Find full-test data

[jfy133]

Description of feature

These should be 3-5 'real life' samples that you would profile against.

Ideally these would be shortread/long read pairs.

[jfy133]

https://lomanlab.github.io/mockcommunity/

And @sofstam will help look for 'real' illumina/nanopore stuff :)

[sofstam]

I have asked at our site if we are allowed to use some of our data as test-data. Meanwhile, I found those for Nanopore:

https://www.ebi.ac.uk/ena/browser/view/PRJNA312719

[sofstam]

@jfy133 I got a response from our site that we cannot use any of our data as test data right now, we need an ethical approval that we will be working on during fall.

[jfy133]

OK lets just look for already published stuff 👍

Maybe: https://www.nature.com/articles/s41597-019-0287-z ?

[sofstam]

I will have a look at this!

Sounds good with the dataset from this article. Since the dataset is focused on bacteria, it might be good to have test data for viruses as well? https://www.ebi.ac.uk/ena/browser/view/PRJNA670157?show=reads

[Midnighter]

Beyond the mock communities, I'm personally interested in using the CAMI data.

[jfy133]

And also need to decide databases, and where to store them (presumably aws...?)

[Midnighter]

I think Zenodo could be a good place if we want to publish the benchmark. Then every database has a DOI.

[jfy133]

I fear the file sizes for some will be too large for Zenodo (50GB limit) but we can see

[sofstam]

Minimum criteria for full-test data:

• Fastq files, 5 Illumina, 2 Nanopore
• Sequencing depth > 10M
• Shotgun experiment
• Multiple run accessions for one sample
• Host removal (contaminant preferably)

[jfy133]

To 'borrow' from the MAG full-test data we can pick 2-3 illumina and 2-3 ONT samples/runs from here: https://www.ebi.ac.uk/ena/browser/view/PRJEB29152

Have both Illumina and Nanopoe, and sequencing depth is >10m, and is shotgun

[sofstam]

https://www.nature.com/articles/s41597-019-0287-z

I post the article here so we do not forget.

[jfy133]

Meslier2022 is what we are going for: https://www.nature.com/articles/s41597-022-01762-z

I did a test run:

nextflow run nf-core/taxprofiler
		 -profile mpcdf,raven
		 -r combine-kreports-fix
		 --input fulltest_samplesheet.csv
		 --databases fulltest_dbsheet.csv
		 --outdir ./results
		 --save_preprocessed_reads
		 --perform_shortread_qc
		 --shortread_qc_mergepairs
		 --perform_shortread_complexityfilter
		 --save_complexityfiltered_reads
		 --perform_longread_qc
		 --perform_shortread_hostremoval
		 --perform_longread_hostremoval
		 --hostremoval_reference 'ftp://ftp.ncbi.nlm.nih.gov/genomes/all/GCA/000/819/615/GCA_000819615.1_ViralProj14015/GCA_000819615.1_ViralProj14015_genomic.fna.gz'
		 --save_hostremoval_index
		 --save_hostremoval_mapped
		 --save_hostremoval_unmapped
		 --perform_runmerging
		 --save_runmerged_reads
		 --run_centrifuge
		 --centrifuge_save_reads
		 --run_diamond
		 --run_kaiju
		 --run_kraken2
		 --kraken2_save_reads
		 --kraken2_save_readclassification
		 --kraken2_save_minimizers
		 --run_krakenuniq
		 --krakenuniq_save_reads
		 --krakenuniq_save_readclassifications
		 --run_bracken
		 --run_malt
		 --malt_save_reads
		 --malt_generate_megansummary
		 --run_metaphlan3
		 --run_motus
		 --run_profile_standardisation
		 --run_krona
		 -ansi-log false
		 -with-tower
		 -resume
		 --run_metaphlan3
		 --run_motus

With a the following samplesheet:

sample,run_accession,instrument_platform,fastq_1,fastq_2,fasta
MOCK_001_Minion_R9,1,OXFORD_NANOPORE,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/000/ERR9765780/ERR9765780.fastq.gz,,
MOCK_002_Minion_R9,1,OXFORD_NANOPORE,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/001/ERR9765781/ERR9765781.fastq.gz,,
MOCK_003_Minion_R9,1,OXFORD_NANOPORE,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/002/ERR9765782/ERR9765782.fastq.gz,,
MOCK_001_Illumina_Hiseq_3000,1,ILLUMINA,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/006/ERR9765746/ERR9765746_1.fastq.gz,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/006/ERR9765746/ERR9765746_2.fastq.gz,
MOCK_002_Illumina_Hiseq_3000,1,ILLUMINA,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/007/ERR9765747/ERR9765747_1.fastq.gz,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/007/ERR9765747/ERR9765747_2.fastq.gz,
MOCK_003_Illumina_Hiseq_3000,1,ILLUMINA,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/008/ERR9765748/ERR9765748_1.fastq.gz,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/008/ERR9765748/ERR9765748_2.fastq.gz,
MOCK_003_Illumina_Hiseq_3000,2,ILLUMINA,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/009/ERR9765749/ERR9765749_1.fastq.gz,ftp://ftp.sra.ebi.ac.uk/vol1/fastq/ERR976/009/ERR9765749/ERR9765749_2.fastq.gz,

And database sheet

tool,db_name,db_params,db_path
bracken,bracken-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/bracken.tar.gz
centrifuge,centrifuge-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/centrifuge.tar.gz
diamond,diamond-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022/diamond/diamond.dmnd
kaiju,kaiju-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/kaiju.tar.gz
kraken2,kraken2-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/kraken2.tar.gz
krakenuniq,krakenuniq-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/krakenuniq.tar.gz
malt,malt-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/malt.tar.gz
metaphlan3,metaphlan3-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/metaphlan3.tar.gz
motus,motus-db,,/raven/ptmp/jfellowsy/databases/taxprofiler_full_test/meslier2022_final_dbs/tars/motus.tar.gz

The databases TARs basd on these instructions

And it mostly ran! Observations in next commetn

[jfy133]

• The multiQC report looks weird in a few places though
  • The nanopore reads has the the ERR* name not the sample nam1e in the general stats MultiQC Tweaks after full-test experiments #223
  • We still have the bracken entries in the general stats that we need to turn off MultiQC Tweaks after full-test experiments #223
  • Porechop plots all say 'not able to plot data' Tweak MultiQC config [skip ci] #227
    [2023-01-20 18:53:22,590] multiqc.plots.bargraph [WARNING] Tried to make bar plot, but had no data: porechop-starttrim-barplot [2023-01-20 18:53:22,590] multiqc.plots.bargraph [WARNING] Tried to make bar plot, but had no data: porechop-endtrim-barplot [2023-01-20 18:53:22,590] multiqc.plots.bargraph [WARNING] Tried to make bar plot, but had no data: porechop-middlesplit-barplot
• I agree with @Midnighter that we need a separate MQC report for Nanopore data, but I don't want to mess with pipeline code atm and focus on bug fixing (separate issue: Split up MultiQC run reports between long and short reads #218 )
• We only picked up the Bracken results, and no Kraken2 etc. for some reason (I guess the file names we used wasn't correct ) https://github.com/nf-core/taxprofiler/pull/223/files

Output files:

• Centrifuge ran despite no multiQC report, however .results.txt files in that directory are empty
• Kaiju is a complete fail, no output!
• MetaPhlan3 runs fine for short-reads, but all long reads fail with 100% unknown, maybe we should filter this out (I suspect long-reads are unsuitable because long-reads trying to be mapped against short genes witha s hort read aligner ) Filter out nanopore data prior Metaphlan3 #221
• mOTUs did not execute because when supplying in TAR not accounted for here Allow all tool's databases to be provided as .tar.gz files #219
• Separate short-read and long-read parameters for the databases. [v1.1!] Split up longread- and short-read parameters in databases #226

Overall we get 35% reads classified with Brakcken so I think this a good sign this is reasonable dataset

[jfy133]

• Add to docs we recommend running SR/LR separtely (although running together is supported) Add comment about recommendation to split SR/LR #220

[sofstam]

Shall we close this?

[jfy133]

Not yet, lets get the samplesheets and databases seets upload to test-datasets and then we can close it :)