’ parameter that is
specified within the config file (but outside a profile) itself - regardless
of profile order (see above).
For example, see the example adapted from the SHH nf-core/eager pipeline-specific configuration.
This pipeline-specific profile is automatically loaded if nf-core/eager detects
we are running eager, and that we specified the profile as shh
.
If you run with nextflow run -profile shh
to specify to use an
institutional-level nf-core config, the parameters will be read as --bwaalnn 0.04
and --bwaalnl 32
as these are the defaults ‘fall back’ params as
indicated in the example above.
If you specify as nextflow run -profile pathogen_loose,shh
, as expected
Nextflow will resolve the two parameters as 0.01
and 16
.
Importantly however, if you specify -profile shh,pathogen_loose
the
pathogen_loose
profile will still take precedence over just the
‘global’ params.
Equally, a process-level defined parameter (within the nf-core/eager code
itself) will take precedence over the fallback parameters in the config
file.
This is also described in the Nextflow documentation
here
This is because selecting a profile
will always take precedence over the
values specified in a config file, but outside of a profile.
Tutorial Profiles - Writing your own profile
We will now provide an example of how to write, use and share a project specific profile. We will use the example of Andrades Valtueña et al. 2016.
In it they used the original EAGER (v1) to map DNA from ancient DNA to the genome of the bacterium Yersinia pestis.
Now, we will generate a profile, that, if they were using nf-core/eager they could share with other researchers.
In the methods they described the following:
… reads mapped to Y. pestis CO92 reference with BWA aln (-l 16, -n 0.01, hereby referred to as non-UDG parameters). Reads with mapping quality scores lower than 37 were filtered out. PCR duplicates were removed with MarkDuplicates.”
Furthermore, in their ‘Table 1’ they say they used the NCBI Y. pestis genome ‘NC_003143.1’, which can be found on the NCBI FTP server at: https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/009/065/GCF_000009065.1_ASM906v1/GCF_000009065.1_ASM906v1_genomic.fna.gz
To make a profile with these parameters for use with nf-core/eager we first need to open a text editor, and define a Nextflow ‘profile’ block.
Next we need to define the name of the profile. This is what we would write in
-profile
. Lets call this AndradesValtuena2018.
Now we need to make a params
‘scope’. This means these are the parameters you
specifically pass to nf-core/eager itself (rather than Nextflow configuration
parameters).
You should generally not add non-params
scopes
in profiles for a specific project. This is because these will normally modify
the way the pipeline will run on the computer (rather than just nf-core/eager
itself, e.g. the scheduler/executor or maximum memory available), and thus not
allow other researchers to reproduce your analysis on their own
computer/clusters.
Now, as a cool little trick, we can use a couple of nf-core specific parameters
that can help you keep track which profile you are using when running the
pipeline. The config_profile_description
and config_profile_contact
profiles
are displayed in the console log when running the pipeline. So you can use these
to check if your profile loaded as expected. These are free text boxes so you
can put what you like.
Now we can add the specific nf-core/eager parameters that will modify the mapping and deduplication parameters in nf-core/eager.
Once filled in, we can save the file as AndradesValtuena2018.config
. This you
can use yourself, or upload alongside your publication for others to use.
To use the profile you just need to specify the file containing the profile you wish to use, and then the profile itself.
For example, Aida (Andrades Valtueña) on her cluster sdag
at the MPI-SHH
(shh
) in Jena could run the following:
Then a colleague at a different institution, such as the SciLifeLab, could run the same profile on the UPPMAX cluster in Uppsala with:
And that’s all there is to it. Of course you should always check that there are no other ‘default’ parameters for your given pipeline are defined in any pipeline-specific or institutional profiles. This ensures that someone re-running the pipeline with your settings is as close to the nf-core/eager defaults as possible, and only settings specific to your given project are used. If there are ‘upstream’ defaults, you should explicitly specify these in your project profile.
Tutorial - How to set up nf-core/eager for human population genetics
Tutorial Human Pop-Gen - Introduction
This tutorial will give a basic example on how to set up nf-core/eager to perform initial screening of samples in the context of ancient human population genetics research.
⚠️ this tutorial does not describe how to install and set up nf-core/eager For this please see other documentation on the nf-co.re website.
We will describe how to set up mapping of ancient sequences against the human reference genome to allow sequencing and library quality-control, estimation of nuclear contamination, genetic sex determination, and production of random draw genotypes in eigenstrat format for a specific set of sites, to be used in further analysis. For this example, I will be using the 1240k SNP set. This SNP set was first described in Mathieson et al. 2015 and contains various positions along the genome that have been extensively genotyped in present-day and ancient populations, and are therefore useful for ancient population genetic analysis. Capture techniques are often used to enrich DNA libraries for fragments, that overlap these SNPs, as is being assumed has been performed in this example.
⚠️ Please be aware that the settings used in this tutorial may not use settings nor produce files you would actually use in ‘real’ analysis. The settings are only specified for demonstration purposes. Please consult the your colleagues, communities and the literature for optimal parameters.
Tutorial Human Pop-Gen - Preparation
Prior setting up the nf-core/eager run, we will need:
- Raw sequencing data in FASTQ format
- Reference genome in FASTA format, with associated pre-made
bwa
,samtools
andpicard SequenceDictionary
indices (however note these can be made for you with nf-core/eager, but this can make a pipeline run take much longer!) - A BED file with the positions of the sites of interest.
- An eigenstrat formatted
.snp
file for the positions of interest.
We should also ensure we have the very latest version of the nf-core/eager pipeline so we have all latest bugfixes etc. In this case we will be using nf-core/eager version 2.2.0. You should always check on the nf-core website whether a newer release has been made (particularly point releases e.g. 2.2.1).
It is important to note that if you are planning on running multiple runs of nf-core/eager for a given project, that the version should be kept the same for all runs to ensure consistency in settings for all of your libraries.
Tutorial Human Pop-Gen - Inputs and Outputs
To start, lets make a directory where all your nf-core/eager related files for this run will go, and change into it.
The first part of constructing any nf-core/eager run is specifying a few generic parameters that will often be common across all runs. This will be which pipeline, version and profile we will use. We will also specify a unique name of the run to help us keep track of all the nf-core/eager runs you may be running.
For the -profile
parameter, I have indicated that I wish to use Singularity as
my software container environment, and I will use the MPI-SHH institutional
config as listed on
nf-core/configs,
using the profile for the ‘sdag’ cluster. These profiles specify settings
optimised for the specific cluster/institution, such as maximum memory available
or which scheduler queues to submit to. More explanations about configs and
profiles can be seen in the nf-core
website and the profile
tutorial.
Next we need to specify our input data. nf-core/eager can accept input FASTQs files in two main ways, either with direct paths to files (with wildcards), or with a Tab-Separate-Value (TSV) file which contains the paths and extra metadata. In this example, we will use the TSV method, as to simulate a realistic use-case, such as receiving paired-end data from an Illumina NextSeq of double-stranded libraries. Illumina NextSeqs sequence a given library across four different ‘lanes’, so for each library you will receive four FASTQ files. The TSV input method is more useful for this context, as it allows ‘merging’ of these lanes after preprocessing prior mapping (whereas direct paths will consider each pair of FASTQ files as independent libraries/samples).
Our TSV file will look something like the following:
You can see that we have a single line for each pair of FASTQ files representing
each Lane
, but the Sample_Name
and Library_ID
columns identify and group
them together accordingly. Secondly, as we have NextSeq data, we have specified
we have 2
for Colour_Chemistry
, which is important for downstream processing
(see below). See the nf-core/eager
parameter documentation above for more specifications on how to set up a
TSV file (e.g. why despite NextSeqs
only having 4 lanes, we go up to 8 in the example above).
Alongside our input TSV file, we will also specify the paths to our reference FASTA file and the corresponding indices.
We specify the paths to each reference genome and it’s corresponding tool
specific index. Paths should always be encapsulated in quotes to ensure Nextflow
evaluates them, rather than your shell! Also note that as bwa
generates
multiple index files, nf-core/eager takes a directory that must contain these
indices instead.
Note the difference between single and double
-
parameters. The former represent Nextflow flags, while the latter are nf-core/eager specific flags.
Finally, we can also specify the output directory and the Nextflow work/
directory (which contains ‘intermediate’ working files and directories).
Tutorial Human Pop-Gen - Pipeline Configuration
Now that we have specified the input data, we can start moving onto specifying
settings for each different module we will be running. As mentioned above, we
are pretending to run with NextSeq data, which is generated with a two-colour
imaging technique. What this means is when you have shorter molecules than the
number of cycles of the sequencing chemistry, the sequencer will repeatedly see
‘G’ calls (no colour) at the last few cycles, and you get long poly-G ‘tails’ on
your reads. We therefore will turn on the poly-G clipping functionality offered
by fastp
, and any pairs of files
indicated in the TSV file as having 2
in the Colour_Chemistry
column will be
passed to fastp
. We will not change the default minimum length of a poly-G
string to be clipped.
Since our input data is paired-end, we will be using DeDup
for duplicate
removal, which takes into account both the start and end of a merged read before
flagging it as a duplicate. To ensure this happens works properly we first need
to disable base quality trimming of collapsed reads within Adapter Removal. To
do this, we will provide the option --preserve5p
. Additionally, Dedup should
only be provided with merged reads, so we will need to provide the option
--mergedonly
here as well. We can then specify which dedupper we want to use
with --dedupper
.
We then need to specify the mapping parameters for this run. The default mapping
parameters of nf-core/eager are fine for the purposes of our run. Personally, I
like to set --bwaalnn
to 0.01
, (down from the default 0.04
) which reduces
the stringency in the number of allowed mismatches between the aligned sequences
and the reference.
We may also want to remove ambiguous sequences from our alignments, and also
remove off-target reads to speed up downstream processing (and reduce your
hard-disk footprint). We can do this with the samtools filter module to set a
mapping-quality filter (e.g. with a value of 25
to retain only slightly
ambiguous alignments that might occur from damage), and to indicate to discard
unmapped reads.
Next, we will set up trimming of the mapped reads to alleviate the effects of DNA
damage during genotyping. To do this we will activate trimming with
--run_trim_bam
. The libraries in this underwent ‘half’ UDG treatment. This
will generally restrict all remaining DNA damage to the first 2 base pairs of a
fragment. We will therefore use --bamutils_clip_half_udg_left
and
--bamutils_clip_half_udg_right
to trim 2bp on either side of each fragment.
To activate human sex determination (using
Sex.DetERRmine.py) we will
provide the option --run_sexdeterrmine
. Additionally, we will provide
sexdeterrmine with the BED file of our SNPs of interest using the
--sexdeterrmine_bedfile
flag. Here I will use the 1240k SNP set as an example.
This will cut down on computational time and also provide us with an
error bar around the relative coverage on the X and Y chromosomes.
If you wish to use the same bedfile to follow along with this tutorial,
you can download the file from here.
Similarly, we will activate nuclear contamination estimation with
--run_nuclear_contamination
. This process requires us to also specify the
contig name of the X chromosome in the reference genome we are using with
--contamination_chrom_name
. Here, we are using hs37d5, where the X chromosome
is simply named ‘X’.
Because nuclear contamination estimates can only be provided for males, it is
possible that we will need to get mitochondrial DNA contamination estimates for
any females in our dataset. This cannot be done within nf-core/eager (v2.2.0)
and we will need to do this manually at a later time. However, mtDNA
contamination estimates have been shown to only be reliable for nuclear
contamination when the ratio of mitochondrial to nuclear reads is low
(Furtwängler et al. 2018). We can
have nf-core/eager calculate that ratio for us with --run_mtnucratio
, and
providing the name of the mitochondrial DNA contig in our reference genome with
--mtnucratio_header
. Within hs37d5, the mitochondrial contig is named ‘MT’.
Finally, we need to specify genotyping parameters. First, we need to activate
genotyping with --run_genotyping
. It is also important to specify we wish to
use the trimmed data for genotyping, to avoid the effects of DNA damage. To
do this, we will specify the --genotyping_source
as 'trimmed'
. Then we can
specify the genotyping tool to use with --genotyping_tool
. We will be using
'pileupCaller'
to produce random draw genotypes in eigenstrat format. For this
process we will need to specify a BED file of the sites of interest (the same as
before) with --pileupcaller_bedfile
, as well as an eigenstrat formatted .snp
file of these sites that is specified with --pileupcaller_snpfile
.
With this, we are ready to submit! If running on a remote cluster/server, Make
sure to run this in a screen
session or similar, so that if you get a ssh
signal drop or want to log off, Nextflow will not crash.
Tutorial Human Pop-Gen - Results
Assuming the run completed without any crashes (if problems do occur, check
against #usage that all parameters are as expected, or
check the FAQ), we can now check our results in
results/
.
Tutorial Human Pop-Gen - MultiQC Report
In here there are many different directories containing different output files.
The first directory to check is the MultiQC/
directory. You should
find a multiqc_report.html
file. You will need to view this in a web browser,
so I recommend either mounting your server to your file browser, or downloading
it to your own local machine (PC/Laptop etc.).
Once you’ve opened this you can go through each section and evaluate all the results. You will likely want to check these for artefacts (e.g. weird damage patterns on the human DNA, or weirdly skewed coverage distributions).
For example, I normally look for things like:
General Stats Table:
- Do I see the expected number of raw sequencing reads (summed across each set of FASTQ files per library) that was requested for sequencing?
- Does the percentage of trimmed reads look normal for aDNA, and do lengths after trimming look short as expected of aDNA?
- Does ClusterFactor or ‘Dups’ look high (e.g. >2 or >10% respectively) suggesting over-amplified or badly preserved samples?
- Do the mapped reads show increased frequency of C>Ts on the 5’ end of molecules?
- Is the number of SNPs used for nuclear contamination really low for any individuals (e.g. < 100)? If so, then the estimates might not be very accurate.
FastQC (pre-AdapterRemoval):
- Do I see any very early drop off of sequence quality scores suggesting a problematic sequencing run?
- Do I see outlier GC content distributions?
- Do I see high sequence duplication levels?
AdapterRemoval:
- Do I see high numbers of singletons or discarded read pairs?
FastQC (post-AdapterRemoval):
- Do I see improved sequence quality scores along the length of reads?
- Do I see reduced adapter content levels?
Samtools Flagstat (pre/post Filter):
- Do I see outliers, e.g. with unusually high levels of human DNA, (indicative of contamination) that require downstream closer assessment? Are your samples exceptionally preserved? If not, a value higher than e.g. 50% might require your attention.
DeDup/Picard MarkDuplicates:
- Do I see large numbers of duplicates being removed, possibly indicating over-amplified or badly preserved samples?
DamageProfiler:
- Do I see evidence of damage on human DNA?
- High numbers of mapped reads but no damage may indicate significant modern contamination.
- Was the read trimming I specified enough to overcome damage effects?
SexDetERRmine:
- Do the relative coverages on the X and Y chromosome fall within the expected areas of the plot?
- Do all individuals have enough data for accurate sex determination?
- Do the proportions of autosomal/X/Y reads make sense? If there is an overrepresentation of reads within one bin, is the data enriched for that bin?
Detailed documentation and descriptions for all MultiQC modules can be seen in the the ‘Documentation’ folder of the results directory or here in the output documentation
If you’re happy everything looks good in terms of sequencing, we then look at specific directories to find any files you might want to use for downstream processing.
Note that when you get back to writing up your publication, all the versions of the tools can be found under the ‘nf-core/eager Software Versions’ section of the MultiQC report. But be careful! All tools in the container are listed, so you may have to remove some of them that you didn’t actually use in the set up.
For example, in this example, we have used: Nextflow, nf-core/eager, FastQC, AdapterRemoval, fastP, BWA, Samtools, endorS.py, DeDup, Qualimap, PreSeq, DamageProfiler, bamUtil, sexdeterrmine, angsd, MTNucRatioCalculator, sequenceTools, and MultiQC.
Citations to all used tools can be seen here
Tutorial Human Pop-Gen - Files for Downstream Analysis
You will find the eigenstrat dataset containing the random draw genotypes of
your run in the genotyping/
directory. Genotypes from double stranded
libraries, like the ones in this example, are found in the dataset
pileupcaller.double.{geno,snp,ind}.txt
, while genotypes for any single
stranded libraries will instead be in pileupcaller.single.{geno,snp,ind}.txt
.
Tutorial Human Pop-Gen - Clean up
Finally, I would recommend cleaning up your work/
directory of any
intermediate files (if your -profile
does not already do so). You can do this
by going to above your results/
and work/
directory, e.g.
and running
Tutorial Human Pop-Gen - Summary
In this this tutorial we have described an example on how to set up an nf-core/eager run to preprocess human aDNA for population genetic studies, preform some simple quality control checks, and generate random draw genotypes for downstream analysis of the data. Additionally, we described what to look for in the run summary report generated by MultiQC and where to find output files that can be used for downstream analysis.
Tutorial - How to set up nf-core/eager for metagenomic screening
Tutorial Metagenomics - Introduction
The field of archaeogenetics is now expanding out from analysing the genomes of single organisms but to whole communities of taxa. One particular example is of human associated microbiomes, as preserved in ancient palaeofaeces (gut) or dental calculus (oral). This tutorial will give a basic example on how to set up nf-core/eager to perform initial screening of samples in the context of ancient microbiome research.
⚠️ this tutorial does not describe how to install and set up nf-core/eager For this please see other documentation on the nf-co.re website.
We will describe how to set up mapping of ancient dental calculus samples against the human reference genome to allow sequencing and library quality-control, but additionally perform taxonomic profiling of the off-target reads from this mapping using MALT, and perform aDNA authentication with HOPS.
⚠️ Please be aware that the settings used in this tutorial may not use settings nor produce files you would actually use in ‘real’ analysis. The settings are only specified for demonstration purposes. Please consult the your colleagues, communities and the literature for optimal parameters.
Tutorial Metagenomics - Preparation
Prior setting up an nf-core/eager run for metagenomic screening, we will need:
- Raw sequencing data in FASTQ format
- Reference genome in FASTA format, with associated pre-made
bwa
,samtools
andpicard SequenceDictionary
indices - A MALT database of your choice (see MALT manual for set-up)
- A list of (NCBI) taxa containing well-known taxa of your microbiome (see below)
- HOPS resources
.map
and.tre
files (available here)
We should also ensure we have the very latest version of the nf-core/eager pipeline so we have all latest bugfixes etc. In this case we will be using nf-core/eager version 2.2.0. You should always check on the nf-core website whether a newer release has been made (particularly point releases e.g. 2.2.1).
It is important to note that if you are planning on running multiple runs of nf-core/eager for a given project, that the version should be kept the same for all runs to ensure consistency in settings for all of your libraries.
Tutorial Metagenomics - Inputs and Outputs
To start, lets make a directory where all your nf-core/eager related files for this run will go, and change into it.
The first part of constructing any nf-core/eager run is specifying a few generic parameters that will often be common across all runs. This will be which pipeline, version and profile we will use. We will also specify a unique name of the run to help us keep track of all the nf-core/eager runs you may be running.
For the -profile
parameter, I have indicated that I wish to use Singularity as
my software container environment, and I will use the MPI-SHH institutional
config as listed on
nf-core/configs,
and using the profile for the ‘sdag’ cluster. These profiles specify settings
optimised for the specific cluster/institution, such as maximum memory available
or which scheduler queues to submit to. More explanations about configs and
profiles can be seen in the nf-core
website and the profile
tutorial.
Next we need to specify our input data. nf-core/eager can accept input FASTQs files in two main ways, either with direct paths to files (with wildcards), or with a Tab-Separate-Value (TSV) file which contains the paths and extra metadata. In this example, we will use the TSV method, as to simulate a realistic use-case, such as receiving paired-end data from an Illumina NextSeq of double-stranded libraries. Illumina NextSeqs sequence a given library across four different ‘lanes’, so for each library you will receive four FASTQ files. The TSV input method is more useful for this context, as it allows ‘merging’ of these lanes after preprocessing prior mapping (whereas direct paths will consider each pair of FASTQ files as independent libraries/samples).
Our TSV file will look something like the following:
You can see that we have a single line for each pair of FASTQ files representing
each Lane
, but the Sample_Name
and Library_ID
columns identify and group
them together accordingly. Secondly, as we have NextSeq data, we have specified
we have 2
for Colour_Chemistry
, which is important for downstream processing
(see below). The other columns are less important for this particular context of
metagenomic screening. See the nf-core/eager usage
documentation for more specifications on how to set up a TSV file (e.g. why
despite NextSeqs only having 4 lanes, we go up to 8 in the example above).
Alongside our input TSV file, we will also specify the paths to our reference FASTA file and the corresponding indices.
We specify the paths to each reference genome and it’s corresponding tool
specific index. Paths should always be encapsulated in quotes to ensure Nextflow
evaluates them, rather than your shell! Also note that as bwa
generates
multiple index files, nf-core/eager takes a directory that must contain these
indices instead.
Note the difference between single and double
-
parameters. The former represent Nextflow flags, while double are nf-core/eager specific flags.
Finally, we can also specify the output directory and the Nextflow work/
directory (which contains ‘intermediate’ working files and directories).
Tutorial Metagenomics - Pipeline Configuration
Now that we have specified the input data, we can start moving onto specifying
settings for each different module we will be running. As mentioned above, we
are pretending to run with NextSeq data, which is generated with a two-colour
imaging technique. What this means is when you have shorter molecules than the
number of cycles of the sequencing chemistry, the sequencer will repeatedly see
‘G’ calls (no colour) at the last few cycles, and you get long poly-G ‘tails’ on
your reads. We therefore will turn on the poly-G clipping functionality offered
by fastp
, and any pairs of files
indicated in the TSV file as having 2
in the Colour_Chemistry
column will be
passed to fastp
. We will not change the default minimum length of a poly-G
string to be clipped.
We will keep the default settings for mapping etc. against the reference genome as we will only use this for sequencing quality control, however we now need to specify that we want to run metagenomic screening. To do this we firstly need to tell nf-core/eager what to do with the off target reads from the mapping.
nf-core/eager will now take all unmapped reads after mapping and convert the BAM
file back to FASTQ, which can be accepted by MALT. But of course, we also then
need to tell nf-core/eager we actually want to run MALT. We will also specify
the location of the pre-built database and which ‘min support’
method we want to use (this specifies the minimum number of alignments is needed
to a particular taxonomic node to be ‘kept’ in the MALT output files). Otherwise
we will keep all other parameters as default. For example using BlastN mode,
requiring a minimum of 85% identity, requiring at least 0.01% alignments for a
taxon to be saved (as specified with the --malt_min_support_mode
). More
documentation describing each parameters can be seen in the usage
documentation
Finally, to help quickly assess whether we our sample has taxa that are known to exist in (modern samples of) our expected microbiome, and that these alignments have indicators of true aDNA, we will run ‘maltExtract’ of the HOPS pipeline.
In the last parameters above we’ve specified the path to our list of taxa. This contains something like (for oral microbiomes):
We have also specified the path to the HOPS resources downloaded earlier, and that I want to turn off ‘destacking’ (removal of any read that overlaps the positions of another - something only recommended to keep on when you have high coverage data).
With this, we are ready to submit! If running on a remote cluster/server, Make
sure to run this in a screen
session or similar, so that if you get a ssh
signal drop or want to log off, Nextflow will not crash.
Tutorial Metagenomics - Results
Assuming the run completed without any crashes (if problems do occur, check
against usage that all parameters are as expected, or check
the FAQ), we can now check our results in
results/
.
Tutorial Metagenomics - MultiQC Report
In here there are many different directories containing different output files.
The first directory to check is the MultiQC/
directory. You should
find a multiqc_report.html
file. You will need to view this in a web browser,
so I recommend either mounting your server to your file browser, or downloading
it to your own local machine (PC/Laptop etc.).
Once you’ve opened this you can go through each section and evaluate all the results. You will likely not want to concern yourself too much with anything after MALT - however you should check these for other artefacts (e.g. weird damage patterns on the human DNA, or weirdly skewed coverage distributions).
For example, I normally look for things like:
General Stats Table:
- Do I see the expected number of raw sequencing reads (summed across each set of FASTQ files per library) that was requested for sequencing?
- Does the percentage of trimmed reads look normal for aDNA, and do lengths after trimming look short as expected of aDNA?
- Does ClusterFactor or ‘Dups’ look high suggesting over-amplified or badly preserved samples (e.g. >2 or >10% respectively - however given this is on the human reads this is just a rule of thumb and may not reflect the quality of the metagenomic profile) ?
- Does the human DNA show increased frequency of C>Ts on the 5’ end of molecules?
FastQC (pre-AdapterRemoval):
- Do I see any very early drop off of sequence quality scores suggesting problematic sequencing run?
- Do I see outlier GC content distributions?
- Do I see high sequence duplication levels?
AdapterRemoval:
- Do I see high numbers of singletons or discarded read pairs?
FastQC (post-AdapterRemoval):
- Do I see improved sequence quality scores along the length of reads?
- Do I see reduced adapter content levels?
MALT:
- Do I have a reasonable level of mappability?
- Somewhere between 10-30% can be pretty normal for aDNA, whereas e.g. <1% requires careful manual assessment
- Do I have a reasonable taxonomic assignment success?
- You hope to have a large number of the mapped reads (from the mappability plot) that also have taxonomic assignment.
Samtools Flagstat (pre/post Filter):
- Do I see outliers, e.g. with unusually high levels of human DNA, (indicative of contamination) that require downstream closer assessment?
DeDup/Picard MarkDuplicates:
- Do I see large numbers of duplicates being removed, possibly indicating over-amplified or badly preserved samples?
DamageProfiler:
- Do I see evidence of damage on human DNA? Note this is just a
rule-of-thumb/corroboration of any signals you might find in the metagenomic
screening and not essential.
- If you have high numbers of human DNA reads but no damage may indicate significant modern contamination.
Detailed documentation and descriptions for all MultiQC modules can be seen in the the ‘Documentation’ folder of the results directory or here in the output documentation
If you’re happy everything looks good in terms of sequencing, we then look at specific directories to find any files you might want to use for downstream processing.
Note that when you get back to writing up your publication, all the versions of the tools can be found under the ‘nf-core/eager Software Versions’ section of the MultiQC report. Note that all tools in the container are listed, so you may have to remove some of them that you didn’t actually use in the set up.
For example, in the example above, we have used: Nextflow, nf-core/eager, FastQC, AdapterRemoval, fastP, BWA, Samtools, endorS.py, Picard Markduplicates, Qualimap, PreSeq, DamageProfiler, MALT, MaltExtract and MultiQC.
Citations to all used tools can be seen here
Tutorial Metagenomics - Files for Downstream Analysis
If you wanted to look at the output of MALT more closely, such as in the GUI
based tool
MEGAN6,
you can find the .rma6
files that is accepted by MEGAN under
metagenomic_classification/malt/
. The log file containing the information
printed to screen while MALT is running can also be found in this directory.
As we ran the HOPS pipeline (primarily the MaltExtract tool), we can look in
MaltExtract/results/
to find all the corresponding output files for the
authentication validation of the metagenomic screening (against the taxa you
specified in your --maltextract_taxon_list
). First you can check the
heatmap_overview_Wevid.pdf
summary PDF from HOPS (again you will need to
either mount the server or download), but to get the actual per-sample/taxon
damage patterns etc., you can look in pdf_candidate_profiles
. In some cases
there maybe valid results that the HOPS ‘postprocessing’ script doesn’t pick up.
In these cases you can go into the default
directory to find all the raw text
files which you can use to visualise and assess the authentication results
yourself.
Finally, if you want to re-run the taxonomic classification with a new database
or tool, to find the raw fastq/
files containing only unmapped reads that went
into MALT, you should go into samtools/filter
. In here you will find files
ending in unmapped.fastq.gz
for each library.
Tutorial Metagenomics - Clean up
Finally, I would recommend cleaning up your work/
directory of any
intermediate files (if your -profile
does not already do so). You can do this
by going to above your results/
and work/
directory, e.g.
and running
Tutorial Metagenomics - Summary
In this this tutorial we have described an example on how to set up a metagenomic screening run of ancient microbiome samples. We have covered how to set up nf-core/eager to extract off-target reads in a form that can be used for MALT, and how to additionally run HOPS to authenticate expected taxa to be found in the human oral microbiome. Finally we have also described what to look for in the MultiQC run summary report and where to find output files that can be used for downstream analysis.
Tutorial - How to set up nf-core/eager for pathogen genomics
Tutorial Pathogen Genomics - Introduction
This tutorial will give a basic example on how to set up nf-core/eager to perform bacterial genome reconstruction from samples in the context of ancient pathogenomic research.
⚠️ this tutorial does not describe how to install and set-up nf-core/eager For this please see other documentation on the nf-co.re website.
We will describe how to set up mapping ancient pathogen samples against the reference of a targeted organism genome, to check sequencing and library quality-control, calculation of depth and breath of coverage, check for damage profiles, feature-annotation statistics (e.g. for gene presence and absence), SNP calling, and producing an SNP alignment for its usage in downstream phylogenetic analysis.
I will use as an example data from Andrades Valtueña et al 2017, who retrieved Late Neolithic/Bronze Age Yersinia pestis genomes. This data is very large shotgun data and is not ideal for testing, so running on your own data is recommended as otherwise running this data will require a lot of computing resources and time. However, note the same procedure can equally be applied on shallow-shotgun and also whole-genome enrichment data, so other than the TSV file you can apply this command explained below.
⚠️ Please be aware that the settings used in this tutorial may not use settings nor produce files you would actually use in ‘real’ analysis. The settings are only specified for demonstration purposes. Please consult the your colleagues, communities and the literature for optimal parameters.
Tutorial Pathogen Genomics - Preparation
Prior setting up the nf-core/eager run, we will need:
- Raw sequencing data in FASTQ format
- Reference genome in FASTA format, with associated pre-made
bwa
,samtools
andpicard SequenceDictionary
indices (however note these can be made for you with nf-core/eager, but this can make a pipeline run take much longer!) - A GFF file of gene sequence annotations (normally supplied with reference genomes downloaded from NCBI Genomes, in this context from here)
- [Optional] Previously made VCF GATK 3.5 files (see below for settings), of previously published Y. pestis genomes.
We should also ensure we have the very latest version of the nf-core/eager pipeline so we have all latest bugfixes etc. In this case we will be using nf-core/eager version 2.2.0. You should always check on the nf-core website whether a newer release has been made (particularly point releases e.g. 2.2.1).
It is important to note that if you are planning on running multiple runs of nf-core/eager for a given project, that the version should be kept the same for all runs to ensure consistency in settings for all of your libraries.
Tutorial Pathogen Genomics - Inputs and Outputs
To start, lets make a directory where all your nf-core/eager related files for this run will go, and change into it.
The first part of constructing any nf-core/eager run is specifying a few generic parameters that will often be common across all runs. This will be which pipeline, version and profile we will use. We will also specify a unique name of the run to help us keep track of all the nf-core/eager runs you may be running.
For the -profile
parameter, I have indicated that I wish to use Singularity as
my software container environment, and I will use the MPI-SHH institutional
config as listed on
nf-core/configs,
and using the profile for the ‘sdag’ cluster. These profiles specify settings
optimised for the specific cluster/institution, such as maximum memory available
or which scheduler queues to submit to. More explanations about configs and
profiles can be seen in the nf-core
website and the profile
tutorial.
Next we need to specify our input data. nf-core/eager can accept input FASTQs files in two main ways, either with direct paths to files (with wildcards), or with a Tab-Separate-Value (TSV) file which contains the paths and extra metadata. In this example, we will use the TSV method, as to simulate a realistic use-case, such as both receiving single-end and paired-end data from Illumina NextSeq and Illumina HiSeqs of double-stranded libraries. Illumina NextSeqs sequence a given library across four different ‘lanes’, so for each library you will receive four FASTQ files. Sometimes samples will be sequenced across multiple HiSeq lanes to maintain complexity to improve imaging by of base calls. The TSV input method is more useful for this context, as it allows ‘merging’ of these lanes after preprocessing prior mapping (whereas direct paths will consider each pair of FASTQ files as independent libraries/samples).
Note we also have a mixture of non-UDG and half-UDG treated libraries.
You can see that we have a single line for each set of FASTQ files representing
each Lane
, but the Sample_Name
and Library_ID
columns identify and group
them together accordingly. Secondly, as we have NextSeq data, we have specified
we have 2
for Colour_Chemistry
vs 4
for HiSeq; something that is important
for downstream processing (see below). See the nf-core/eager
parameter documentation above for more specifications on how to set up a TSV
file (e.g. why despite NextSeqs only having 4 lanes, we can also go up to 8 or
more when having a sample sequenced on two NextSeq runs).
Alongside our input TSV file, we will also specify the paths to our reference FASTA file and the corresponding indices.
We specify the paths to each reference genome and it’s corresponding tool
specific index. Paths should always be encapsulated in quotes to ensure Nextflow
evaluates them, rather than your shell! Also note that as bwa
generates
multiple index files, nf-core/eager takes a directory that must contain these
indices instead.
Note the difference between single and double
-
parameters. The former represent Nextflow flags, while the latter are nf-core/eager specific flags.
Finally, we can also specify the output directory and the Nextflow work/
directory (which contains ‘intermediate’ working files and directories).
Tutorial Pathogen Genomics - Pipeline Configuration
Now that we have specified the input data, we can start moving onto specifying
settings for each different module we will be running. As mentioned above, some
of our samples were generated as NextSeq data, which is generated with a
two-colour imaging technique. What this means is when you have shorter molecules
than the number of cycles of the sequencing chemistry, the sequencer will
repeatedly see ‘G’ calls (no colour) at the last few cycles, and you get long
poly-G ‘tails’ on your reads. We therefore will turn on the poly-G clipping
functionality offered by fastp
, and any
pairs of files indicated in the TSV file as having 2
in the Colour_Chemistry
column will be passed to fastp
(the HiSeq data will not). We will not change
the default minimum length of a poly-G string to be clipped.
We then need to specify the mapping parameters for this run. Typically, to
account for damage of very old aDNA libraries and also sometimes for
evolutionary divergence of the ancient genome to the modern reference, we should
relax the mapping thresholds that specify how many mismatches a read can have
from the reference to be considered ‘mapped’. We will also speed up the seeding
step of the seed-and-extend approach by specifying the length of the seed. We
will do this with --bwaalnn
and --bwaalnl
respectively.
As we are also interested at checking for gene presence/absence (see below), we will ensure no mapping quality filter is applied (to account for gene duplication that may cause a read to map equally to to places) by setting the threshold to 0. In addition, we will discard unmapped reads to reduce our hard-drive footprint.
While some of our input data is paired-end, we will keep with the default of
Picard’s MarkDuplicates’for duplicate removal, as DeDup takes into account
both the start and end of a merged read before flagging it as a duplicate -
something that isn’t valid for a single-end read (where the true end of the
molecule might not have been sequenced). We can then specify which dedupper we
want to use with --dedupper
. While we are using the default (which does not
need to be directly specified), we will put it explicitly in our command for
clarity.
Alongside making a SNP table for downstream phylogenetic analysis (we will get to this in a bit), you may be interested in generating some summary statistics of annotated parts of your reference genome, e.g. to see whether certain virulence factors are present or absent. nf-core/eager offers some basic statistics (percent and and depth coverage) of these via Bedtools. We will therefore turn on this module and specify the GFF file we downloaded alongside our reference fasta. Note that this GFF file has a lot of redundant data, so often a custom BED file with just genes of interest is recommended. Furthermore
Next, we will set up trimming of the mapped reads to alleviate the effects of
DNA damage during genotyping. To do this we will activate trimming with
--run_trim_bam
. The libraries in this example underwent either no or
‘half’-UDG treatment. The latter will generally restrict all remaining DNA
damage to the first 2 base pairs of a fragment. We will therefore use
--bamutils_clip_half_udg_left
and --bamutils_clip_half_udg_right
to trim 2
bp on either side of each fragment. For the non-UDG treated libraries we can
trim a little more to remove most damage with the --bamutils_clip_none_udg_<*>
variants of the flag. Note that there is a tendency in ancient pathogenomics to
trim damage prior mapping, as it allows mapping with stricter parameters to
improve removal of reads deriving from potential evolutionary diverged
contaminants (this can be done nf-core/eager with the Bowtie2 aligner), however
we do BAM trimming instead here as another demonstration of functionality.
Here we will use MultiVCFAnalyzer for the generation of our SNP table. A MultiVCFAnalyzer SNP table allows downstream assessment of the level of multi-allelic positions, something not expected when dealing with a single ploidy organism and thus may reflect cross-mapping from multiple-strains, environmental relatives or other contaminants.
For this we need to run genotyping, but specifically with GATK UnifiedGenotyper 3.5 (as MultiVCFAnalyzer requires this particular format of VCF files). We will therefore turn on Genotyping, and check ploidy is set 2 so ‘heterozygous’ positions can be reported. We will also need to specify that we want to use the trimmed bams from the previous step.
Finally we can set up MultiVCFAnalyzer itself. First we want to make sure we specified that we want to report the frequency of the given called allele at each position so we can assess cross mapping. Then, often with ancient pathogens, such as Y. pestis, we also want to include to the SNP table comparative data from previously published and ancient genomes. For this we specify additional VCF files that have been generated in previous runs with the same settings and reference genome. We can do this as follows.
For the two min_allele_freq
parameters we specify that anything above 90%
frequency is considered ‘homozygous’, and anything above 10% (but below 90%) is
considered an ambiguous call and the frequency will be reported. Note that you
would not normally use this SNP table with these parameters for downstream
phylogenetic analysis, as the table will include ambiguous IUPAC codes, making
it only useful for fine-comb checking of multi-allelic positions. Instead, set
both parameters to the same value (e.g. 0.8) and use that table for downstream
phylogenetic analysis.
With this, we are ready to submit! If running on a remote cluster/server, Make
sure to run this in a screen
session or similar, so that if you get a ssh
signal drop or want to log off, Nextflow will not crash.
Tutorial Pathogen Genomics - Results
Assuming the run completed without any crashes (if problems do occur, check
against #usage that all parameters are as expected, or
check the FAQ), we can now check our results in
results/
.
Tutorial Pathogen Genomics - MultiQC Report
In here there are many different directories containing different output files.
The first directory to check is the MultiQC/
directory. You should
find a multiqc_report.html
file. You will need to view this in a web browser,
so I recommend either mounting your server to your file browser, or downloading
it to your own local machine (PC/Laptop etc.).
Once you’ve opened this you can go through each section and evaluate all the results. For example, I normally look for things like:
General Stats Table:
- Do I see the expected number of raw sequencing reads (summed across each set of FASTQ files per library) that was requested for sequencing?
- Does the percentage of trimmed reads look normal for aDNA, and do lengths after trimming look short as expected of aDNA?
- Does the Endogenous DNA (%) columns look reasonable (high enough to indicate you have received enough coverage for downstream, and/or do you lose an unusually high reads after filtering )
- Does ClusterFactor or ’% Dups’ look high (e.g. >2 or >10% respectively - high values suggesting over-amplified or badly preserved samples i.e. low complexity; note that genome-enrichment libraries may by their nature look higher).
- Do you see an increased frequency of C>Ts on the 5’ end of molecules in the mapped reads?
- Do median read lengths look relatively low (normally <= 100 bp) indicating typically fragmented aDNA?
- Does the % coverage decrease relatively gradually at each depth coverage, and does not drop extremely drastically
- Does the Median coverage and percent >3x (or whatever you set) show sufficient coverage for reliable SNP calls and that a good proportion of the genome is covered indicating you have the right reference genome?
- Do you see a high proportion of % Hets, indicating many multi-allelic sites (and possibly presence of cross-mapping from other species, that may lead to false positive or less confident SNP calls)?
FastQC (pre-AdapterRemoval):
- Do I see any very early drop off of sequence quality scores suggesting problematic sequencing run?
- Do I see outlier GC content distributions?
- Do I see high sequence duplication levels?
AdapterRemoval:
- Do I see high numbers of singletons or discarded read pairs?
FastQC (post-AdapterRemoval):
- Do I see improved sequence quality scores along the length of reads?
- Do I see reduced adapter content levels?
Samtools Flagstat (pre/post Filter):
- Do I see outliers, e.g. with unusually low levels of mapped reads, (indicative of badly preserved samples) that require downstream closer assessment?
DeDup/Picard MarkDuplicates:
- Do I see large numbers of duplicates being removed, possibly indicating over-amplified or badly preserved samples?
PreSeq:
- Do I see a large drop off of a sample’s curve away from the theoretical complexity? If so, this may indicate it’s not worth performing deeper sequencing as you will get few unique reads (vs. duplicates that are not any more informative than the reads you’ve already sequenced)
DamageProfiler:
- Do I see evidence of damage on the microbial DNA (i.e. a % C>T of more than ~5% in the first few nucleotide positions?) ? If not, possibly your mapped reads are deriving from modern contamination.
QualiMap:
- Do you see a peak of coverage (X) at a good level, e.g. >= 3x, indicating sufficient coverage for reliable SNP calls?
MultiVCFAnalyzer:
- Do I have a good number of called SNPs that suggest the samples have genomes with sufficient nucleotide diversity to inform phylogenetic analysis?
- Do you have a large number of discarded SNP calls?
- Are the % Hets very high indicating possible cross-mapping from off-target organisms that may confounding variant calling?
Detailed documentation and descriptions for all MultiQC modules can be seen in the the ‘Documentation’ folder of the results directory or here in the output documentation
If you’re happy everything looks good in terms of sequencing, we then look at specific directories to find any files you might want to use for downstream processing.
Note that when you get back to writing up your publication, all the versions of the tools can be found under the ‘nf-core/eager Software Versions’ section of the MultiQC report. Note that all tools in the container are listed, so you may have to remove some of them that you didn’t actually use in the set up.
For example, in the example above, we have used: Nextflow, nf-core/eager, FastQC, AdapterRemoval, fastP, BWA, Samtools, endorS.py, Picard Markduplicates, Bedtools, Qualimap, PreSeq, DamageProfiler, MultiVCFAnalyzer and MultiQC.
Citations to all used tools can be seen here
Tutorial Pathogen Genomics - Files for Downstream Analysis
You will find the most relevant output files in your results/
directory. Each
directory generally corresponds to a specific step or tool of the pipeline. Most
importantly you should look in deduplication
for your de-duplicated BAM files
(e.g. for viewing in IGV), bedtools for depth (X) and breadth (%) coverages of
annotations of your reference (e.g. genes), multivcfanalyzer
for final SNP
tables etc that can be used for downstream phylogenetic applications.
Tutorial Pathogen Genomics - Clean up
Finally, I would recommend cleaning up your work/
directory of any
intermediate files (if your -profile
does not already do so). You can do this
by going to above your results/
and work/
directory, e.g.
and running
Tutorial Pathogen Genomics - Summary
In this this tutorial we have described an example on how to set up an nf-core/eager run to process microbial aDNA for a relatively standard pathogen genomics study for phylogenetics and basic functional screening. This includes preform some simple quality control checks, mapping, genotyping, and SNP table generation for downstream analysis of the data. Additionally, we described what to look for in the run summary report generated by MultiQC and where to find output files that can be used for downstream analysis.