AMPcombi : A parsing, filtering and annotation workflow for tools predicting AntiMicrobial Peptide genes

AMPcombi and its submodules provide a command-line interface to parse the results of antimicrobial peptide (AMP) prediction tools into a single table, aligns the AMP hits against a reference AMP database for functional classifications, filters the AMP hits according to their physicochemical properties, clusters the filtered hits and predicts signaling peptides if present.

---

Introduction

(A) For parsing and filtering; AMPcombi is developed to parse and filter the output of only the following AMP prediction tools:

Tool	Version
Ampir	1.1.0
AMPlify	1.0.3
Macrel	1.1.0
HMMsearch	3.3.2
EnsembleAMPpred	-
NeuBI	-
AMPgram	-
AMPtransformer	-

(B) For classification; AMPcombi is developed to provide the functional annotation of the detected AMPs by alignment to an AMP reference databases using diamond blastp v.2.0.15:

Tool	Version
DRAMP	3.0
Diamond	2.0.15

⚠️ If no database is provided by the user, AMPcombi will automatically download the DRAMP db and use the files for classification.

(C) For structural and physical annotations; corresponding to the molecular weight, isoelectric point, hydrophobicity, pH and the fraction of helix turns and beta sheets were calculated using:

Tool	Version
BioPython - ProteinAnalysis	1.80

Any transporter gene if present in the vicinity of the AMP on the contig is also reported.

(D) For filtering AMP hits: AMPcombi is developed to filter hits based on their AMP probabilities given by each AMP tool - E-values in case of HMMsearch - and the presence of stop-codons downstream and upstream of the AMP hit on the contig.

⚠️ If contig or sample details/information are available this can be added here, for example, metadata, pydamage, contig taxonomic classification, etc.

(E) For clustering AMP hits: AMPcombi is developed to cluster the AMPs predicted and remove singletons and clusters with a minium number of members or clusters with only a specific string in their name. This is done using:

Tool	Version
MMseqs2	15.6f452

(F) For signal peptide detection: Signal peptides can be predicted from clustered AMPs given the user installs SignalP separately. For licensing issues, SignalP can only be downloaded and used by academic users; other users are requested to contact DTU Health Technology Software Package before using it. Please refer to SignalP documentation. For obtaining SignalP please follow:

Tool	Version
SignalP	6.0

(F) For visualization: a shiny app is accessible through ./pyshiny. This is a user interface that renders the AMPcombi summary table and genrates a number of figures for data analysis.

Tool	Version
Shiny	0.7.1

---

Installation

To install AMPcombi:

• Using conda:

conda create -n ampcombi python==3.11 diamond==2.0.15 mmseqs==15.6f452 ampcombi

or

conda env create -f ./ampcombi/environment.yml

• Using singularity and docker:

singularity pull ampcombi:0.2.2--pyhdfd78af_0

• From git repository:

git clone https://github.com/Darcy220606/AMPcombi.git

---

Usage and Output:

For full usage documentaion of ampcombi and it's submodules please refer to the help documentation:

ampcombi --help

Submodules

---

ampcombi parse_tables

This subcommand parses and filters the output files generated by different AMP prediction tools, aligns the amino-acid sequences to the reference database and estimates the physicochemical annotations .

To get a full list of options available and their defaults please refer to the help documentation of the submodule

ampcombi parse_tables --help

Example usage (1):

ampcombi parse_tables \
--amp_results path/to/my/result_folder/ \
--faa path/to/sample_faa_files/ \
--gbk path/to/sample_gbk_or_gbff_files/ \
--sample_list sample_1 sample_2 \
--contig_metadata path/to/contig_metadata.tsv
--<tool_1>_file '.tsv' 
--<tool_2>_file '.txt' 
--log true
--threads 10

In this case we use the --amp_results option, to supply AMP tool prediction results from many samples in a folder format. The folder must follow this structure:

amp_results/
├── tool_1/
|   ├── sample_1/
|   |   └── sample_1.tsv
|   └── sample_2/
|   |   └── sample_2.tsv
├── tool_2/
|   ├── sample_1/
|   |   └── sample_1.txt
|   └── sample_2/
|   |   └── sample_2.txt
├── tool_3/
    ├── sample_1/
    |   └── sample_1.fasta
    └── sample_2/
        └── sample_2.fasta

• --_file the should be changed to ampir, macrel, amplify, neubi, hmmsearch, ensemblamppred, ampgram, amptransformer. The argument value should be a suffix of the files generated by that tool. We have assigned defaults to each tool, however the user can change those defaults according to the user input file extensions.
• --contig_metadata a *.tsv file that must contain the sample name in the first column and the contig ID/name in the second column. The column headers are not important.
• --faa a folder containing annotated files of the AMP hits with a suffix *.faa. This can be generated by any annotation tool i.e., PROKKA, PYRODIGAL, etc. NOTE The files have to include the sample name, for example, <samplename>.faa.
• --gbk a folder containing annotated files of the AMP hits with a suffix *.gbk or *.gbff. This can be generated by any annotation tool i.e., PROKKA, PYRODIGAL, etc. NOTE The files have to include the sample name, for example, <samplename>.gbk/gbff.
• --log captures all standard output and standard errors in a log file
• --threads the number of cores to parallelize the job

Example usage (2):

ampcombi parse_tables \
--path_list path_to_sample_1_tool_1.csv path_to_sample_1_tool_2.txt \
--sample_list sample_1 \
--faa path/to/sample_faa_files/sample_1.faa \
--gbk path/to/sample_gbk_or_gbff_files/sample_1.<gbk><gbff> \
--<tool_1>_file '.tsv' 
--<tool_2>_file '.txt' 

In this case we use the --path_list option, to supply AMP tool prediction results from a single sample in a list format.

Optional arguments:

command	definition	default	example
--amp_cutoff	Probability cutoff to filter AMPs by probability (not applicable for hmmsearch)	0.0	0.5
--hmm_evalue	Probability cutoff to filter AMPs by E-value (only applicable for HMMsearch)	None	0.05
--db_evalue	Probability cutoff to filter database classifications by E-value - any hit with an E-value below this will have it's database classification removed	None	0.05
--aminoacid_length	Probability cutoff to filter AMP hits by the length of the amino acid sequence	100	60
--window_size_stop_codon	The length of the window size required to look for stop codons downstream and upstream of the CDS hits	60	40
--window_size_transporter	The length of the window size required to look for a 'transporter' e.g. ABC transporter downstream and upstream of the CDS hits	11	20
--remove_stop_codons	Removes any AMP hits that don't have a stop codon found in the window downstream or upstream of the CDS assigned by '--window_size_stop_codon'. Must be turned on if hits are to be removed	False	True
--sample_metadata	Path to a tsv-file containing sample metadata, e.g. 'path/to/sample_metadata.tsv'. The metadata table can have more information for sample identification that will be added to the output summary. The table needs to contain the sample names in the first column.	None	./sample_metadata.tsv/
--contig_metadata	Path to a tsv-file containing contig metadata, e.g. 'path/to/contig_metadata.tsv'. The metadata table can have more information for contig classification that will be added to the output summary. The table needs to contain the sample names in the first column and the contig_ID in the second column. The metadata table can be the output from MMseqs2, pydamage and MetaWrap.	None	./contig_metadata.tsv/
--amp_database	Path to the folder containing the reference database files: (1) a fasta file with <.fasta> file extension and (2) the corresponding table with functional and taxonomic classifications in <.tsv> file extension	DRAMP 'general amps'	./custom_amp_ref_database/

⚠️ With regards to the reference database supplied to --amp_database:

• The fasta file corresponding to the AMP database should not contain any characters other than ['A','C','D','E','F','G','H','I','K','L','M','N','P','Q','R','S','T','V','W','Y'].
• The reference database table should be tab delimited.

Output:

The output will be written into your working directory, containing the following files and folders:

<pwd>/
├── amp_ref_database/
|   ├── amp_ref.dmnd
|   ├── general_amps_<DATE>_clean.fasta
|   └── general_amps_<DATE>.tsv
├── sample_1/
|   ├── contig_gbks/
|   ├── sample_1_amp.faa
|   ├── sample_1_ampcombi.tsv
|   ├── sample_1_diamond_matches.txt
|   └── sample_1_ampcombi.log
├── sample_2/
|   ├── contig_gbks/
|   ├── sample_2_amp.faa
|   ├── sample_2_ampcombi.tsv
|   ├── sample_2_diamond_matches.txt
|   └── sample_2_ampcombi.log
└── Ampcombi_parse_tables.log

---

ampcombi complete

This subcommand concatenates the ampcombi summaries generated by running ampcombi parse_tables, in a single complete summary.

To get a full list of options available and their defaults please refer to the help documentation of the submodule

ampcombi complete --help

Example usage (1):

ampcombi complete \
--summaries_directory path/to/ampcombi_parse_tables_results_folder/ 

In this case we use the --summaries_directory option, to supply the samples' result folder from --ampcombi parse_tables which should contain the folder structure from ampcombi parse_tables in a parent folder for example named ./ampcombi/....

Example usage (2):

ampcombi complete \
--summaries_files path/to/ampcombi_parse_tables/sample_1_ampcombi.tsv path/to/ampcombi_parse_tables/sample_2_ampcombi.tsv/ 

In this case we use the --summaries_files option, to supply the ampcombi_parse_tables AMPcombi summary files in a list format.

Output:

The output will be written into your working directory, containing the following files and folders:

<pwd>/
└── Ampcombi_summary.tsv
└── Ampcombi_complete.log

---

ampcombi cluster

This subcommand clusters the AMPcombi complete summary generated by running ampcombi complete. As this uses mmseqs cluster in the background some parameters that were deemed important for the purpose of AMPcombi were incorporated as optional arguments.

To get a full list of options available and their defaults please refer to the help documentation of the submodule

ampcombi cluster --help

Example usage:

ampcombi cluster \
--ampcombi_summary path/to/Ampcombi_summary.tsv 

The input option --ampcombi_summary requires the 'Ampcombi_summary.tsv' which is generated by the ampcombi complete submodule.

Optional arguments:

command	definition	default	example
--cluster_cov_mode	This assigns the cov. mode to the mmseqs2 cluster module- More information can be obtained in mmseqs2 docs here.	0	2
--cluster_mode	This assigns the cluster mode to the mmseqs2 cluster module- More information can be obtained in mmseqs2 docs here.	1	2
--cluster_coverage	This assigns the coverage to the mmseqs2 cluster module- More information can be obtained in mmseqs2 docshere.	0.8	0.9
--cluster_seq_id	This assigns the seqsID to the mmseqs2 cluster module- More information can be obtained in mmseqs2 docs here.	0.4	0.7
--cluster_sensitivity	This assigns sensitivity of alignment to the mmseqs2 cluster module- More information can be obtained in mmseqs2 docs here	4.0	7.0
--cluster_remove_singletons	This removes any hits that did not form a cluster.	True	False
--cluster_retain_label	This removes any cluster that only has a certain label in the sample name. For example if you have sample labels with 'S1_metaspades' and 'S1_megahit', you can retain clusters that have samples with suffix '_megahit' by running '--retain_clusters_label megahit'.	''	'megahit'
--cluster_min_member	This removes any cluster that has a hit number lower than assigned here.	3	1

Output:

The output will be written into your working directory, containing the following files and folders:

<pwd>/
└── Ampcombi_summary_cluster.tsv
├── Ampcombi_summary_cluster_representative_seq.tsv
└── Ampcombi_cluster.log

• Ampcombi_summary_cluster.tsv includes the contents of the complete summary plus a column with cluster IDs.
• Ampcombi_summary_cluster_representative_seq.tsv contains a table with all the representative hits from each cluster.

---

ampcombi signal_peptide

This subcommand predicts whether a signal peptide was found on the filtered and clustered AMP hits. This only works if the user installs SignalP separately. For licensing issues, SignalP can only be downloaded and used by academic users; other users are requested to contact DTU Health Technology Software Package before using it. Please refer to SignalP documentation.

To get a full list of options available and their defaults please refer to the help documentation of the submodule

ampcombi signal_peptide --help

Example usage:

ampcombi signal_peptide \
--signalp_model path/to/signalp_model/ \
--ampcombi_cluster path/to/Ampcombi_summary_cluster.tsv \
--log true

The input option --ampcombi_cluster requires the 'Ampcombi_summary_cluster.tsv' which is generated by the ampcombi cluster submodule.

Output:

The output will be written into your working directory, containing the following files and folders:

<pwd>/
└── Ampcombi_summary_cluster_SP.tsv
├── Ampcombi_summary_cluster_SP_onlyclusterswithSP.tsv
├── Ampcombi_summary_cluster_SP_onlyclusterswithSP.tsv
├── signalp
|   ├── output_*.png/
|   ├── prediction_results_index.tsv
|   ├── prediction_results.tsv
|   ├── representative_seq.txt
└── Ampcombi_signalpeptide.log

• Ampcombi_summary_cluster_SP.tsv includes the contents of the cluster summary plus a column with yes/no indicating the presence of a signal peptide sequence.
• Ampcombi_summary_cluster_SP_onlyclusterswithSP.tsv includes the contents of the cluster summary plus a column with yes/no indicating the presence of a signal peptide sequence. But in this case clusters are retained only if they contain a hit or more with a signaling peptide.
• signalp contains the results from the tool signalp in *.png format showing the location of the predicted signaling peptide and the prediction_results.tsv which contains a table with the location of the signaling peptide and the identity. The prediction_results_index.tsv contains a table that gives an index number to every hit found in ./AMPcombi_summary_ao_human_nonhuman_clusters_SP_onlyclusterswithSP.tsv. This can be used to rename the localcolabfold results downstream in the workflow.

---

Example runs using test files:

To test the function and output for AMPcombi, we provide test files that can be downloaded as described below, distributed in the three packages named test_faa, test_gbk and test_files.

Step1: Download the test directories and unzip :

wget https://github.com/Darcy220606/AMPcombi/tree/main/test_faa.tar.gz 
wget https://github.com/Darcy220606/AMPcombi/tree/main/test_gbk.tar.gz 
wget https://github.com/Darcy220606/AMPcombi/tree/main/test_files.tar.gz

tar -xzvf test_faa.tar.gz
tar -xzvf test_gbk.tar.gz
tar -xzvf test_files.tar.gz

Step2: Parse tables from all AMP tools. This can be produced by running the AMP workflow from FUNCSCAN - a pipeline for predicting functional genes in metagenomes.

ampcombi parse_tables --amp_results ./test_files/ --faa ./test_faa/ --gbk ./test_gbk/ --sample_list sample_1 sample_2 --ampir_file '.tsv' --amplify_file '.tsv' --macrel_file '.tsv' --neubi_file '.fasta' --hmmsearch_file '.txt' --ampgram_file '.tsv' --amptransformer_file '.txt' --threads 28 --log true

Step3: Concatenate all AMPcombi summary files from all samples:

ampcombi complete --summaries_directory ./test_ampcombi/ --log true

Step4: Cluster the hits and remove singletons:

ampcombi cluster --ampcombi_summary Ampcombi_summary.tsv --log true

---

Visualization:

To visualize the result tables from AMPcombi, a pyshiny app can be rendered by running:

cd ./pyshiny
pip install -r requirements.txt
shiny run --port 36317 --reload app.py

⚠️ the port can be changed accordingly

The user can upload the Ampcombi_summary_cluster_SP.tsv to generate tables and figures ready for publication. 3D structures in PDB format can also be uploaded to generate an overlay structure.

---

References for tools, packages and databases used in AMPcombi:

• Steinegger M and Soeding J. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nature Biotechnology, doi: 10.1038/nbt.3988 (2017).

• Steinegger M and Soeding J. Clustering huge protein sequence sets in linear time. Nature Communications, doi: 10.1038/s41467-018-04964-5 (2018).

• Mirdita M, Steinegger M and Soeding J. MMseqs2 desktop and local web server app for fast, interactive sequence searches. Bioinformatics, doi: 10.1093/bioinformatics/bty1057 (2019).

• Mirdita M, Steinegger M, Breitwieser F, Soding J, Levy Karin E: Fast and sensitive taxonomic assignment to metagenomic contigs. Bioinformatics, doi: 10.1093/bioinformatics/btab184 (2021).

• Teufel, F., Almagro Armenteros, J.J., Johansen, A.R. et al. SignalP 6.0 predicts all five types of signal peptides using protein language models. Nat Biotechnol 40, 1023–1025 doi: 10.1038/s41587-021-01156-3 (2022).

• Buchfink B, Reuter K, Drost HG, Sensitive protein alignments at tree-of-life scale using DIAMOND. Nature Methods 18, 366–368 doi:10.1038/s41592-021-01101-x (2021).

• Shi G., Kang X., Dong F., Liu Y., Zhu N., Hu Y., Xu H., Lao X., Zheng H., DRAMP 3.0: an enhanced comprehensive data repository of antimicrobial peptides, Nucleic Acids Research, 50,D1, doi: 10.1093/nar/gkab651 (2022).

• The Shiny development team, Shiny for Python, https://shiny.posit.co/py/, license:https://github.com/posit-dev/py-shiny/blob/main/LICENSE v.0.8.0

• Inc., P. T. Collaborative data science. Montreal, QC: Plotly Technologies Inc. Retrieved from https://plot.ly (2015)

• Upsetplot, https://github.com/jnothman/UpSetPlot license: https://github.com/jnothman/UpSetPlot/blob/master/LICENSE v.0.9.0

• py3Dmol, https://github.com/avirshup/py3dmol license: https://github.com/avirshup/py3dmol/blob/master/LICENSE.txt

---

Contribution:

AMPcombi is a tool developed for parsing results from published AMP prediction tools. We therefore welcome fellow contributors who would like to add new AMP prediction tools results for parsing and alignment.

Adding a new tool to AMPcombi:

In ampcombi/reformat_tables.py

• add a new tool function to read the output to a pandas dataframe and return two columns named contig_id and prob_<toolname>
• add the new function to the read_path function

In ampcombi/ampcombi.py

• add a new parameter equivalent to the tool <--tool_file>.

---

Authors and credits:

The tool was written mainly by @louperelo and @darcy220606 with major scientific contributions from @RosaLuzia.

---

Funding:

This project was funded by Werner Siemens Foundation grant 'Palaeobiotechnology'

---