PHASTEST

PHASTEST Statistics

Quick Links

Table 1: PHASTEST's performance upgrades
Table 2: Feature comparison between PHAST, PHASTER and PHASTEST
Table 3: PHASTEST's evaluation (summary)
Table 4: PHASTEST's evaluation on GenBank annotated genomes
Table 5: PHASTEST's evaluation on unannotated genomes
Table 6: Comparison of Numbers of Proteins Annotated by Lite (Swissprot) and Deep (PHAST-BSD) Annotation Mode
Figure 1: Numbers of correct and false positive phage regions for E. coli
Figure 2: Performance Comparison of Prodigal and GLIMMER

Table 1: PHASTEST's performance upgrades

Cumulative set of performance enhancements	BLAST vs. phage DB runtime (s)	BLAST vs. bacterial DB runtime (s)	Total runtime on GenBank annotated genome (s)	Total runtime on unannotated genome (s)
PHAST (baseline) current DBs, no other upgrades	191	576	270	899
PHASTER (Past) - 2016 Data	47	48	126	227
PHASTER (Baseline) - Current DBs, no other upgrades	116	83	162	277
PHASTEST (Upgrade 1) - BLAST+ parameter adjustment	82	82	144	229
PHASTEST (Upgrade 2) - Whole-sequence Prodigal	81	71	141	201
PHASTEST (Upgrade 3) - Parallel Diamond*	84	124	266	118
PHASTEST (Upgrade 4) - Swissprot DB*	80	64	110	195

Details of PHASTEST’s performance upgrades, and its impact on 5.5 Mbp test genome (Escherichia coli O157:H7,GenBank accession NC_002655). Pipeline configuration capable of full genome-wide annotation is marked with (*).

Table 2: Feature comparison between PHAST, PHASTER and PHASTEST

Feature	PHAST (as of Jan 2011)	PHASTER	PHASTEST
Viral sequence database	~45,000 sequences	~187,000 sequences	~410,000 sequences
Bacterial sequence database	~4 million sequences	~9 million sequences, streamlined through CD-HIT filtering	Swissprot database, ~560,000 sequences
Computing cluster	32 CPU cores	112 CPU cores	116 CPU cores
BLAST	Legacy version 2.2.16	BLAST+ version 2.3.0+	BLAST+ version 2.3.0+, Diamond v2.0.14
Cluster use optimization	Rudimentary	Smart partitioning of query sequences and target bacterial DB; optimized execution parameters	Optimized execution parameters
Front-end server	Shared, single CPU	50% faster, dedicated	Dedicated 4 CPU cores
Front-end website	Perl and CGI	Ruby on Rails	Ruby on Rails
Genome viewer	Adobe Flash	JavaScript, AngularPlasmid and D3	CGView.js
Queuing system	Flat file	Uses Sidekiq for threading submissions	Uses Sidekiq for threading submissions
Recall previous user submissions	Bookmark page	“My Searches” feature or bookmark	"My Searches" feature or bookmark
Pre-computed genome results for quick query searching	0	>14,000	>14,000
Retrieve previously annotated genome results	GenBank accession or GI number only	GenBank accession, GI number, or full sequence	GenBank accession, GI number, or full sequence
Metagenomic data handling	NA	For raw sequence files only	For raw sequence files and whole-genome shotgun GenBank records
Annotation Target	Phage region only	Phage region only	Full genome

Table 3: PHASTEST's evaluation (summary)

	PHAST (2011)		PHASTER (2016)		PHASTEST (2022)
Input data	GenBank annotated genome	Sequence only	GenBank annotated genome	Sequence only	GenBank annotated genome	Sequence only
Sensitivity	85.4%	79.4%	86.9%	85.0%	87.6%	85.0%
Positive predictive value (PPV)	94.2%	86.5%	91.0%	87.3%	91.4%	91.2%
Prophages annotated in evaluation set	267	267	267	267	267	267
Prophages matched	228	212	232	227	234	227
Predicted prophages not present in evaluation set	14	33	23	33	22	22
Predicted prophages not in the evaluation set with evidence suggestive of being true prophages	N/A	N/A	12	11	8	7
Adjusted sensitivity	N/A	N/A	87.5%	85.6%	88.1%	85.4%
Adjusted PPV	N/A	N/A	95.7%	91.5%	94.5%	94.0%

Comparison of accuracy between the PHAST (2011), PHASTER (2016), and PHASTEST (2022). Evaluation was performed against manual prophage annotations in 54 bacterial genomes from Casjens, S. (2003) Molecular Microbiology, 49, 277-300. Because these manual annotations are a number of years old, predicted prophages were manually inspected, and it was determined that PHASTER and PHASTEST finds good evidence that several of these are in fact true prophages (these are marked below in Tables 4 and 5). Once the evaluation standard was adjusted to count these prophages as true positives, adjusted sensitivity and PPV values were calculated (last two rows).

Table 4: PHASTEST's evaluation on GenBank annotated genomes

Organism	Reference	PHAST	PHASTER	PHASTEST
>NC_000962, Mycobacterium tuberculosis H37Rv, 4411532 bp	2970551-2981576	2970065-2983874	2970063-2983874	2970063-2984654
>NC_000962, Mycobacterium tuberculosis H37Rv, 4411532 bp	1780643-1788505	1780643-1788505	1766989-1788505	1766987-1788505
>NC_000913, Escherichia coli K12, 4639675 bp	262552-296320	262124-296432	262898-297206	262898-297719
	2464567-2475651	2464378-2475651	2465301-2477629	2465301-2477629
	2754181-2775804	2753821-2780748	2754896-2777782	2746434-2777782
	564038-584856	563980-585282	564755-586057	564755-586057
	1410024-1432281	1404587-1432838	1395952-1435051	1395952-1435051
	1196090-1210402	1191881-1218961	1196867-1216671	1196867-1216671
	1631063-1650732	1631063-1662537	1619557-1656744	1619557-1656744
	2556793-2563354	False Negative	False Negative	2558575-2567625
	2064329-2076158	False Negative	False Negative	2068952-2082547
	False Positives	4505466-4540762	4491197-4548442	4491197-4548442 **
	False Positives	n/a	3716547-3722049	n/a
>NC_003112, Neisseria meningitidis MC58, 2272360 bp
	1001560-1005455	998362-1007364	False Negative	False Negative
	1099910-1133980	1101164-1133760	1101164-1133760	1101164-1133760

Detailed evaluation results for PHAST (2011) and PHASTER (2016) run on GenBank annotated genomes. Evaluation was made against manual prophage annotations in 54 bacterial genomes from Casjens, S. (2003) Molecular microbiology, 49, 277-300. Prophages predicted by PHASTER that were absent in the evaluation set were manually inspected, and several (marked below with **) were deemed to have good evidence of being true prophages rather than being false positives.

For statistics on more organisms click here.

Table 5: PHASTEST's evaluation on unannotated genomes

Organism	Reference	PHAST	PHASTER	PHASTEST
>NC_000962, Mycobacterium tuberculosis H37Rv, 4411532 bp	2970551-2981576	2970065-2980835	2970063-2980833	2970063-2984654
>NC_000962, Mycobacterium tuberculosis H37Rv, 4411532 bp	1780643-1788505	False Negative	1766987-1788115	1766987-1788487
>NC_000913, Escherichia coli K12, 4639675 bp	262552-296320	False Negative	False Negative	False Negative
	2464567-2475651	False Negative	2465301-2477629	2465301-2477629
	2754181-2775804	False Negative	False Negative	False Negative
	564038-584856	563980-585282	564755-586057	564755-586057
	1410024-1432281	1409925-1432985	1411899-1434959	1411969-1434812
	1196090-1210402	1197865-1215896	1198640-1216671	1198640-1216671
	1631063-1650732	1616881-1650732	1627517-1644304	1618855-1653915
	2556793-2563354	False Negative	False Negative	False Negative
	2064329-2076158	False Negative	False Negative	False Negative
	False Positives	n/a	3716547-3722049	n/a
>NC_003112, Neisseria meningitidis MC58, 2272360 bp
	1001560-1005455	998266-1007364	False Negative	False Negative
	1099910-1133980	1099110-1133760	1101164-1133966	1098337-1133966
	False Positives	916111-930967	916109-929811	n/a

Detailed evaluation results for PHAST (2011) and PHASTER (2016) run on unannotated genome sequences i.e. not genbank annotated (only the sequence is used). Evaluation was made against manual prophage annotations in 54 bacterial genomes from Casjens, S. (2003) Molecular microbiology, 49, 277-300. Prophages predicted by PHASTER that were absent in the evaluation set were manually inspected, and several (marked below with **) were deemed to have good evidence of being true prophages rather than being false positives.

For statistics on more organisms click here.

Table 6: Comparison of Numbers of Proteins Annotated by Lite (Swissprot) and Deep (PHAST-BSD) Annotation Modes

Sample	Swissprot Hits	Swissprot Misses	PHAST-BSD Hits	PHAST-BSD Misses
NC_000907.1	1647	22	1666	3
NC_000913.3	3955	28	3959	24
NC_000962.3	2832	938	3763	7
NC_000964.3	3698	48	3737	9
NC_002488.3	1174	574	1748	0

Comparison of numbers of non-phage proteins annotated via PHASTEST's lite mode (using Swissprot database) and deep mode (using PHAST-BSD database). Swissprot Misses denote numbers of proteins present in deep annotation mode but not found in the lite annotation, whereas PHAST-BSD Misses denote numbers of proteins present in lite annotation but not found in deep annotation mode.

For statistics on more organisms click here.

Figure 1: Numbers of correct and false positive phage regions for E. coli

Numbers of true and false positive phage regions for E. coli O157:H7 (NC_002655.2) contig inputs of varying length. Graph shows numbers of phage regions correctly identified relative to the E. coli O157:H7 full sequence annotation result. E. coli O157:H7 full sequence is predicted to possess 15 phage regions total.

Figure 2: Performance Comparison of Prodigal and GLIMMER

Accuracy comparison between Prodigal (used in PHASTEST) and parallel GLIMMER (used in PHASTER). Both programs were provided with FASTA format sequence input of 54 genome samples outlined in the table. Result of the ORF prediction was compared against pre-annotated CDS regions outlined in Genbank records of each genome. Sensitivity formula follows (amount of correct region prediction / total CDS identified in Genbank record), whereas accuracy was determined by (amount of correct predictions / (correct predictions + false positives)).

For detailed statistics on more organisms click here.

We require the use of cookies for essential features like storing your previously submitted PHASTEST queries. Rejecting the usage of cookies will result in certain features being disabled. By clicking ACCEPT or continuing to use the website you are agreeing to our use of cookies.