Skip to content

Spacepharer documentation

Spacepharer

As described by the github, Spacepharer is a program that uses the search method of MMSEQ2 to query spacer sequences against a database with the intent to find host-phage matches. Normally Spacer sequences have quite short sequences which confound more conventional methods such as BLASTN and therefore require stringent filtering methods. However phage databases are frequently incomplete and phages are generally highly mutable, necessitating that a certain amount of mismatches are allowed. This is what spacepharer overcomes by prioritising open reading frames which bypass the need for highly similar matches.

Spacepharer was therefore chosen as the method in CRISPRscape to identify spacer to phage matches compared to conventional methods such as BLASTN. With similar rationale, spacepharer was also chosen to match spacers with plasmids as they were expected to have similar problems such as virality causing high mutability and consisting mostly of coding sequences.

Databases

Spacepharer requires databases to match against. For this, CRISPRscape uses the scripts download_phage_database.sh and download_plasmid_database.sh, which download the Phagescope and PLSDB databases for phage and plasmid matching, respectively. These are also automatically prepared for use in Spacepharer. It is possible to use other databases but this requires multiple changes in the workflow code.

Phagescope

Phagescope is a large database containing a large set of bacteriophages combined from several sources and further annotated with information such as taxonomy. Specifically this database was chosen as it has a large variety of host species, sufficient annotation for CRISPRscape's goal and is relatively large compared to other databases. Furthermore, depending on further research, the extra annotations are extensive and could be used for further research.

As of writing Phagescope contains 873.718 phage sequences from 4.723 host species.

The full download comprises about 41GB, of which merged_metadata.tsv (127MB) and the fasta files are the most important.

  • merged_metadata.tsv lists for each phage its genome length, GC-content, taxonomy, completeness (e.g., 'High-quality'), host species, lifestyle (e.g, 'virulent'), and which cluster and subcluster it belongs to.

From these, a custom database is build, which uses 155GB.

$ ls -sh resources/phagescope/
total 41G
1.4G CHVD.fasta      4.0G GOV2.fasta   8.2G IMG_VR.fasta              97M RefSeq.tar.gz
400M CHVD.tar.gz     1.1G GOV2.tar.gz  2.4G IMG_VR.tar.gz             98M STV.fasta
 12M DDBJ.fasta      5.1G GPD.fasta    8.4G MGV.fasta                 28M STV.tar.gz
3.6M DDBJ.tar.gz     1.6G GPD.tar.gz   2.6G MGV.tar.gz               2.5G TemPhD.fasta
8.5M EMBL.fasta      560M GVD.fasta    127M phagescope_metadata.tsv  770M TemPhD.tar.gz
2.6M EMBL.tar.gz     172M GVD.tar.gz   224M PhagesDB.fasta
147M Genbank.fasta   241M IGVD.fasta    68M PhagesDB.tar.gz
 45M Genbank.tar.gz   70M IGVD.tar.gz  318M RefSeq.fasta

PLSDB 2024_05_31_v2

PLSDB is an up to date database containing plasmids taken from NCBI and INSDC that have been scrubbed of chromosomal sequences. Furthermore PLSDB has extensive annotation including taxonomy and antimicrobial resistance genes.

This database was chosen for its wide variety of host species, origin location and being of sufficient size. Furthermore, as of writing it is still getting updates.

The version used by CRISPRscape (PLSDB 2024_05_31_v2) contains 72.360 plasmid sequences from 9436 different host species.

The full download uses 14GB of disk space, but the essential files for CRISPRscape are smaller:

  • sequences.fasta: 7GB

  • self-explanatory: fasta file with the plasmid sequences

  • nuccore.csv: 17MB

  • Metadata file with unique ID, accession number, description, creation date, predicted completeness, 'Genome' (plasmid), length, and 14 more columns with a.o. source information

  • taxonomy.csv: 2.3MB

  • Taxonomic tree for all included plasmids as comma-separated table, including both taxonomic names for each rank and taxID numbers.

These are used to build a custom database of 68GB.

$ ls -sh resources/PLSDB/
total 14G
 37M amr.tsv                         8.0K ecopaths.csv          1.7G proteins.fasta
3.1M assembly.csv                    780K nucc_identical.csv     16K README.md
 16K biosample_attributes_plsdb.csv   17M nuccore.csv           7.0G sequences.fasta
 12M biosample.csv                    16M plasmidfinder.csv     2.3M taxonomy.csv
2.5M changes.tsv                      14M plsdb_mashdb_sim.tsv   23M typing.csv
1.7M disease_terms.csv               569M plsdb_sketch.msh       75M typing_markers.csv
3.5G download_meta.tar.gz            610M proteins.csv

Output

Spacepharers output consists of a singular .tsv file showing all matches to a phage_ID within the database provided. The expected output takes the form of:

#prok_acc  phage_acc   S_comb      num_hits
>spacer_acc      phage_acc   p_bh    spacer_start      spacer_end  phage_start phage_end   5'_PAM|3'_PAM    5'_PAM|3'_PAM(reverse strand)

where the # indicates the prokaryotic match. However due to the way that input is provided, this ends up being redundant. CRISPRscape therefore filters out the # sections retaining only the individual matches.

of the matches the columns consist of: spacer accession, phage accession, p besthit, spacer start and end, phage start and end, possible 5’ PAM|3’ PAM, possible 5’ PAM|3’ PAM on the reverse strand

Lastly the matches are then matched with their respective metadata and added onto the table. Crisprscape will by default take the metadata from Phagescope and PLSDB and add the relevant columns.

For Phagescope these columns are: Length, GC_content, taxonomy, completeness, host, lifestyle

For PLSDB the column added is: Host taxonomy