Where is ssu rrna found

A total of sequences belonging to Euglenida were recovered, all of them from nominally free-living organisms Figure 1 C, Figure S6 and Figure S7. Of these, almost all environmental— A total of sequences including environmental sequences fell among the clades of phagotrophic euglenids.

This bias toward Petalomonadida is likely due to the divergent nature of most other Euglenida SSU rRNA gene sequences, which makes it hard to capture this diversity using universal primers Additionally, the V4 region of Euglenida tends to be massively expanded past the bp limit of current high-throughput sequencing technology, but Petalomonadida represents an exception to this trend.

The majority of Euglenophyceae sequences are from freshwater, while few sequences come from soil and marine environments.

This might reflect their natural distribution or point to an undersampling of diversity from soil and marine environments.

They are a clade of heterotrophic flagellates overwhelmingly known from marine habitats. Of these, all but one, belonging to the uncultured type species Eupelagonema oceanica 45 , lack any formal taxonomic description.

An overwhelming majority of the Diplonemea sequences was environmental, with most obtained from marine plankton, some from a hydrothermal plume or oxygen-depleted sea water, while only six were benthic Figures 1 B and D.

Only a handful of sequences Diplonema papillatum, Hemistasia phaeocysticola , two strains of Rhynchopus and one sequence belonging to the Eupelagonemidae were found to be likely host-associated.

It is predicted that a vast undiscovered diversity is hidden within the taxon Eupelagonemidae KIN1 comprises 16 environmental sequences from various deep-marine environments. Metakinetoplastina contain four subgroups Neobodonida, Parabodonida, Eubodonida, and Trypanosomatida , with the Neobodonida being the deepest branching clade, and the Trypanosomatida being most closely related to the Eubodonida 46— Some of these clades especially Dimastigella spp.

Parabodonida contains four clades corresponding to four described genera: Parabodo free-living organisms from terrestrial and freshwater biomes and potential parasites found in plant sap , Cryptobia endoparasites of snails , Procryptobia free-living marine organisms , and Trypanoplasma fish blood parasites; including some species originally assigned to Cryptobia.

Eubodonida are free-living bacterivorous protists found in soil, freshwater and marine habitats. In Metakinetoplastina, most of the sampling and sequencing efforts have been dedicated to parasitic Trypanosomatida, and the level of misannotations was generally very low. We have improved on the GenBank classification and corrected the assignment of many sequences such that all of the sequences now have rational higher-level taxonomic labels.

A majority of described Metamonada and Discoba species are characterized exclusively by morphological features with no molecular data available; consequently, we have a limited ability to assign species labels to the molecular diversity of Excavata in environmental samples. The presented EukRef databases highlight several interesting patterns in the diversity of excavates and exposes the uneven and varied historical effort of protistologists in the study of these groups.

First of all, for several groups, there is a large imbalance between environment-, culture-, or isolate-derived data Figure 1 B. Most of the Heterolobosea, Fornicata, and Parabasalia sequences come from uncultured isolates, while jakobids and diplonemids are mostly derived from environmental sequences, but most Euglenida come from established cultures in culture collections.

These differences likely represent biases caused by the fact that representatives of some groups are generally easier to cultivate than others. The databases are also likely biased toward taxa that predominantly live in easily accessible environments e. North America and Europe. North America and Europe are the most commonly sampled locations, while Africa, the Arctic and Antarctica are the least sampled locations, which suggests that tropical and polar biodiversity is poorly investigated Figure 1 E.

There is also an apparent lack of environmental studies aimed at predominantly host-associated groups such as Fornicata, Preaxostyla or Parabasalia Figures 1 C and D. This is particularly disconcerting as a large part of the known diversity of Preaxostyla and Parabasalia is described from terrestrial animals, and these protists are typically transferred directly from host-to-host without having a free-living or cyst stage as opposed to many parasites of aquatic organisms.

Consequently, a significant part of the diversity of these clades will be excluded from most environmental studies, along with the hosts. The current trend in sequencing short amplicons encourages analyses of SSU rRNA from host environments, which can reveal novel diversity; however, due to the divergence of SSU rRNA in many eukaryotes, unknown groups may be missed and longer sequences are necessary to more fully resolve the taxonomic and phylogenetic affinities of eukaryotes.

These disparities in diversity could accurately represent the extant diversity. However, as discussed above, these differences are more likely caused by experimental and sampling biases, therefore much more diversity is awaiting discovery. The high genetic divergence of the SSU rRNA gene might be partially responsible for this bias and proper investigation of the diversity of some of the excavate groups might require using group-specific primers and longer sequences than the usual short sequence tags.

Following the EukRef pipeline eukref. For each dataset, several distantly related SSU rRNA gene sequences were included as outgroups to root the phylogenetic tree. The phylogenetic tree for each dataset was visually inspected, and sequences resulting in long, errant branches were removed from the dataset. Using the SSU rRNA reference trees, a full classification was manually assigned to each tip based on phylogenetic support and currently accepted taxonomy.

Classification was assigned for each taxonomic level for each sequence based on the position of each sequence in a phylogenetic clade. If necessary, the taxonomy for a sequence was modified to be consistent with all other sequences in a clade following the best studied or type species.

Empty taxonomic ranks were labeled with the taxon name of the higher rank. Well-supported clades consisting of unclassified environmental sequences were given abbreviated names a prefix of the lowest assigned taxon in capital letters followed by a number according to EukRef guidelines. A taxon name was not entered for a taxonomic level in the database if it was not possible to distinguish sequences associated with this taxon name as a distinct clade from the SSU rRNA trees, particularly at the species level even though sequences may be derived from organisms described as different species, or even deeper taxa, based on other data.

For these cases, these taxonomic levels were filled with the taxon name at the higher level. Metadata for each sequence was obtained from GenBank entries or by referring to research publications and culture collection databases. Statistical support for each branch was calculated based on bootstrap replicates.

We also thank Ana Sandoval Tamayo for assistance in generating figures. Supplementary data are available at Database Online. Science , , Google Scholar. Del Campo J. PLoS Biol. Simpson A. Hampl V. USA , , — Adl S. Heiss A. Open Sci. Cavalier-Smith T. Brown M. Genome Biol. Kolisko M. BMC Evol. Handbook of the Protists. Springer , Cham , pp. Google Preview. Maritz J. Brune A. Flegontova O. Gawryluk R. Park J. Protist , , — Guillou L.

Nucleic Acids Res. Quast C. Moriya S. Zhang Q. Keeling P. Silberman J. Leger M. Gile G. Lara E. Kamikawa R. Yabuki A. Yang J. Heterolobosea , a novel marine anaerobic protist with strikingly derived morphology and life cycle. De Jonckheere J. Protist , , 89 — Lax G. Lukomska-Kowalczyk M. Tashyreva D. Okamoto N.

Moreira D. Deschamps P. Yazaki E. Genes Genet. Callahan H. Ichthyobodo Necator. Mukherjee I. FEMS Microbiol. Goodwin J. Katoh K. Bioinformatics , 9 , — Capella-Gutierrez S. Bioinformatics , 25 , — Stamatakis A. Bioinformatics , 30 , — Citation details: Kolisko, M.

Database Vol. Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide. Sign In or Create an Account. Sign In. Advanced Search. Search Menu. Article Navigation. Close mobile search navigation Article Navigation.

Volume Article Contents Abstract. Results and discussion. Concluding remarks. Experimental procedures. Supplementary Data. Martin Kolisko , Martin Kolisko. Oxford Academic. Olga Flegontova. Anna Karnkowska.

Gordon Lax. Julia M Maritz. Department of Zoology, Charles University. Jane M Carlton. Alastair G B Simpson. Vera Tai. Tree-based assignment is, therefore in theory, a more robust approach [28] and current FLX Titanium longer reads now make it possible to extract phylogenetic information with a high degree of reliability [29]. On the other hand, phylogenetic analyses allow for the description of clades, which may lead to new insights into the structure and functioning of ecosystems, as previously mentioned.

Although more robust, these methods are less frequently used than BLAST or probabilistic classifiers, as they require more computing resources Table S1. Though large computational capacity is now more accessible e.

In this work, we introduce a tree-based treatment designed for analyzing massively parallel sequencing outputs that automatically affiliates sequences from SSU rRNA gene amplicons and builds phylogenetic trees composed of very large numbers of sequences. As short-read sequence data e. Designed for the analysis of any microorganism protists, Bacteria and Archaea , the value of this treatment is highlighted here on the protist diversity as the pipelines dedicated to the study of eukaryotic pyrotags are still scarce.

Indeed, 16S rRNA gene reads were widely investigated in previous studies [31] — [33] to assess bacterial diversity, which enhanced the development of specific 16S rRNA gene analytical tools. However, 18S rRNA gene surveys and tools allowing for the accurate and rapid taxonomic affiliation of protists from NGS data are needed because the number of studies dealing with protists diversity is currently increasing e.

Secondly, the different methods of taxonomic assignment i. Thirdly, the accuracy of phylogenetic affiliation was compared on amplicons covering different variable regions V1 to V9 , and finally, a dataset of original pyrosequencing data obtained from lacustrine small protists was analyzed by the tree-based approach that was developed.

In the analysis of near full-length reference sequences of 18S rRNA gene, taxonomic groups were found in similar proportions to those initially present in the samples. At the finest phylogenetic level studied i. Thus, our phylogenetic affiliation method outperformed the other methods on near full-length sequences.

However, as environmental sequences are generally quite divergent from referenced ones and their affiliation needs to be checked manually, sequences belonging to freshwater clades [6] , [7] were also processed by our phylogenetic affiliation method to evaluate how it behaved on these datasets.

Accuracy of the phylogenetic affiliation in relation with the variable region targeted. Because the V8—V9 region is often missing in public databases, the results obtained from this region were based only on sequences included in the reference database. The affiliation results at the genus rank differed according to length, variability within the studied region and method used for taxonomic affiliation Table 1. Considering the affiliation methods, LCA specificity was higher than that of NN for fragments of bp only for the V1 and V8 regions, and LCA specificity was always better for fragments of bp.

In this last region, we observed a decrease in the accuracy of the affiliation, coupled with a sharper decline for the phylogeny-based affiliations. The specificity therefore varied between In addition to the accuracy of assignment, this phylogenetic affiliation method was developed to optimize processing time for large datasets. The run time increased with the number of OTUs, regardless of the length. The reliability of affiliations was compared for bp reads spanning the 18 S rRNA gene for four taxonomic groups: Alveolata, Stramenopiles, Fungi and Viridiplantae at the genus level Figure 1.

Generally, the fragment affiliation depended on the taxonomic group and the region considered. According to previous results, the regions from V5 to V6 gave, on average, the weakest accuracy. Another general trend observed in this analysis was a poor taxonomic restitution for sequences belonging to Viridiplantae compared to other groups, between The best specificity values for Stramenopiles, Alveolata and Fungi were obtained in different regions: V1—V2 The taxonomic affiliation for these three groups from the V8—V9 region was relatively similar, from In silico simulations have shown that primers NSF and NSR, used to target the V4 region of the 18S rRNA gene captured the greatest diversity data not shown and that the region amplified by these primers is suitable for taxonomic affiliation Table 1.

The diversity and richness indexes obtained for each environment are shown in Table 2. The lowest and highest richness indexes Chao1 were found on Anterne Lake and Villerest Lake respectively, whereas the normalized indexes based on sequences showed that Bourget Lake harboured the largest number of species Table 2.

This normalization also had an effect on the richness estimates in Godivelle Lake and Geneva Lake. These mean values mask some disparities between lakes. Thus, Anterne Lake harboured mainly reads affiliated to Fungi These data demonstrate the presence of Chlorophyta and Haptophyta in all of the lakes studied, with the exception of Anterne Lake, which is characterised by an over-representation of Fungi and an absence of Haptophyta.

This tree-based approach allows for the study of beta-diversity from phylogenies. Fungi and the total eukaryotes. This analysis permit to differentiate environments according to their taxonomic composition. For example, Lake Godivelle seems to be different from the other lakes for the Stramenopiles, while it is similar for all eukaryotes.

In a comparison of the OTUs found in this study to those present in previous studies on the small protists, only 4. Moreover, new light is shed on putative clades of small protists. Specifically, these clades include the chlorophycean group of Mamiellophyceae, represented in Figure 4 ; Foraminifera Rhizaria ; Dictyochophyceae Stramenopiles ; and Euglenida Euglenozoa.

The distribution of the OTUs among different lakes shows a main presence of clade 1 in Lake Pavin while clade 2 is mainly present in Lake Godivelle. As the interplay between evolution and ecology receives more attention in ecosystem studies [38] , there is greater interest in phylogenetic approaches for deciphering the mechanisms that govern the diversity and functioning of communities and ecosystems.

However, the phylogenetic methods that are typically applied to Sanger-sequenced SSU rRNA are computationally expensive and cannot be readily used to handle NGS datasets; therefore, pyrosequencing reads are mainly analyzed by other approaches. The method described in this study is a response to the challenge of analyzing hundreds of thousands of SSU rRNA genes in a phylogenetic framework, inferring taxonomies from sister sequences and describing clades.

This method has been implemented and tested for microorganisms with an emphasis on protists, which are not well served by bioinformatics tools dedicated to NGS data, although the early focus on bacterial and archaeal diversity has recently broadened to include eukaryotic microorganisms [39] , [40] ; thus, the database provided in PANAM includes reference sequences from protists, Bacteria and Archaea and can be used for taxonomic assignment of all microorganisms.

Our taxonomic affiliations were compared with BLAST, a tool commonly used for the identification of microorganisms especially microeukaryotes e. This method, based on ClustalW alignments and PHYML phylogenies, is a standard method for taxonomic affiliations based on phylogenetic analyses. The RDP Classifier [42] is often considered to be restricted to bacterial and archaeal taxa [26] and therefore, is not used for eukaryotic classification of SSU rRNA genes after amplicon pyrosequencing.

We used this tool for the first time for taxonomic affiliation of 18S rRNA gene amplicons generated with high-throughput pyrotag sequencing. Surprisingly, trimming the reference database to the primer region did not result in an improvement of classification for 18S rRNA gene sequences data not shown , in contrast to the results of Werner et al.

The weak performance on the truncated sequences could thus be explained by the limited number of 18S rRNA gene sequences in public databases compared with 16S rRNA gene sequences, particularly for the V9 region see the discussion below.

The comparison of the tree-based method proposed with these tools in the context of taxonomic affiliation of 18S rRNA gene amplicons shows that regardless of the method that is used, taxonomic reliability depends on the sequence length and amplicon location on the SSU rRNA gene sequence.

Our results mostly illustrate the impact of sequence length on phylogenetic methods, which appears to be the main limitation of this approach. According to Liu et al. However, by comparing different affiliation methods, they also noted that the short reads generated by pyrosequencing i. However, our analysis, similar to the one proposed by Jeraldo et al. Phylogenetic methods are generally considered superior to other approaches for taxonomic affiliation [45] as they assess relatedness between a set of sequences.

They are also considered to be difficult to automate as i their reliability greatly depends on the quality of the alignments, which need to be validated by experts in the field, and ii they use intensive, time-consuming methods for tree building. In this study, we use the curated alignments sequences provided by SILVA, which is, at least for eukaryotic sequences, the only up-to-date curated database. All high-quality and near full-length aligned sequences suitable for in-depth phylogenetic analysis were selected.

However, the guide-tree for eukaryotes provided by SILVA, in contrast to the other domains, represents only an approximate phylogeny. Tree-based approaches can implement other tools based on the tree-insertion methods like pplacer [46] as proposed by Bik et al. Similarly to STAP, this tool analyzes one sequence at a time. Thus, clades may be, at best, approximated from a frozen backbone tree, while the addition of distant taxa, as can be expected from environmental sequences, may require a re-evaluation of the phylogenetic tree [46].

In terms of processing time, we demonstrated that the tree-based method described here can process 1 M sequences in a reasonable about three hours time scale. However, while a pyrosequencing run can produce up to 1. Additionally, in diversity studies, the raw sequences are first cleaned i. Consequently, in current studies of diversity, the effective number of sequences to be affiliated is on the order of tens of thousands, which can be processed in a few hours on a personal computer.

However, some studies [32] , [48] suggest that the V6 region is not optimal for taxonomic affiliation as it overestimates richness and the number of OTUs at different cut-offs [49]. In the microeukaryotic field, the regions V2—V3 [13] , V3 [14] , [34] , V4 [22] , [23] , [39] and V9 [21] , [22] , [24] , [25] , [39] were investigated with limited in silico analysis.

Behnke et al. However, the inclusion of at least some part of the variable regions of the SSU RNA gene is necessary for the methods to retrieve sufficient signal for taxonomic affiliation. Liu et al. Interestingly, the accuracy of the taxonomic affiliation of the main phyla varied with the region analyzed, but regardless of the variable region analyzed, simulated amplicons from Viridiplantae were always difficult to affiliate reliably at the genus level.

Thus, the bias observed between variable regions [22] could be due to primers that may not anneal uniformly to all groups, but also to the bioinformatic process used for the taxonomic identification. In summary, with the exception of Viridiplantae, the V8—V9 region appears to be a good candidate for the study of protist diversity because the reliability of the taxonomic affiliation did not differ according to the phyla considered i. However, sequence databases such as GenBank contain many fewer sequences that include the V9 region than other variable regions.

In this analysis, our goal was not to explain the spatial pattern of the protist community composition PCC but to characterize the structure of these communities richness, diversity and composition by high-throughput SSU rRNA gene amplicon sequencing and sequence affiliation utilizing a tree-based method. We focused on the optimization of processing environmental data and on the description of the general picture of protists diversity obtained for these lakes.

For an in-depth analysis of this PCC from lacustrine ecosystems, we introduced environmental sequences and taxonomies in the reference database to delineate specific clades as defined in previous publications e. Phylogenetic methods provide a clear edge in describing under-studied and complex communities.

However, as with other methods, the precision of sequence mapping falls off when experimental sequences lie distant from reference SSU rRNA gene sequences [51]. This observation is particularly true for environmental sequences, for which the availability of close relatives and well-annotated sequences in reference databases is limited, as is the case for the V9 region. If the referenced trees do not include known relatives branching close to experimental reads, divergent lineages form long-branch taxa with no close reference sequences at relatively deep internal nodes.

This phenomenon results in a less precise taxonomic affiliation of these sequences; however, clades of interest could still be drawn, as very similar sequences i. Most eukaryotic species are defined on morphological differences, however, as the majority of existing microorganisms on Earth have not yet been cultured, their phenotypic traits can hardly be described.

Thus, environmental microbial species are delineated according to a sequence similarity cut-off based on comparisons of SSU rRNA gene sequences to demarcate operational taxonomic units [52]. These authors defined this similarity threshold after studying the distribution of intra- and inter-specific variations of the 18S rRNA gene in protistan communities.

However, as they pointed, this cut-off is a conservative estimator of species richness, and may mask considerable physiological diversity in some OTUs. However, this value has been defined for delineating a species from the full-length 16S rRNA gene. Finally, in a previous study, Mangot et al. Indeed, all the amplicons derived from this sequence clustered in one OTU at this cut-off. Our tree-based treatment applied to NGS sequences demonstrated that few OTUs have been previously described by the traditional cloning-sequencing CS method.

As these OTUs represent taxa present in relatively low abundance in many environments, little information is available about them. These novel OTUs were contained in a broad range of higher level taxa, including i well-established clades such as Cryptomycota, ii in phyla rarely detected by cultivation-independent sequencing e.

Thus, according to this study, the OTUs representing the most abundant sequences were found among Fungi, Alveolata, Stramenopiles, Cryptophyta and Rhizaria. More precisely, the phylogenetic affiliation allows to delineate three of the four previously defined freshwater Cryptophyta clades [6]. Within the Fungi, numerous OTUs were associated with Cryptomycota [57] or Chytridiomycota, which include both parasitic and saprotrophic organisms [58]. The presence of Chlorophyta and Haptophyta was confirmed in most of the lake environments sampled in this study.

By the CS method used for describing PCC, Chlorophyta and Haptophyta were often absent [59] , [60] or found at a very low proportion [6] , [37] , whereas these phyla represented a significant proportion of PCC when counting methods such as FISH were used [61]. Such a bias has also been highlighted in marine environments since epifluorescence microscopy reveals a dominance of phototrophic or mixotrophic cells over heterotrophic cells [62].

Another example of phyla rarely described yet detected here is the Ichthyosporea phylum, which was found only in hyper-eutrophic conditions [63]. Finally, some clades supported by high bootstrap values in our phylogenies, e. To our knowledge this is the first time that a clade closely associated to Mamiellales, as defined by Marin and Melkonian [64] , has been detected in lakes. The freshwater counterpart of this group, the Monomastigales, is rarely recovered from environmental samples and likely requires new molecular approaches that will specifically target photosynthetic organisms in the environment [64].

Freshwater Foraminifera, a group of granuloreticulosan protists largely neglected until now have already been detected by using specific primers in one study of freshwater ecosystems [67]. Among the biases commonly assigned to CS, other than the variability in the cell lysis efficiency, the rRNA gene copy number, which range from 1 to 12, [68] is certainly the most important and may result in an over-representation of heterotrophic organisms notably of the alveolate taxa [34].

However, even if these differences in copy number distort the interpretation in number of reads and OTUs for both the CS and NGS methods, the massively parallel sequencing can at least increase detection of rare lineages or organisms with low gene copy numbers thanks to the increased depth of sequencing.

We can hypothesize that this copy number could be more homogeneous at a specific lower taxonomic level for example Alveolata , and the various indexes were therefore computed for each phylum instead of considering the whole protistan community Table S3. Thus, the tree-based method presented in this work, applied to the whole spectrum of microorganisms diversity i.

The data originating from simulations and pyrosequencing were processed by a pipeline, referred to as PANAM Phylogenetic Analysis of Next-generation AMplicons that is based on publicly available programs. In addition to the phylogenetic analysis, this pipeline allows for the complete analysis of a full pyrosequencing run, including raw data processing, sequence clustering into OTUs and generating phylogenies for the taxonomic affiliation.

It is written in Perl and can be run on Linux. The pyrosequencing reads can be cleaned according to different methods commonly used in the field of molecular microbial ecology. Pyrosequencing errors can therefore be reduced by removing the primers e. Short sequences and sequences with low-quality scores are removed using PANGEA scripts [16] and only sequences with a primer match percentage above a defined threshold are selected using Fuznuc [70].

Alternatively, other quality filtering methods can be implemented; the platform does not depend upon the filtering approach described above. When several samples are analyzed, the checked sequences are split into different files depending on their bar code or tag. For the phylogenetic affiliation, a dedicated database of reference sequences, verified taxonomy and alignments was built using sequences extracted from the SSURef database of the SILVA project [72].

The sequence quality score defined by SILVA is a combination of the percentages of ambiguities, homopolymers longer than 4 bases and possible vector contaminations, and the pintail value corresponds to the probability that the rRNA sequence is chimeric. The complete database, after filtering according to the criteria above, contains , sequences Archaea: 11,; Bacteria: ,; and Eukaryota: 21, together with their taxonomy.

Each profile corresponds to the first rank beneath that of domain. As the taxonomy of Bacteria and Archaea follow standardized taxonomic paths, the monophyletic profiles of these two domains correspond to phylum, the first level occurring after the domain. For Eukaryota domain, the taxonomy does not necessarily fit this organization, and the position of the taxon in the taxonomic hierarchy does not imply rank as it is the case with Bacteria and Archaea.

Therefore, for the eukaryotic profiles, we opted for the rank position the first one after the eukaryotic domain and the monophyly, regardless to the taxonomic level. For each of the 37 phyletic groups, an outgroup containing one sequence from each other group belonging to the same domain plus 2 external sequences were added to the alignment to root the phyletic tree to be produced and to specify the relatedness of early diverging sequences from the root of the group.

To broaden the targeted diversity, the user can add specific environmental sequences to the database and the profiles. Using this dedicated database, the phylogenetic affiliation is carried out following the different stages described in the Figure 5. Next, a file containing aligned reads and sequences from the corresponding group is generated by processing a profile alignment by HMMER. As this first step does not intend to provide an exact affiliation, but rather to give a first approximation to perform a rapid and accurate phylogenetic analysis, the query sequences are sorted according to the taxonomy of their best hits, whatever their similarity score.

Several files are generated, each containing the reads and their 5 best hits, assigned to one of the 37 specific phyletic groups.

Synthetic files, which include the reference sequences and the aligned experimental reads, are generated. The trees are then parsed to generate files containing the taxonomy of the inserted sequences and files reporting the clades that could be identified from reads forming monophyletic groups. Two methods for taxonomy assessment are implemented: lowest common ancestor LCA and nearest neighbor NN.

In this last method, for each query sequence, all the nodes containing the sequence are scanned from the most recent to the deepest.

The closest neighbor is defined as the first referenced sequence starting from the lowest node. The query sequence will acquire the complete taxonomy of its nearest neighbor. For LCA [32] each node holds only the common taxonomy between all of its descendants and thus may be incomplete. Each query sequence will inherit the taxonomy of its lowest node. The final taxonomy assignment is based on the phylogeny.

The relatedness between all sequences both experimental and referenced are re-evaluated, and the similarity based assignments proposed on stage 1 are therefore revised to provide a more phylogeny-driven affiliation. Regarding the clades, their definition differs according to authors e.