"The Opera of PhAnToMe": Phage Annotation Tools at the 20th Biennial Evergreen International Phage Meeting

The Opera of PhAnToMe

Ramy K. Aziz (Twitter: @azizrk)Aug 04 2013

opus (LT) = work (Pl. opera)

The environment, the toolbox, and the community

Phage Genomics Workshop, Evergreen 2013

08/04/2013

Past,

Phage Genomics - Evergreen 2013

NSF-funded, 3-year project (09-12) to develop

PhageAnnotationTools andMethods

Four Centers:- SDSU, San Diego, CA- VCU, Richmond, VA- USF, St. Pete FL- UA, Tucson, AZ

http://www.phantome.org

08/04/2013

… present, ...


?TBA

08/04/2013

… and future


Aims• Direct

– Discuss the concepts behind RAST– Quickly preview several tools developed under (or

under influence of) the PhAnToMe project– Demonstrate online, community annotation using

SEED

• Indirect {hidden agenda ;)}– PhAnToMe 2.0?– Establish community annotation efforts/

crowdsourcing– Seek Funding? Crowdfunding?

08/04/2013 Phage Genomics - Evergreen 2013

Outline• The environment (the SEED)

– The SEED and the ‘Subsystems Technology’

• The toolbox (PhAnToMe and sequels)– PHAST and RAST– PhACTS– PhiSPy– iVireons

• The community– Online annotation process – Annotation jamboree(s)– Course design


$$

Writing proposals, applying for grants

I. THE ENVIRONMENTThe Opera of PhAnToMe


I. The Environment: SEED

http://theseed.org

08/04/2013

Aziz RK,, et al. (2012) PLoS ONE 7(10): e48053. doi:10.1371/journal.pone.0048053


SEED: Main concept

One genome

All genomes


SEED: Main concept

One genome

All genomes


“Subsystems-based technologies were developed in the SEED with the view that the interpretation of one genome can be made more efficient and consistent if hundreds of genomes are simultaneously annotated in one subsystem at a time”

Aziz RK,, et al. (2012) PLoS ONE 7(10): e48053. doi:10.1371/journal.pone.0048053

SEED: Main concept• Protein-based database

Jargon: PEG = protein-encoding gene

• The subsystems approach

and• FIGfams: protein families based on

– sequence similarity– chromosomal co-occurrence, gene order,

synteny– human curation, evidence-based expert

assertions08/04/2013 Phage Genomics - Evergreen 2013

RAST: automated annotation


08/04/2013

What is a subsystem?• “A subset of functional roles studied across genomes”• A spreadsheet where:

– each row represents a genome– each column represents a functional role/ feature/ protein– different patterns = variants

Function 1 Function 2 … Function n

Genome a

Genome b

…

Genome z


08/04/2013

What is a subsystem?


Advantages of subsystems

Subsystems-basedannotation


Annotation Reconstruction

from genome from metagenome


Incomplete

frameshift

- complete- accurate

Credit: Andrew Kropinski Credit: Bas Dutilh

faulty assembly

Annotation Reconstruction

from genome from metagenome

08/04/2013

Incomplete faulty assembly

frameshift

- complete- accurate


Credit: Andrew Kropinski Credit: Bas Dutilh

II. THE TOOLBOXThe Opera of PhAnToMe


II. PhAnToMe ToolBoxhttp://www.phantome.org


II. The ToolBox: RAST• (At least) Four ways to annotate a genome via

RAST:

– myRAST (local)

• uses the server but you can edit offline)

– RAST (http://rast.nmpdr.org)

• annotates online, saves your genome on server

– “PhAST” (http://www.phantome.org/PhageSeed/Phage.cgi?page=phast)

• optimized gene-calling

– Use your favorite gene caller then upload gbk file to RAST


http://rast.nmpdr.org


phiRAST complaints• ORF/Gene calling

• tRNA– bug fixed, but still follow Andrew’s advice

• Too many hypotheticals, etc. – manual annotation, see later

– need for expert annotations, community contribution

– funding


“PhAST”: some improvement?


“PhAST”: some improvement?


PHAST: Disambiguation


Other tools• PHACTS:

– classifies and predicts lifestyle

• PhiSpy: – finds prophages

• iVireons– predicts phage structural proteins, holins,

more to come


II. The ToolBox: PHACTS• PHAge Classification Tool Set

• Uses a novel similarity algorithm and a supervised Random Forest classifier to predict whether the lifestyle of a phage, described by its proteome, is virulent or temperate.

• The similarity algorithm creates a training set from phages with known lifestyles and along with the lifestyle annotation, trains a Random Forest to classify the lifestyle of a phage.

• PHACTS predictions have had a 99% precision rate.

08/04/2013 Phage Genomics - Evergreen 2013 Kate McNair

PHACTS

• Out of the 227 phages with a known lifestyle, PHACTS was able to confidently and correctly calculate the lifestyle of 197 phages.

• Only 2 phages were predicted confidently wrong: The two phages that were confidently incorrectly classified were both virulent phages that contained a functional integrase


PHACTS• http://www.phantome.org/PHACTS/

• Other applications• Host prediction: whether a phage infects a Gram

positive or Gram negative bacteria• Taxonomy prediction: a phage’s Family


http://www.phantome.org/PHACTS/

PHACTS


II. The ToolBox: PhiSpy

Calculate genomic characteristics

Classifyprophage region

Evaluate predicted prophages

• Transcriptional Strand Orientation• Customized AT skew• Customized GC skew• Protein length • Abundance of Phage words

• Random Forest• Pre calculated training genome• Input bacterial genome

• Produce a rank for each gene

• Phage insertion points• Similarity of phage proteins

08/04/2013 Phage Genomics - Evergreen 2013 Sajia Akhter

PhiSpy

• Performance comparison in 50 complete bacterial genomes

Applications %Identified %FN %FP

Prophinder 89% 11% 12%

Phage_finder 82% 18% 1.33%

PhiSpy 94% 6% 0.66%

08/04/2013 Phage Genomics - Evergreen 2013 Sajia Akhter

• Download: PhiSpy – http://sourceforge.net/projects/phispy

• PhiSpy is on Kbase– http://kbase.science.energy.gov

• Web version under final development

• Ran PhiSpy on 4,335 bacterial genomes

• Predicted 12,826 prophages in 3,203 genomes

– 9,101 known prophages

– 3,723 undefined prophages08/04/2013 Phage Genomics - Evergreen 2013

PhiSpy

Sajia Akhter

http://kbase.science.energy.gov/

iVIREONS – http://vdm.sdsu.edu/ivireons

Victor Seguritan

Victor Seguritan

Application of Artificial Neural Networks (ANNs)

to Viral Dark Matter

Viral Hypothetical Protein Sequences

Known

eval <= 0.001

Conserved Domain DB (rpsblast)

Keep sequences ≥ 200 aa

no hit OR e-value > 0.001

no hit OR e-value > 0.001

eval <= 0.001

Reference Sequence DB(tblastp)

Artificial Neural Networks (ANNs)

Remove ≥ 80% identical sequences

Synthesize ANN-predicted Hypothetical Protein Genes

Clone in E.coli

Purification By Cobalt Affinity

Validation by TEM or X-ray Crystallography


“FAMILIES” OF ANNs

1) General structural proteins:

2) Phage major capsid proteins

3) Phage tail/tail fibers/collar etc.

4) Holins

5) Portals

• Trained with all types of proteins• Both phages & viruses


Victor Seguritan

1


2Enter User Info

VibrioPhage

[email protected]

3Upload Sequences

Victor Seguritan

4 View Results

5Copy Results to a Spreadsheet


- Structural 1:1- MCP 1:1- MCP 2:1- MCP 3:1- MCP 4:1- MCP 7:1- MCP 22:1

(lambda)- Tail 1:1- Tail 2:1- Tail 4:1- Tail 7:1- Tail 6.6:1

(lambda)

Stringencies Reported


III. THE COMMUNITYThe Opera of PhAnToMe


SEED allows continuous annotation

08/04/2013

SEED

RAST

GenomesSubsystems

SEED Viewer

New Genomes

Subsystems Editor


SEED allows community annotation


Later in the meeting, • Who might be interested in putting

together:a) an outline for an annotation jamboree/

workshop with phage experts

b) a syllabus/outline for a course to get undergraduate/graduate students to annotate specific subsystems

c) a proposal to get funding for community annotation efforts

d) all above


POST SCRIPTUMThe Opera of PhAnToMe


Aims• Direct

– Discuss the concepts behind RAST– Quickly preview several tools developed under (or

under influence of) the PhAnToMe project– Demonstrate online, community annotation using

SEED

• Indirect {hidden agenda ;)}– PhAnToMe 2.0?– Establish community annotation efforts/

crowdsourcing– Seek Funding? Crowdfunding?


If you use, please cite• SEED, RAST, myRAST, phiRAST, PHAST:

– RAST, BMC Genomics 2008 and SEED servers: PLoS ONE 2011

• Other tools– PHAST: McNair et al. PMID: 22238260; PhiSpy: Akhter et al. PMID:

22584627; iVireons: Seguritan et al. PMID: 22927809

• Letters of support


AcknowledgmentsRobert A. Edwards, PhD

• PhiRAST development: Ross Overbeek, Robert Olson, Gordon Pusch, Terry Disz, Bruce Parrello

• Phage annotators (Phantomers): Bhakti Dwivedi, Mya Breitbart, et al.

• FIG and all SEED annotators:VeronikaV, SvetaG, OlgaV/Z, et al.

Sajia Akhter

08/04/2013

$$


& NSF

$$& NSF

Acknowledgments

• PHAST

Victor Seguritan

08/04/2013

Katelyn McNair

• iVireons


"The Opera of PhAnToMe": Phage Annotation Tools at the 20th Biennial Evergreen International Phage Meeting

Technology

phage annotation tools

phage structural proteins

lifestyle annotation

seed http

annotation reconstruction

automated annotation

manual annotation

community annotation