AMPSphere : the worldwide survey of prokaryotic antimicrobial peptides

INTRODUCTION

AMPSphere is a comprehensive catalog of antimicrobial peptides predicted using Macrel (DOI: 10.7717/peerj.10555) from 63,410 public metagenomes, ProGenomes v2.2 database (82,400 high-quality microbial genomes), and c.a. 4k non-whitelisted microbial genomes from NCBI.

GENERATION

Peptides were predicted using Macrel. Singleton peptides were removed, except those with a direct hit to the DRAMP database.

Redundant peptides were hierarchically clustered using CD-HIT (version 4.6) at 100%, 85%, 75%, and 50% of amino acid identity (and 90% of overlap of the shorter peptide). The obtained clusters were sorted by decrescent size and
numbered as families. Each level of clustering was called SPHERE and was used to understand the AMPs' structure accordingly to their orthology.

Nucleotide sequences from the most frequent variants per AMP also were included in this version of AMPSphere.

STATISTICS

AMPSphere v.2021-03 contains 863,498 sequences (avg length: 36 amino acids, range 8-98). DRAMP database was used to find confirmed sequences with strict homology to reference. This approach showed that 2,488 peptides were previously confirmed in our dataset.

IDENTIFIERS

Peptides are named:

>AMP10.XXX_XXX

Where `XXX_XXX` is a unique numerical identifier (starting at zero). Numbers were assigned in order of increasing
number of copies. So that the lower the number, the higher the number of copies of that peptide was present in the input data.

Annotations were also provided as separated fields in the fasta file, containing their:

- SPHERE families at level 4 (corresponding to hierarchically obtained clusters using 100%, 85%, and 75% of identity with a minimum overlap of 90% of the shorter gene).

Example of the header:

>AMP10.000_000 | SPHERE-III.001_396

VERSION

This version v.2021-03 brings the validated SPHERE families. Using the families from level III, containing 8 sequences or more, we calculated the alignments and trees in Newick format, which were included in this version too. No major changes were made in the other files besides the naming.

Files included:

AMPsphere_v.2021-03_families_tree_nwk.tar

AMPsphere_v.2021-03_families_alignment.tar

SPHERE_v.2021-03_levels.assessment.tsv

Files changed:

AMPsphere_v.2021-03.faa  -- Included SPHERE lv.III information

README -- update the information about the new files

Files eliminated:

No files were eliminated from this version when compared to the previous.

FILES

README.md
This file.

AMPSphere_v.2021-03.fna
Multi-fasta with AMPSphere gene sequences (nucleotide).

AMPSphere_v.2021-03.faa
Multi-fasta with AMPSphere peptide sequences (amino acid).

AMPSphere_v.2021-03.features.tsv
Table relating AMP name, and the features used for its prediction.
Columns:

1. AMP accession
2. tinyAA
3. smallAA
4. aliphaticAA
5. aromaticAA
6. nonpolarAA
7. polarAA
8. chargedAA
9. basicAA
10. acidicAA
11. charge
12. pI
13. aindex
14. instaindex
15. boman
16. hydrophobicity
17. hmoment
18. SA.Group1.residue0
19. SA.Group2.residue0
20. SA.Group3.residue0
21. HB.Group1.residue0
22. HB.Group2.residue0
23. HB.Group3.residue0

For more details about these features see the Macrel manuscript

AMPSphere_v.2021-03.origin_samples.tsv
TSV table relating AMP accession, sequence, and their origins in terms of prokaryotic genome or metagenome sample.

Columns:

1. AMP accession
2. GMSC accession  (comma separated list)
3. metagenome samples  (comma separated list)
4. proGenomes2 genomes  (comma separated list)

AMPSphere_v.2021-03.species.tsv
TSV table relating AMP name, sequence, and the species from which they were detected.

Columns:

1. AMP accession
2. proGenomes2 genomes
3. SpecI cluster

Note that AMPSphere was generated from the complete proGenomes v2 database.

However, after the initial release, many genomes were removed due to quality-control issues, leading to version 2.2 used for constructing this table.

DRAMP_anno_AMPSphere_v.2021-03.parsed.tsv
TSV table relating AMP name as query and the hits obtained with Blast against DRAMP database. The format is blast `outfmt6`.

Columns:

1. query
2. target
3. identity
4. alignment length
5. misalignment
6. gaps
7. query start
8. query end
9. target start
10. target end
11. e-value
12. score
13. target annotation
14. target function
15. target biochemical targets
16. target origin reference

AMPSphere_v.2021-03.hosts.tsv
TSV table relating AMPs with the hosts of host-associated metagenomes via metadata.

Columns:

1. AMP accession
2. host common name
3. host scientific name
4. host NCBI taxid                                                                            
5. counts

Column 5 (counts) measures are in the number of identical variants of a given peptide assigned to a common host.
   
 

AMPSphere_v.2021-03.locations.tsv
TSV table relating AMP name and their geographic location from metadata annotation of metagenome samples.

Columns:

1. AMP accession,
2. geographic location,
3. copies

Geographic location refers to the locale where the gene was found through metagenomics. It was assigned as a broad location such as country, ocean, continent (e.g. US, Atlantic Ocean, Arctic, Australia).
                                              
Counts are the number of identical variants of a given peptide assigned to a common location.

AMPSphere_v.2021-03.microontology.tsv
Table relating AMP name and their habitat of origin.

Columns:

1. AMP accession,
2. microontology,
3. counts

Microontology is a scheme used to annotate environments, it has different levels of complexity separated by ':'
                                              
Counts measures are in the number of identical variants of a given peptide assigned to a common habitat.

SPHERE_v.2021-03_levels.assessment.tsv
TSV table relating AMP name, sequence, and the hierarchically obtained clusters per level.

Columns:

1. AMP accession
2. evaluation vs. representative
3. SPHERE_fam level I
4. SPHERE_fam level II
5. SPHERE_fam level III

Levels of each SPHERE family:

I: contains clusters obtained with 100% of identity cut-off and 90% of overlap of the shorter sequence;

II: contains clusters obtained with the unclustered sequences and the representatives from level I at 85% of identity and 90% of overlap of the shorter sequence;

III: contains clusters obtained with the unclustered sequences and the representatives from level II at 75% of identity and 90% of overlap of the shorter sequence;

`evaluation vs. representatives` shows the percent of identity the sequence has in an alignment against the cluster representative, and also the overlap in percent.

Example:

* -- This means: this sequence is the cluster representative.

OR something like this:

77.50%,1:40:1:40 -- This means: alignment identity against the representative of the cluster equals 77.5% and the alignment start and end position for the query (1 and 40, respectively), and target (1 and 40, respectively).

AMPsphere_v.2021-03_families_alignment.tar

Compressed folder (`.tar.gz`) containing peptide alignments in FASTA format used for phylogenetic reconstruction. These alignments were built with Muscle for all SPHERE families larger than 7 sequences.

For more information, see the AMPsphere code, where you can find all parameters used in this step.

AMPsphere_v.2021-03_families_tree_nwk.tar
Compressed folder (`.tar.gz`) containing phylogenetic trees in Newick (`.nwk`) calculated from the peptide alignments built with Muscle per SPHERE family larger than 7 sequences.
    
Trees were built using FastTree with 1000 bootstraps as node support.

For more information, see the AMPsphere code, where you can find all parameters used in this step.

CONTACT

You can contact us via our discussion group.

AMPSphere main developers:

- Célio Dias Santos Júnior
- Yiqian Duan
- Luis Pedro Coelho

COPYRIGHT NOTICE

AMPSphere - the worldwide survey of prokaryotic antimicrobial peptides.

This work is a joint effort of the Big Data Biology group from the Institute of
Science and Technology for Brain-Inspired Intelligence (ISTBI) - Fudan
University, Shanghai, China, and the Structural and Computational Biology Unit
(Heidelberg) - European Molecular Biology Laboratory (EMBL).

Copyright (C) 2019-2021 The Authors

   AMPSphere IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
   EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
   OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
   IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
   DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
   OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
   USE OR OTHER DEALINGS IN THE SOFTWARE.

   This database is free; you can redistribute it and/or modify it
   as you wish, under the terms of the CC BY 4.0 license.

   You are allowed to:

   Share — copy and redistribute the material in any medium or format

   Adapt — remix, transform, and build upon the material for any purpose,
                 even commercially.

   You may also obtain a copy of the CC BY 4.0 license here.

REFERENCES CITED

1. Macrel: Santos-Júnior CD, Pan S, Zhao X, Coelho LP. 2020. Macrel: antimicrobial peptide screening in genomes and metagenomes. PeerJ 8:e10555. https://doi.org/10.7717/peerj.10555
2. ProGenomes: Mende DR, Letunic I, Maistrenko OM et al. 2020. proGenomes2: an improved database for accurate and consistent habitat, taxonomic and functional annotations of prokaryotic genomes.  Nucleic Acids Research 48(D1): D621–D625. https://doi.org/10.1093/nar/gkz1002
3. DRAMP: Kang X, Dong F, Shi C et al. 2019. DRAMP 2.0, an updated data repository of antimicrobial peptides. Sci Data 6, 148. https://doi.org/10.1038/s41597-019-0154-y