• Describe the bug
  With the current model and settings the average read quality is about 8. This is (1) out-of-date with the current real avg read quality ONT sequencers produce and (2) it is so low that some pipeline just filter out those reads entirely.

• To Reproduce

1. Run cievad hap --read_type ont
2. Run poreCov
3. Check Nanoplot quality report in results/1.Read_quality/

• Expected behavior
  Something like avg 15-20 sould be default for simulation.

• Screenshots
  

• Possible solution
  Add new model or check settings of nanosim to improve avg read quality.

Can you please provide a small generic example of the of a sample sheet in the Wiki?
I think it will help the lesser experienced user.

Situation:
With the new feature #42 the sample_sheet parameter got introduced as an alternative to the callset_dir. These input formats should be mutually exclusive.

Improvement:
Test different input combinations of callset_dir and sample_sheet with GH Actions to ensure intended behavior. This is a good opportunity to also test invalid inputs (empty sample_sheet, missing files, etc).

The Snakemake rule to run amplisim is currently using a compiled binary. This must be changed to a conda installation before release.

Ideally, this will be solved along with Issue #19 .

Minimal changes:

• change this line to simply ´amplisim´
• add a conda env to this rule

Document how to install and launch new NF design

Please replace the static binary (<projectDir>/bin/pbsim) with the conda package.
Find at

conda search -c bioconda pbsim3 

Check for equal results under same model + seed.

• Is your feature request related to a problem? Please describe.
  At the moment the executor.cpus and task.cpus are hardcoded. Reserving a bigger maschine (eg on a cluster) doesn't improve the runtime.

• Describe the solution you'd like
  The a max_cores parameter in the config, that is passed to the executors. Also, change relevant modules to scale accordingly at the task.cpus.

Problem:
Neither hap.nf nor eval.nf have a help page. The user has to dive into the nextflow.config file in order to find parameters that one can tune.

Solution:
Provide help pages in a UNIX CLI style, like nextflow run hap.nf --help

• Update README with explanation that read depth and qual can be tuned. Hint to FAQ
• Example case for documentation how to run with non-default parameter
• Update wiki
• Update minor release

The project could be tested with an end-to-end test for good practice. Particularly the CLI.
Better wait until a decision is made on Issue #17 though.

• Is your feature request related to a problem? Please describe.
  Sometimes processes do not success due to hardware failures. Then, a fraction of the synthetic samples is not generated.

• Describe the solution you'd like
  There are errorStrategy and maxRetries directives for Nextflow processes s.t. this type of failure can be caught. This should be added to relevant (all?) processes.

I think the tool suite would benefit from having one joint conda environment over multiple individual environments, one for each module or snakemake job. It would reduce the runtime for first-ever module activation which is particularly useful for CI purposes. The question is if all requirements and dependencies can be satisfied in a joint environment. Attempt to realize this should be trivial to test once a CI workflow is in place, i.e. issue #18 resolved.

Redesign the CI for the nextflow branch, i.e. new conda env and MWEs

• Is your feature request related to a problem? Please describe.
  No problem, just a feature for convenience.

• Describe the solution you'd like
  Nextflow provides a git checkout-free solution for nextflow pipelines via nextflow run <organization>/<repository>. CIEVaD provides multiple pipelines, so maybe something is possible in the style of nextflow run rki-mf1/cievad-hap, and same with -eval, respectively.

The current architecture of the program was initially designed for a much larger project with a lot more code re-use. Therefore, the folder structure and the split between creating configs and running a workflow is somewhat over-engineered. For long term maintenance I highly advise to reduce and restructure the code base.

Currently, the callsets have to reside or be linked inside the samples' data folders, like data/sample_hap001/callset.vcf.gz with particularly strict naming convention of the filename. That requirement should be less constrained, i.e. a minimal improvement should be

• flexible filename
• uncompressed input possible

• Is your feature request related to a problem? Please describe.
  Reviewer request

• Describe the solution you'd like
  Report N stretches (number, size) in the provided genome.

• Is your feature request related to a problem? Please describe.
  Illumina read error distribution cannot be trivially modified yet.

• Describe the solution you'd like
  Include the corresponding mason simulator parameters into the CLI and nextflow.config .

Align FASTQs and check if the read simulator generates heterozygous mutations

In the conda environment "snakemake7" in dev branch the tools mason and bcftools are missing. Once I installed them manually in the conda env the script was able to run successfully.

Issue:
Currently, the documentation says the user only requires a Nextflow v20.04+. This statement was based on the Nextflow documentation and the Nextflow features used. However, in 2021 somethings changed on the execution level of Nextflow e.g.

• the -dsl2 parameter usage for DSL2 language features
• singularity env variables (not yet relevant but will be in case we support containerized execution)

Solution:
I think something like v22.04+ is appropriate. This can be precisely verified with a version matrix in GH Actions.

Test invalid inputs to the modules, e.g.

• missing ref (HAP)
• invalid read type (HAP)
• #64
• missing callsets (EVAL)
• missing truthset (EVAL)
• malformatted sample sheet (EVAL)
• wrong relative path to callset dir (EVAL)
• deterministic results

Describe the bug

If neither --callsets_dir not --sample_sheet is specified the eval.nf it is supposed to run into the else-case and abort with exit 1. However, I accidentally tested this and nothing happened. The wf ran successfully without anything being done.

To Reproduce

Steps to reproduce the behavior:

$nextflow run eval.nf -profile local,conda --callset_dir data
N E X T F L O W  ~  version 23.10.1
Launching `eval.nf` [prickly_kilby] DSL2 - revision: e4728a89a7
$

(mind that --callset_dir was the mistake here. It must be --callsets_dir)

Expected behavior

As none of the valid input formats are specified, this should end up in an exit 1 with an error message.

Setup (please complete the following information):

• cievad v0.4.0
• nextflow 23.10.1

• Is your feature request related to a problem? Please describe.
  Currently, the output of eval.nf is precision/recall statistics per sample (<sample>.sompy.stats.csv). I'd be helpful to have a summary statistic across all simulated samples.

• Describe the solution you'd like
  Ideally, there would be a file containing an average and/or median for each feature in the individual samples' statistics.

PROBLEM:

The final stats report by som.py misses a first column in the header. It's misleading when converting and looking at the CSV table, e.g. with

sed 's/,/\t/g' sample01.sompy.stats.csv | column -t | less -S 

because all the lines below the header are shifted with respect to the header.

FIX:

The first column of the header needs content, e.g. "idx" in the example below:

idx,type,total.truth,total.query,tp,fp,fn,...
0,indels,140,113,113,0,27,0,0,...
1,SNVs,320,305,305,0,15,0,0,...
5,records,460,422,418,4,42,0,0,...

IMPLEMENTATION:

Add a process that modifies the header after the som.py process.

Make the --seed an optional command line parameter in the hap-simu and ngs-simu features.

• Is your feature request related to a problem? Please describe.
  The BAM file from the simulation wf is vital for visually inspecting the simulation, e.g. with IGV. Currently, the BAM file returned from the NGS data generation is unsorted and comes without the BAI index.

• Describe the solution you'd like
  After the mason_simulator process, sort and index the BAM file by default.