oschwengers / bakta Goto Github PK

Would it be possible to also output the nucleotide sequence for each CDS, similar to what is already output for the translated protein?

Add VFDB protein sequences to protein sequence blast expert annotation system.

• rank: 75
• min identity: 90
• min query coverage: 90
• min model coverage: 90

Provide a CLI command to download or update the database from within Bakta.

Superseede UniRef100/UniRef90 annotations by the following specialized plasmid protein DBs:

Mobilization proteins:

• MobScan: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2937521 https://castillo.dicom.unican.es/mobscan_about/ https://castillo.dicom.unican.es/mobscan_about/MOBfamDB.gz

Conjugation proteins:

• CONJScan https://github.com/gem-pasteur/Macsyfinder_models/tree/master/models/Conjugation http://dx.doi.org/10.1371/journal.pone.0110726 http://dx.doi.org/10.1038/srep23080

Improve the internal DB download workflow:

1. Improve the progress feedback during the DB download process by using alive-progress.
2. Improve download speed by increasing the download chunk size.

For complete genomes, rotate chromosomes and plasmids to dnaA and RepA/RepB genes, respectively.

Ideas:

• UniRef90: Search for dnaA and RepA/RepB genes in anotations
• Dfast: https://github.com/nigyta/dfast_core/blob/master/dfc/components/DnaAfinder.py
• Unicycler: https://github.com/rrwick/Unicycler/blob/master/unicycler/unicycler.py

If this is conducted before any annotation, Prodigal could be executed with the -c option to disallow the prediction of genes running off edges.

Dear all,
Thank you for developing this tool, it looks very comprehensive and the documentation is really thorough.
I went through your documentation and I see you already have an option --skip-cds where you can skip CDS prediction and annotation. However, I was wondering if it would be possible to have the option to still have the CDS prediction but to skip the posterior CDS functional annotation.
This is simply because I'd like to use an alternative protein function annotation tool but would still need to have the CDS.
In this regard, if you do add this option, which DBs would not be necessary?

Thanks and regards,
Pedro

Add SwissProt annotations:

1. parse uniprot_sprot.xml.gz
2. calc aa seq hash
3. lookup UPS entries
4. a) either update UPS gene/product if no UniRef90 id available in UPS
   b) otherwise update PSC via UPS-UniRef90

Available annotation:

•  <recommendedName>
     <fullName evidence="1">5'-deoxynucleotidase YfbR</fullName>
     <ecNumber evidence="1">3.1.3.89</ecNumber>
   </recommendedName>```
  

•  <gene>
     <name evidence="1" type="primary">yfbR</name>
     <name type="ordered locus">SeAg_B2472</name>
   </gene>```
  

•  <dbReference type="GO" id="GO:0005737">
     <property type="term" value="C:cytoplasm"/>
     <property type="evidence" value="ECO:0000501"/>
     <property type="project" value="UniProtKB-SubCell"/>
   </dbReference>
   <dbReference type="GO" id="GO:0002953">
     <property type="term" value="F:5'-deoxynucleotidase activity"/>
     <property type="evidence" value="ECO:0000501"/>
     <property type="project" value="InterPro"/>
   </dbReference>
   <dbReference type="GO" id="GO:0046872">
     <property type="term" value="F:metal ion binding"/>
     <property type="evidence" value="ECO:0000501"/>
     <property type="project" value="UniProtKB-KW"/>
   </dbReference>
   <dbReference type="GO" id="GO:0000166">
     <property type="term" value="F:nucleotide binding"/>
     <property type="evidence" value="ECO:0000501"/>
     <property type="project" value="UniProtKB-KW"/>
   </dbReference>```
  

Add check if required dependencies are available at runtime.

Also check the dependencies' software versions.

Bakta needs an IS element detection and annotation feature.
A promising candidate (detection, annotation, recently updated, BioConda) is ISEScan:

• https://github.com/xiezhq/ISEScan
• https://doi.org/10.1093/bioinformatics/btx433

Annotate all clusters by KEGG's KoFams:
https://www.genome.jp/ftp/db/kofam/

In assembly GCF_002007885.1 a riboswitch is detected and duplicated in the final annotation.

Position: [126426,126511] on sequence NZ_CP019870.1

Products:

• c-di-GMP-II-GAG riboswitch, score 48.9
• Cyclic di-GMP-II riboswitch, score 74.3

Hi,

I'm trying to test out Bakta on our newly sequenced bacterial genome (1 chromosome, 6 other genomic compartments in one multifasta file), hopefully comparing against existing Prokka+custom database pipeline. However, the process exits with error message once it gets to 'conduct expert systems...' stage under 'predict & annotate CDSs...' I can replicate this issue on both our lab's server and on my personal laptop (both running Ubuntu, 20.04 and 21.04).

Bakta was installed with conda/bioconda in both of the instances using below command, running on Ubuntu.

conda create -n bakta bakta

The annotation process command was:

bakta --db ~/bakta_data/db/ --output test --genus genusname --species speciesname --threads 20 --replicon replicon.tsv our_assembly.fasta

The process exits with below error message

Traceback (most recent call last):
  File "/home/user/miniconda3/envs/bakta/bin/bakta", line 10, in <module>
    sys.exit(main())
  File "/home/user/miniconda3/envs/bakta/lib/python3.9/site-packages/bakta/main.py", line 276, in main
    expert_amr_found = exp_amr.search(genome['features'][bc.FEATURE_CDS], cds_fasta_path)
  File "/home/user/miniconda3/envs/bakta/lib/python3.9/site-packages/bakta/expert/amrfinder.py", line 33, in search
    raise Exception(f'amrfinder error! error code: {proc.returncode}')
Exception: amrfinder error! error code: 1

Please let me know if you need more info/tests on my end for troubleshooting.

Thank you!

Analyze hypotheticals via:

• hmmsearch vs Pfam-A
• compute protein statistics via BioPython
  • molecular weight
  • isoelectric point
• trans-membrane analysis via tmhmm?

Catch thread argument values larger than the number of available threads.

Thanks to @LuisFF bringing this up in #62

The variable in this line should be argcount

Describe the bug
When building the docker image with buildah an error message occurs that states that the "SHELL" command is not OCI compliant:
ERRO[0000] SHELL is not supported for OCI image format, [bash -l -c] will be ignored. Must use docker format

Should we change this to be OCI compliant?

Uuse DeepSig (GPL3) to predict signal peptides and cleavage sites.

• GH: https://github.com/BolognaBiocomp/deepsig
• DOI: https://doi.org/10.1093/bioinformatics/btx818

Add skip options for each feature type:

• --skip-trna
• --skip-tmrna
• --skip-rrna
• --skip-ncrna
• --skip-cds
• --skip-sorf

First of all:, thanks for this great tool.
I was wondering if would be feasible to add the fasta sequence at the end of the gff3 file (similarly to prokka gff output). It would be useful for downstream analysis (e.g. roary pangenome pipelines.). Thanks for your effort.
Paolo

Dear all,
Thank you for this amazing annotation tool for bacterial genomes! I want to known if you can add an option for orthosearch if we have Genbank or Protein FASTA file(s) that we want to annotate genes from as the first priority.
This option has been supplied in softwares dfast-core as "--references" (https://github.com/nigyta/dfast_core) and prokka as "--proteins".

Thanks and regards,
Tonny_Z

A configuration file for all replicons within an assembly / genome should be parsed in order to use this information in output files, e.g. GenBank.

Format idea:

Available short cuts:

• chromosome: c
• plasmid: p
• circular: c
• linear: l

<empty> values (- / '') will be replaced by defaults. If new locus id is empty, a new contig name will be autogenerated.

Defaults:
replicon type: contig
topology: linear

Example:

Usage:

bakta --replicons <file.tsv>

Add a CRISPR detetion with either:

• CRT - http://dx.doi.org/10.1186/1471-2105-8-209
• CRISPRCasFinder - https://doi.org/10.1093/nar/gky425

analyze hypotheticals...
Traceback (most recent call last):
  File "<xyz>/bin/bakta", line 10, in <module>
    sys.exit(main())
  File "<xyz>/bakta/main.py", line 291, in main
    pfam_hits = feat_cds.predict_pfam(hypotheticals)
  File "<xyz>/bakta/features/cds.py", line 264, in predict_pfam
    raise Exception(f'hmmsearch error! error code: {proc.returncode}')
Exception: hmmsearch error! error code: 1

Thanks Matthew Croxen for pointing out.

Add a comprehensive logging of all initialization steps and applied annotations.
This logging file will be publicly provided in order to check all annotations in the bakta db.

For fragmented draf assemblies as well as short plasmids the insufficient amount of genetic information might result in far from optimal training data for subsequent CDS prediction.
For these cases, it might be useful and beneficial to provide pre-calculated Prodigal training files for bacterial taxa (genus or species level) as well as plasmids maybe clustered to Inc groups.

Detect oriC and oriT for (complete) chromosomes and plasmids.

• oriC sequences: DoriC
• oriT sequences: MOB-suite

Introduce an abstract system for protein sequence based expert systems providing:

• reference sequences
• minimal alignment thresholds: identity, query coverage, subject coverage
• gene label
• product description
• DbXrefs: EC, etc ...
• rank: a precedence rank to order & select potentially multiple annotations from several expert systems

Implement NCBI BlastRules as a first protein sequence based expert annotation systems

Add some general summarizing genome stats at the end of the annotation process

As tRNAscan-SE 2.0 only detects tRNAs we need to predict tmRNAs separately via e.g. aragorn

Hi,
I'm in the process of submitting annotated genomes with Bakta to the NCBI, hence while checking for the quality and errors of annotation (via table2asn and the gff3 file https://www.ncbi.nlm.nih.gov/genbank/genomes_gff/#run), I encountered several issues with it.
I understand if it's not something this tool will be compatible with as it can be quite tricky but anyhow, here is a list of some of the issues that I had that may be addressed in the gff file:

• add a gene line

contig_1	Prodigal	CDS	3	179	.	-	0	ID=DOCECA_00005;locus_tag=DOCECA_00005;product=hypothetical protein
contig_1	Bakta	gene	3	179	.	-	0	ID=DOCECA_00005;locus_tag=DOCECA_00005

• remove commas in the "product=" category with the exception of EC numbers

Commas that are intended to be part of a name should be encoded (%2C) according to the GFF3 specifications. However, literal commas should only be included when they are part of enzymatic names. Semi-colons generally should not be included in product names.

• Add an option to remove the SO: in dbxref as they are not yet recognized (https://www.ncbi.nlm.nih.gov/genbank/collab/db_xref/)

Best
Greg

Add detection/prediction of Rho (in)dependent translation terminators

Rho independent

• transtermhp: https://genomebiology.biomedcentral.com/articles/10.1186/gb-2007-8-2-r22; Bioconda:https://anaconda.org/bioconda/transtermhp

hmmsearch --tblout out.tsv --noali --cpu 4 --cut_ga AntiFam_Bacteria.hmm aa.faa

Some annotations don't have an ID in the 9th field
Majority of the annotations do have ID, but a few don't. I can't use the annotation readily for RNAseq counting using HTSeq. Is this expected?

Here are all of them in one of my annotations without ID (last one is a sample annotation with an ID):

Add integration tests. Therefore, a tiny mock db as well as some small test genomes are needed.

Detect assembly gaps (>5 Ns) and add annotations (assembly_gap) accordingly.

https://github.com/nigyta/dfast_core/blob/master/dfc/tools/gap.py

First hints for ideas:

• https://github.com/nigyta/dfast_core
• https://github.com/filip-husnik/pseudofinder#how-does-pseudofinder-detect-pseudogene-candidates

Pre-annotation of UniRef90 clusters could be improved using a dedicated phage database like:

• VOGDB: http://vogdb.org/download
• PHROGs: https://phrogs.lmge.uca.fr/READMORE.php

Does this tool also work for archeal genomes or metagenome-assembled genomes? Thanks!

A refinement of the rRNA detection is necessary as currently too many truncated and false positive sequences are predicted.

Inspired by Barrnap and certain issues (tseemann/prokka#423, tseemann/barrnap#39), the cmsearch based approach (cmsearch --noali --cut_tc --notrunc) therefore gets replaced by a cmscan approach in glocal mode (cmscan --noali --cut_tc -g --nohmmonly --rfam)

Many features overlap with each other. A hierarchical bacteria specific overlap detection filter needs to be implemented.
Current hierarchical overlap preference (descending):

Add a --debug option in order to keep tmp files and activate additional debugging logs.

Implement common a ground for expert annotation systems as for example AMRFinderPlus thus allowing the incorporation of certain expert knowledge.

Additional "expert systems" might be:

• NCBI BlastRules
• HAMAP
• UniProt annotation rules

With Ubuntu 20.04, the version in the repo of tRNAscan-SE is 2.0.5-1

I got this error in the call of bakta

ERROR: Wrong tRNAscan-SE version installed. Please, either install tRNAscan-SE version v2.0.6 or use ['--skip-trna']!

With Ubuntu 21.04, the version in the repo of tRNAscan-SE is 2.0.7+ds-1, i dont encounter this error in bakta.

1. Execute AMRFinderPlus with a custom $TMPDIR location pointing to Bakta's internal tmp dir.
2. Write AMRFinderPlus's db to a subdir within the Bakta database directory via amrfinder_update --database <db>/amrfinderplus/ --force_update and amrfinder --protein <input> --database <db>/amrfinderplus/latest as suggested in https://github.com/oschwengers/bakta/discussions/64

For common genomes, use Prodigal training files pre-trained on high-quality reference genomes.
These could be selected by a quick Mash lookup via a RefSeq db of complete reference and representative genomes.