Use of profile hidden Markov models in viral discovery: current insights

Figures & data

Figure 1 Multiple sequence alignment of VP1 (major capsid protein) sequences from Alpavirinae phages.

Notes: Multiple sequence alignment was performed with ClustalX using default parameters.⁵⁷ Colors indicate conservation of residues according to the ClustalX color scheme.⁵⁸

Table 1 Probability values for positions 46–49 from the multiple sequence alignment depicted in Figure 1

Amino acid	Position 46	Position 47	Position 48	Position 49
A	0.05	0.00	0.00	0.00
C	0.00	0.00	0.00	0.05
F	0.00	0.00	0.00	0.25
I	0.00	0.20	0.00	0.10
L	0.95	0.10	0.00	0.00
N	0.00	0.00	0.00	0.20
P	0.00	0.00	1.00	0.00
T	0.00	0.10	0.00	0.20
V	0.00	0.60	0.00	0.20
Y	0.00	0.00	0.00	0.00

Table 2 Probability values with pseudocounts for positions 46–49 from the multiple sequence alignment depicted in Figure 1

Amino acid	Position 46	Position 47	Position 48	Position 49
A	0.0500	0.0167	0.0167	0.0167
C	0.0167	0.0167	0.0167	0.0500
F	0.0167	0.0167	0.0167	0.1833
I	0.0167	0.1500	0.0167	0.0833
L	0.6500	0.0833	0.0167	0.0167
N	0.0167	0.0167	0.0167	0.1500
P	0.0167	0.0167	0.6833	0.0167
T	0.0167	0.0833	0.0167	0.1500
V	0.0167	0.4167	0.0167	0.1500
Y	0.0167	0.0167	0.0167	0.0167

Note: The matrix was constructed using pseudocounts of 0.5.

Figure 2 Diagram representing a profile hidden Markov model (profile HMM).

Notes: Match states are represented as red rectangles, deletion (silent) states as green circles, and insertion states as blue diamonds. The red numerical values next to the arrows indicate transition probabilities. The equalities inside the states indicate amino acid probabilities, generally called emission probabilities. These emission probabilities do not include the use of pseudocounts. Match states use emission probabilities computed from the original alignment; insertion states use background amino acid probability values of 1/20. The transition probabilities highlighted with red circles indicate the probabilities described in the text. The other transition probabilities were arbitrarily set to make the figure more homogeneous and to increase clarity.

Table 3 Web resources of viral profile HMM databases and tools

Database	Description	Download availability	Last update	Reference
GyDB	Gypsy Database of mobile genetic elements and viruses; comprises a lineage-specific collection of 314 profile HMMs; provides several online tools, including similarity searches against the profile HMMs Source: http://gydb.org/	No	2010	Llorens et al^Citation55
HIVAlign	HIVAlign is one of the numerous online tools offered by HIV databases. The program accepts as input aligned or unaligned sequences, which are then aligned against a curated reference profile HMM database Source: https://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html	No	2016	Foley et al^Citation56
vFam	vFam is a database of profile HMMs built from all viral protein sequences available at RefSeq. Viral protein sequences, annotations, and profile HMMs are provided Source: http://derisilab.ucsf.edu/software/vFam/	Yes	2014	Skewes-Cox et al^Citation35
Viral OGs	Subset of eggNOG v. 4.5, composed of viral sequences, annotations, alignments, trees, and profile HMMs Source: http://eggnogdb.embl.de/#/app/viruses	Yes	Current	Huerta-Cepas et al^Citation37
pVOGs	Prokaryotic Virus Orthologous Groups is a database of orthologous groups built from genomes of viruses that infect bacteria and archaea. Provides accession IDs of viral proteins, lists of orthologous groups, alignments, and profile HMMs Source: http://dmk-brain.ecn.uiowa.edu/pVOGs	Yes	Current	Grazziotin et al,^Citation36 Kristensen et al,^Citation38 Kristensen et al^Citation39

Abbreviation: profile HMMs, profile hidden Markov models.

Figure 3 Distribution of orthologous groups from vFam³⁵ (A) and pVOGs³⁶ (B) according to the viral families.

Notes: To obtain quantitative data, the number of corresponding profile HMM/orthologous groups was determined for each viral family based on the annotation provided in the database files. Profile HMMs from the original databases are derived from viruses of either single or multiple families.
Abbreviations: pVOGs, Prokaryotic Virus Orthologous Groups; vFAM, viral profile HMM database; profile HMMs, profile hidden Markov models.

Table 4 Publicly available targeted assembly tools that use profile HMM seeds

Program	Comments	Assembly	Documented	Reference
HMM-GRASPx	Gene-sized assemblies; implements its own novel profile HMM-based search; can be used either as a search tool or as an assembler; further extends contigs based on pre-constructed extension links (overlap graph) Source: https://sourceforpe.net/project.s/hmm-graspx/	Targeted: gene-sized	Partially	Zhong et al^48
MegaGTA	Iterative and succinct de Bruijn graph-based; tests multiple k-mers in each run; demands high memory; free from false-positive k-mers; gene-sized assemblies Source: https://github.com/HKU-BAL/megagta	Targeted: gene-sized	No	Li et al^45
SAT-Assembler	Uses profile HMM seeds for gene-sized assembly; builds and traverses overlap graph for contig generation Source: https://github.com/zhangy72/SAT-Assembler	Targeted: gene-sized	No	Zhang et al^43
Snowball	Aims at reconstructing different strains in metagenomic datasets; uses profile HMM seeds; algorithm based on read overlaps; uses overlapping ends paired- end library; gene-sized assemblies Source: https://github.com/hzi-bifo/snowball	Targeted: gene-sized	No	Gregor et al^44
Xander	Based on De Bruijn graph; uses only one k-mer in each run; demands high memory for good false-positive performance; gene-sized assemblies Source: https://github.com/rdpstaff/Xander assembler	Targeted: gene-sized	Yes	Wang et al^46
GenSeed-HMM	Integrates targeted and progressive assembly: generates contigs potentially much longer than original seed; seed-driven (DNA, protein, or profile HMM); depends on external assemblers; low memory and processor use Source: https://sourceforpe.net/projects/penseedhmm/	Targeted and progressive: unrestrained contig length	Yes	Alves et al^41

Notes: We have considered as properly documented those programs that present an external file (and not just help on the command-line or a short README file) with detailed description of program installation, operation, and options. Partially documented programs include a README file (either in the downloadable package or directly online) with reasonably detailed installation, option, and usage information (that is not simply a repetition of the command-line help). GenSeed-HMM also includes a detailed step-by-step tutorial with included example files.

Abbreviation: profile HMMs, profile hidden Markov models.

Figure S1 Distribution of number of proteins per orthologous group for vFam¹ (A) and pVOGs² (B).

Notes: Data were obtained from the annotation files provided by the database authors and bins of size 10 were used for building the histograms. For increased readability, pVOGs data are shown only up to 1,000 proteins per orthologous group (just six groups presented numbers larger than that, up to a maximum of 8,131 proteins in the largest group).
Abbreviations: pVOGs, Prokaryotic Virus Orthologous Groups; vFAM, viral profile HMM database.

Larkin MA, Blackshields G, Brown NP, et al. Clustal W and Clustal X version 2.0. Bioinformatics. 2007;23(21):2947–2948.

 PubMed Web of Science ®Google Scholar

Gibson T, Higgins D, Thompson J [homepage on the Internet]. General help for CLUSTAL X (2.0). Available from: http://www.clustal.org/download/clustalx_help.html. Accessed May 22, 2017.

 Google Scholar

Llorens C, Futami R, Covelli L, et al. The Gypsy Database (GyDB) of mobile genetic elements: release 2.0. Nucleic Acids Res. 2011;39(Database issue):D70–D74.

 Google Scholar

Foley B, Leitner T, Apetrei C, et al, editors. HIV Sequence Compendium 2013. New Mexico: Theoretical Biology and Biophysics Group, Los Alamos National Laboratory; 2013.

 Google Scholar

Skewes-Cox P, Sharpton TJ, Pollard KS, DeRisi JL. Profile hidden Markov models for the detection of viruses within metagenomic sequence data. PLoS One. 2014;9(8):e105067.

 Web of Science ®Google Scholar

Huerta-Cepas J, Szklarczyk D, Forslund K, et al. eggNOG 4.5: a hierarchical orthology framework with improved functional annotations for eukaryotic, prokaryotic and viral sequences. Nucleic Acids Res. 2016;44(D1):D286–D293.

 PubMed Web of Science ®Google Scholar

Grazziotin AL, Koonin EV, Kristensen DM. Prokaryotic Virus Orthologous Groups (pVOGs): a resource for comparative genomics and protein family annotation. Nucleic Acids Res. 2017;45(Database issue):D491–D498.

 Google Scholar

Kristensen DM, Cai X, Mushegian A. Evolutionarily conserved orthologous families in phages are relatively rare in their prokaryotic hosts. J Bacteriol. 2011;193(8):1806–1814.

 Web of Science ®Google Scholar

Kristensen DM, Waller AS, Yamada T, Bork P, Mushegian AR, Koonin EV. Orthologous gene clusters and taxon signature genes for viruses of prokaryotes. J Bacteriol. 2013;195(5):941–950.

 Web of Science ®Google Scholar

Skewes-Cox P, Sharpton TJ, Pollard KS, DeRisi JL. Profile hidden Markov models for the detection of viruses within metagenomic sequence data. PLoS One. 2014;9(8):e105067.

 Web of Science ®Google Scholar

Grazziotin AL, Koonin EV, Kristensen DM. Prokaryotic Virus Orthologous Groups (pVOGs): a resource for comparative genomics and protein family annotation. Nucleic Acids Res. 2017;45(Database issue):D491–D498.

 Google Scholar