# NCBI Taxonomy Virus Evolution: Step-by-Step Guide By the end of this guide you will be able to take a virus name, pull every relevant genome and metadata record out of NCBI, place those sequences on the official taxonomy tree, build a aligned dataset, reconstruct a maximum-likelihood phylogeny, estimate an evolutionary rate and time to the most recent common ancestor, run a selection-pressure scan, and deposit your own sequences back into NCBI so other researchers can reuse them. You will need a computer with internet access, roughly 20 to 50 GB of free disk space for large viral datasets, and the free command-line tools described in Step 1. No prior scripting experience is assumed, though familiarity with the terminal helps. The core idea is simple. NCBI maintains the official virus Taxonomy database, the GenBank nucleotide archive, and the dedicated NCBI Virus resource. Nearly every published viral evolutionary study in the last decade began by retrieving sequences from one of these databases. The Chikungunya virus global phylogeography study pulled all publicly available CHIKV genomes from NCBI to trace the origin and intercontinental spread of its three genotypes [1]. The classical swine fever virus analysis used 1,053 full-length E2 sequences retrieved from the NCBI database to define subgenotypes, estimate evolutionary rates, and map antigenic variability [2]. The workflow below compresses what those teams did into repeatable steps you can adapt to your own virus of interest. ## Understanding the NCBI Databases You Will Touch Before running any commands, you need a mental map of the system. This is the part of any ncbi databases guide that saves the most time later, because each database answers a different question. **NCBI Taxonomy** is the authoritative classification database at ncbi.nlm.nih.gov/taxonomy. It assigns every named organism, including viruses, a unique numeric Taxonomy ID (taxid). For example, each recognized viral species carries one taxid, and higher ranks (genus, family, order) carry their own. The tree is a consensus classification curated by taxonomy specialists, not a computed phylogeny of the sequences themselves. A recent workshop co-organized by NCBI, the CDC, BV-BRC, and NIAID examined how formal taxonomy should extend below the species level into clades, genotypes, and lineages, using systems such as Pango lineages for respiratory viruses as models [3]. Treat the Taxonomy database as the index that tells you what exists and what it is called. Treat your own phylogenetic analysis as the evidence for how those entities are related. **GenBank (Nucleotide database)** is the raw sequence archive. Search it through the Entrez system, programmatically or on the web. Records carry accession numbers, source metadata (host, country, collection date when submitter-declared), and the sequence itself. **NCBI Virus** (ncbi.nlm.nih.gov/labs/virus) is a curated front end that aggregates viral sequences from GenBank and SRA, lets you filter by taxid, host, geographic region, collection date, and sequence completeness, and exports both FASTA files and tab-delimited metadata tables. Most modern viral studies retrieve data here. The Rift Valley fever virus study in The Lancet Microbe combined new sequences from their own BioProject with publicly available datasets retrieved from the [NCBI GenBank](/blog/guides/ncbi-genbank) nucleotide database using length-filtered queries [4]. The dengue surveillance study in Thailand pulled historical DENV sequences from the NCBI Virus database alongside 92 newly sequenced isolates to track serotype and genotype turnover [5]. The RVFV molecular adaptation study retrieved 596 segment sequences (L, n=173, M, n=196, S, n=227) from 13 African countries directly from the NCBI Virus database [6]. **SRA (Sequence Read Archive)** holds raw reads. You only go here when no assembled genome exists and you plan to assemble one yourself. **PubMed and PMC** are where you find the papers that already annotated the sequences you are about to download. The West Nile virus genotype comparison in Iran and neighboring countries, for example, started from 93 sequences downloaded from the NCBI Virus database and applied metadata-based filtering down to 89 usable records [7]. That filtering step is the part beginners skip and later regret. Finally, third-party tools can sit on top of NCBI. ExTaxsI, for instance, retrieves and visualizes viral biodiversity data stored across NCBI databases, demonstrated with monkeypox virus as a case study [8]. Specialized mirrors such as IMG/VR complement NCBI for uncultivated viruses. One comparison of over 15 million viral genomes across the NCBI and IMG/VR v4 datasets found no close relative for a newly isolated phage, and showed that approximately only half of NCBI virus isolates are detectable in metagenomic datasets, which tells you the two resources genuinely overlap but do not cover each other [9]. ## Step-by-Step NCBI Taxonomy Virus Evolution Workflow ### Step 1. Install the Toolchain Open a terminal and install NCBI Entrez Direct, the official command-line interface to Entrez, plus a aligner and tree builder. On macOS or Linux: ```bash sh -c "$(curl https://ftp.ncbi.nlm.nih.gov/entrez/entrezdirect/install-edirect.sh)" export PATH=$HOME/edirect:$PATH conda install -c bioconda mafft iqtree seqkit ``` Entrez Direct gives you the `esearch`, `efetch`, and `elink` utilities. MAFFT performs [multiple sequence alignment](/blog/guides/multiple-sequence-alignment). IQ-TREE builds maximum-likelihood trees and selects the best-fit substitution model. Verify the installation: ```bash esearch -version mafft --version iqtree2 --version ``` Expected output is a version string from each program. If `esearch` returns nothing, the PATH export line from the installer output was not applied to your current shell session. Add it to your `.bashrc` or `.zshrc`. ### Step 2. Find the Taxonomy ID for Your Virus Everything downstream keys off the taxid, so resolve it first. Search the Taxonomy database: ```bash esearch -db taxonomy -query "Chikungunya virus[Rank=species]" | efetch -db taxonomy -format docsum ``` Expected output (trimmed): ``` ... Species "Chikungunya virus" taxid: 37124 ``` You can also browse interactively at the NCBI Taxonomy website. Search the virus name, confirm you selected the species entry and not a strain or isolate, and record the taxid. This matters for segmentation viruses and species complexes where a loose text search of Nucleotide returns sequences from related species you did not intend to include. For viruses with multiple segments, like Rift Valley fever virus (family Phenuiviridae, three segments), note the segment names now because you will filter for them later [4]. ### Step 3. Retrieve Sequences and Metadata Two routes exist. The NCBI Virus web interface is fastest for a one-off download. Entrez Direct is reproducible and scriptable, which matters when you need to rerun the download after new genomes appear. Web route: go to NCBI Virus, enter your taxid or virus name in the search box, apply filters on the left panel (Complete genomes only, Host, Country, Collection date), then use Download to grab both the "Genomic sequences, FASTA nucleotide" file and the "Data table (TSV)" metadata file. The TSV is essential. Every downstream analysis needs accession, host, country, and collection date joined to each sequence. Command-line route for a segmented virus, using length filters exactly as published for RVFV [4]: ```bash esearch -db nucleotide \ -query '"Rift Valley fever virus segment L"[Title] AND 6404[SLEN]' | \ efetch -db nucleotide -format fasta > rvfv_L.fasta esearch -db nucleotide \ -query '"Rift Valley fever virus segment M"[Title] AND 3885[SLEN]' | \ efetch -db nucleotide -format fasta > rvfv_M.fasta esearch -db nucleotide \ -query '"Rift Valley fever virus segment S"[Title] AND 1520:1690[SLEN]' | \ efetch -db nucleotide -format fasta > rvfv_S.fasta ``` The `[SLEN]` field restricts results to a sequence-length range, which removes partial, truncated, and badly annotated records in one stroke. Adjust the numbers for your virus after checking the expected segment length in a reference record. Pull matching metadata for every retrieved accession: ```bash efetch -db nucleotide -format gb > rvfv_L.gb ``` The GenBank flat file contains the source features. Extract host, country, and collection date from each record's `/host=`, `/country=`, and `/collection_date=` qualifiers, or download the pre-built dataflow metadata table from NCBI Virus, which ships those fields already parsed. For a single-gene study, the pattern is the same. The H5N1 evolution study retrieved 78 hemagglutinin, 62 nucleoprotein, and 61 neuraminidase gene sequences from humans, poultry, and mammals across the NCBI database and analyzed them for cross-host clade structure and positive selection [10]. The classical swine fever virus study retrieved 1,053 full-length E2 sequences [2]. Decide at this stage whether you are building a whole-genome dataset, a single-gene dataset, or a multi-segment dataset, because that choice propagates through every later step. ### Step 4. Curate the Dataset Raw NCBI downloads always contain records you cannot use. Filter on four axes: 1. **Completeness.** Drop sequences shorter than a sensible threshold for your gene or genome. The West Nile virus study filtered 93 candidate sequences down to 89 before tree building [7]. 2. **Metadata.** Sequences without collection date cannot be used in temporal (molecular clock) analyses. Sequences without a country cannot be used in phylogeography. Decide whether to exclude them or keep them for topology-only trees. 3. **Redundancy.** Identical sequences from the same outbreak inflate sampling. Deduplicate with seqkit: ```bash seqkit seq -m 5000 rvfv_L.fasta > rvfv_L_minlen.fasta seqkit seq -r -g rvfv_L_minlen.fasta | head # sanity check seqkit fx2tab -nl rvfv_L_minlen.fasta # list names and lengths ``` 4. **Labeling.** Rename every FASTA header to a scheme like `Accession|Host|Country|Year` using a spreadsheet join or a short script. Clear labels make the tree readable and make mistakes visible. Check your work: sum the sequence count in your filtered FASTA against your metadata table row count. If they differ, you have orphan records that will break downstream joins. ### Step 5. Confirm Taxonomic Placement With BLAST Before trusting any downloaded sequence's label, and before labeling your own newly sequenced isolates, run a local similarity search. NCBI BLAST compares your query against database sequences and returns ranked hits with similarity statistics. The HCV study in Shandong Province determined patient sample genotypes using NCBI BLAST against reference databases [11]. The [avian influenza](/knowledge/bacteria/avian-bacteria/avian-influenza-new-zealand-kerala-outbreaks-jessamine-county-updates) host-tropism study aligned NCBI Virus sequences with BLASTP before phylogenetic analysis [12]. For nucleotide queries against NCBI's public database without installing anything locally: ```bash blastn -query my_isolate.fasta -remote -db nt \ -outfmt "6 stitle pident length evalue bitscore" \ -max_target_seqs 20 -evalue 1e-10 ``` Expected output is a tabular list where each line shows the subject title, percent identity, aligned length, and e-value. A genuine member of your target species typically shows high percent identity (the exact value depends on how conserved your locus is, so establish it empirically from known members rather than a fixed rule) with a near-zero e-value. Hits to a different species or genus mean either a mislabeled record in your dataset or a genuinely novel virus, which happens. The phage study discussed above found its isolate had no close relative among more than 15 million viral genomes, demonstrating that low-abundance, genomically unique viruses can exist entirely outside current database coverage [9]. If your best hits are distant, report that honestly rather than forcing a taxonomic assignment. ### Step 6. Align the Sequences A multiple sequence alignment (MSA) lines up homologous positions across all sequences so that differences can be counted position by position. For whole genomes or coding sequences, MAFFT is the standard choice. The Lassa virus study aligned complete S and L segment genomes with the MUSCLE algorithm, an equally valid alternative, after retrieval from NCBI [13]. ```bash mafft rvfv_L_minlen.fasta > rvfv_L_aln.fasta ``` MAFFT auto-detects sequence number and size and picks an appropriate strategy. For a few hundred sequences of a few kilobases, this finishes in minutes. Check the alignment before proceeding: - Open it in a alignment viewer such as AliView or Aliview's command-line cousin, or scan it with `seqkit seq`. Look for sequences dramatically shorter than the rest, which usually indicates a truncated record that survived Step 4. - For coding sequences, verify the reading frame. Translate the alignment and confirm there are no internal stop codons: ```bash seqkit translate rvfv_L_aln.fasta | grep -c '\*' ``` Internal stops usually mean frame-shift errors, a wrong record, or a non-coding region inside your coordinates. ### Step 7. Trim and Choose a Substitution Model Trim poorly aligned ends and gappy columns. Then choose a nucleotide substitution model, the mathematical description of how one nucleotide changes into another over evolutionary time. Two published viral studies converged on the same general model family from independent model tests: the Lassa virus analysis selected GTR+Gamma based on a statistical model test for its fit to Lassa genome dynamics [13], and the HCV molecular epidemiology study built maximum-likelihood trees under the GTR+Gamma model [11]. GTR (general time reversible) allows unequal base frequencies and unequal exchange rates between nucleotides. Gamma adds site-to-site rate variation, meaning some positions evolve faster than others, which is almost universally true for viral genomes. Let IQ-TREE do the model testing for you rather than assuming: ```bash iqtree2 -s rvfv_L_aln.fasta -m MF -T AUTO ``` `-m MF` runs ModelFinder, which evaluates candidate models and reports the best fit by BIC (Bayesian Information Criterion). Expected output ends with a line naming the best model, often `GTR+F+G4` in IQ-TREE notation, which corresponds to the GTR+Gamma family described above. The H5N1 study, the dengue study, and the West Nile virus study each used model-informed maximum-likelihood or distance methods on NCBI-retrieved datasets, consistent with this general approach [10][5][7]. ### Step 8. Build the Maximum-Likelihood Phylogeny Maximum likelihood (ML) tree inference finds the branching pattern that best explains your alignment under the chosen model. IQ-TREE does inference and bootstrap support in one command: ```bash iqtree2 -s rvfv_L_aln.fasta -m GTR+F+G4 -B 1000 -T AUTO --prefix rvfv_L ``` Alternative: RAxML, used for the Lassa virus ML trees [13]: ```bash raxmlHPC -f a -m GTRGAMMA -p 12345 -x 12345 -N 100 \ -s rvfv_L_aln.fasta -n rvfvL ``` `-B 1000` in IQ-TREE triggers ultrafast bootstrap with 1,000 replicates, and `-N 100` in RAxML runs 100 rapid bootstraps. Bootstrap values estimate how consistently the data support each split. The West Nile virus study assessed phylogenetic grouping stability with 1,000-replication bootstrap analysis under the neighbor-joining method in MEGA 10, a simpler distance-based approach that remains useful for quick screening [7]. For publication-quality inference, ML is the prevailing standard. Expected output files include `rvfv_L.treefile` in Newick format. Visualize it in iTOL (itol.embl.de), FigTree, or ggtree in R. Color the tips by country and shape the tips by host. If sequences cluster by geography and date in a biologically plausible way, your dataset passed its first reality check. The Chikungunya analysis, for example, resolved the West African genotype originating in Nigeria and remaining within Africa, the ECSA genotype originating in Tanzania and spreading globally, and the Asian genotype originating in Thailand and reaching Oceania and the Americas [1]. For segmented viruses, build one tree per segment and compare them. Discordant segment topologies can indicate genome reassortment, and the RVFV phylodynamic study analyzed all three segments separately for exactly this reason [4]. ### Step 9. Date the Tree and Estimate the Evolutionary Rate ![Decision path checking dated tips, root-to-tip signal and effective sample size before reporting a molecular clock estimate.](/article-images/illustrations/ncbi-taxonomy-virus-evolution-clock-readiness.webp) A time-scaled phylogeny (molecular clock analysis) converts genetic distance into calendar time using the accumulation of substitutions over known sampling dates. BEAST implements Bayesian [Markov chain Monte Carlo](/blog/data-analysis/markov-chain-monte-carlo-definition-examples) (MCMC) methods for this, and is the tool behind several published viral rate estimates. The HCV Shandong study analyzed temporal and geographic evolution of subtypes 1b and 2a with BEAST MCMC and used the Bayesian skyline plot to reconstruct past population dynamics [11]. The CSFV study estimated an evolutionary rate of 1.364 × 10⁻³ substitutions per site per year (95% HPD: 1.1808 × 10⁻³ to 1.5602 × 10⁻³) and a tMRCA (time to the most recent common ancestor) of approximately 1816 (95% HPD: 1796 to 1837), and found the effective population size rising from the 1920s through the 1950s [2]. The RVFV study used molecular clock-based phylogenetic analyses to show that nucleotide substitution rates increased significantly after the virus first spread out of Africa in 2000 [14]. Workflow in brief: 1. Install BEAST and BEAUti from the official site. 2. Import your alignment into BEAUti. 3. Assign sampling dates using the tips panel, matching each sequence name to its collection date from your metadata table. Sequences without dates must be excluded here. 4. Select a [molecular clock model](/knowledge/molecular-biology/molecular-clock-model) (strict clock first, relaxed clock if a temporal-rootness test or likelihood comparison justifies it). 5. Choose a tree prior (coalescent Bayesian skyline is common for viral population dynamics). 6. Run the MCMC chain long enough that effective sample sizes for key parameters exceed 200, checking in Tracer. 7. Summarize the posterior tree set with TreeAnnotator. A practical check before committing hours of compute: run a root-to-tip regression (for example with TempEst) on your dated sequences. A positive, roughly linear relationship between root-to-tip genetic distance and sampling date indicates temporal signal. Without temporal signal, molecular clock dates are not interpretable, and you should say so in your results rather than reporting a spurious tMRCA. ### Step 10. Detect Selection Pressures Positive selection at specific codons signals adaptive change, often in response to host immunity or host jumping. The standard statistic is dN/dS, the ratio of nonsynonymous substitution rate (dN, changes that alter the encoded amino acid) to synonymous substitution rate (dS, silent changes). A dN/dS above 1 at a site indicates diversifying (positive) selection at that codon. The Datamonkey web platform (datamonkey.org) runs codon-based models on your uploaded alignment with no installation. The RVFV adaptation study identified 7,339 protein-coding mutations across 596 NCBI-retrieved sequences and ran selection-pressure analyses with codon-based models on Datamonkey, alongside phylogenetic reconstruction in IQ-TREE and host-state reconstruction to map cross-species transmission [6]. The CSFV study performed a site-by-site dN/dS examination and pinpointed residue position 71 in E2 as under positive selection [2]. The CHIKV study identified 15 positively selected sites and 10 nonconservative hydrophobicity-altering mutations across CHIKV proteins [1]. The H5N1 multi-host study found positive selection across the HA, NP, and NA genes consistent with adaptive mutations tied to host adaptation [10]. Interpret selected sites by mapping them onto known protein domains and structures. A positively selected codon inside a known receptor-binding or antibody-escape region carries far more biological weight than one in unannotated territory. The Getah virus study combined selection analysis with virus-host co-evolution analysis to characterize adaptive sites as the virus expanded its host range from horses and pigs to cattle, foxes, and detected antibodies in humans [15]. ### Step 11. Reconstruct Geographic Spread Phylogeography places ancestral locations on the time-scaled tree, turning a dendrogram into a transmission map. The RVFV Lancet Microbe study performed continuous phylogeographic reconstruction and landscape phylogeography on all three genome segments to assess dispersal patterns and environmental drivers across Africa and into the Arabian Peninsula [4]. The CHIKV study applied phylogeographic analysis to establish the origin and spread of each genotype, distinguishing the globally expansive ECSA lineage from the Africa-restricted WA lineage and the Asia-Oceania-Americas trajectory of the Asian genotype [1]. In practice this means running BEAST with a discrete or continuous trait (country or latitude-longitude) attached to each tip, then summarizing ancestral state reconstructions and visualizing them on a map. ### Step 12. Explore Taxonomic Context at Scale When your question is biodiversity rather than a single outbreak, taxonomy-level queries over NCBI become the analysis itself. ExTaxsI retrieves viral biodiversity data from NCBI databases and generates visualizations of ecological and taxonomic relationships, demonstrated on monkeypox virus data [8]. Metagenomic-scale studies work the same way at larger volume. The amphipod RNA virus survey screened an additional 152 amphipod datasets from NCBI alongside newly sampled populations, recovering 412 viral contigs including 134 novel species across twelve RNA virus orders, then analyzed alpha and beta diversity by environment and host species [16]. In silico mining of EST and transcriptome databases has repeatedly turned up novel dsRNA virus sequences in unexpected hosts, expanding known viral diversity without a wet-lab experiment [17]. Metagenomics has reshaped evolutionary virology so thoroughly that the term "metaviromics" was coined for this scale of analysis and its implications for understanding the virus world [18]. ### Step 13. Deposit Your Sequences Back The workflow ends where it started, at NCBI. Submit your assembled genomes with full metadata through BankIt or a submission tool. The Thai monkeypox clade Ib characterization submitted genome sequences to both GISAID and NCBI to support outbreak response and improve understanding of epidemiologic patterns [19]. Include host, collection date, country, isolation source, and sequencing technology in the submission. Records missing these fields cannot participate in the temporal and phylogeographic analyses that other researchers will run, which is exactly the gap you cursed during Step 4. ## Quick Reference Table of Core Tools and Parameters | Step | Tool | Database or model | Key command or option | Purpose | |--|--|--|--|--| | Taxid lookup | Entrez Direct | Taxonomy | `esearch -db taxonomy` | Resolve the official Taxonomy ID | | Sequence retrieval | Entrez Direct | Nucleotide | `[SLEN]` length filter, `efetch -format fasta` | Pull sequences reproducibly | | Sequence retrieval | NCBI Virus web | Aggregated virus data | Filters on taxid, host, date, completeness | Fast filtered downloads with metadata TSV | | Placement check | BLAST+ remote | nt | `-outfmt "6 stitle pident evalue"` | Confirm taxonomic identity | | Alignment | MAFFT or MUSCLE | Your FASTA | default auto mode | Homologous position alignment | | Model choice | IQ-TREE ModelFinder | Alignment | `-m MF` | Best-fit substitution model, often GTR+Gamma class [13][11] | | Tree inference | IQ-TREE or RAxML | Alignment | `-B 1000` bootstrap, GTRGAMMA [13] | Maximum-likelihood phylogeny | | Dating and rates | BEAST | Dated alignment | MCMC, skyline prior [11] | Evolutionary rate, tMRCA, population dynamics [2][14] | | Selection | Datamonkey (HyPhy) | Coding alignment | Site models | dN/dS, positive selection sites [2][6] | | Phylogeography | BEAST | Dated tree + locations | discrete/continuous traits | Ancestral locations and spread routes [1][4] | ## NCBI Taxonomy Virus Evolution Tips These field-tested tips compress the difference between a working analysis and a broken one. **Lock your download date.** NCBI grows continuously. Record the retrieval date in your methods and freeze your FASTA and metadata files. Every published figure should be reproducible against a named snapshot. **Use the metadata TSV from NCBI Virus, not manual curation.** Hand-copying country and date fields for hundreds of sequences produces silent errors. Parse the structured table instead, then spot-check a dozen records against their GenBank flat files. **For segmented viruses, retrieve and analyze each segment separately.** Concatenating segments without confirming that all segments are present in every record produces missing-data artifacts. The RVFV studies handled the L, M, and S segments as separate datasets with separate length filters and separate trees [4][6]. **Search beyond NCBI when the question is total viral diversity.** The comparison across NCBI and IMG/VR v4 that found a novel phage had no relative among more than 15 million genomes also showed roughly half of NCBI virus isolates are undetectable in metagenomic data [9]. Cultivated isolates and metagenomic assemblies occupy partially disjoint spaces, and both belong in a diversity survey. **Anchor lineage nomenclature to the community system.** Below the species level, viral taxonomy increasingly uses lineage schemes such as Pango nomenclature for respiratory viruses, and a multi-institution workshop formalized expectations for accurate, timely updates and non-stigmatizing names [3]. If you name a new clade, connect it to the existing scheme for that pathogen rather than inventing a parallel one. **Validate your tree against biology.** A tree where known reference strains, genotypes from prior studies, or sequences from the same outbreak do not cluster as expected signals a data problem more often than a discovery. ## Common Mistakes and Limitations Every one of these mistakes appears repeatedly in student projects and occasionally in published work. **Mixing taxonomic classification with evolutionary inference.** NCBI Taxonomy is a curated consensus tree, not a phylogeny computed from your sequences. A genus membership in the Taxonomy database does not substitute for building your own tree, and disagreements between your phylogeny and the taxonomy often indicate genuinely unresolved classification rather than an error on your side. **Ignoring sequence quality and mislabeling.** GenBank records can carry misidentified species, wrong collection years, or host metadata that simply reflects the sample's storage location. One study screened 152 datasets from NCBI and processed contigs carefully precisely because archive quality varies [16]. Cross-check suspicious records against the original publication via PubMed before they contaminate your tree. **Retrieving partial sequences without declaring it.** A tree built from full-length genomes is not comparable to one built from 300-bp amplicons. The HCV study targeted three specific regions (5' UTR, Core, NS5B) and analyzed them as defined loci [11]. State your locus and length criteria explicitly, and use `[SLEN]` filters to enforce them at download time [4]. **Running molecular clocks on datasets without temporal signal.** Garbage dates from a garbage clock are a standard failure mode. Run a root-to-tip check first. Also recognize that rate estimates are date-range dependent. A reported rate like the CSFV 1.364 × 10⁻³ substitutions per site per year is a property of that dataset, that gene, and that sampling window [2], not a species constant. **Overreading positive selection.** A site passing a dN/dS threshold is a hypothesis, not a mechanism. The CHIKV authors flagged their 15 positively selected sites as needing further functional investigation [1]. Follow the same convention. **Assuming NCBI has everything.** Cultivation-based discovery still finds viruses metagenomics misses, and roughly half of isolates are not detectable in metagenomic datasets at all [9]. Absence of a close relative in NCBI is evidence of database absence, not necessarily biological rarity. **Neglecting recombination.** RNA viruses recombine, and a single tree can misrepresent a recombinant history. The Lassa virus analysis explicitly identified recombination events and ancestral mutations alongside its tree [13]. Screen for recombination in any dataset where topology conflicts appear between genomic regions. **Treating a single-gene tree as whole-organism history.** Different genes can have different histories in viruses. The H5N1 study built inference across HA, NP, and NA separately and found distinct clade formations by gene and host [10]. Match your conclusions to the region you sequenced. **Forgetting deposition.** Sequences that stay on your hard drive cannot enter anyone else's evolutionary analysis, including the next person trying to reproduce your phylogeny. The monkeypox clade Ib team's dual submission to GISAID and NCBI is the standard to emulate [19]. One general limitation applies to this entire domain: conclusions from NCBI-sampled data inherit the sampling biases of the archive, which overrepresents outbreaks, economically important hosts, and well-funded regions. Phylogenetic statements about global viral evolution are always statements about the sequenced subset of it, and individual clinical or field cases require interpretation by a qualified professional with the actual specimens in hand. ## Frequently Asked Questions ### What is the difference between NCBI Taxonomy and NCBI Virus? NCBI Taxonomy is the classification database that assigns each virus species a Taxonomy ID and positions it in the official tree of named organisms. NCBI Virus is a sequence-focused resource that aggregates viral genomes and metadata from GenBank and SRA with filters for host, geography, date, and completeness. Use Taxonomy to identify and define what you are studying, and NCBI Virus to retrieve the actual sequences for it. ### How do I download all complete genomes for one virus species from NCBI? Search NCBI Virus by species name or taxid, apply the complete-genome filter, and download the FASTA plus the metadata TSV. For scripted retrieval, use Entrez Direct with an `esearch` query combining the species name and a `[SLEN]` length filter, then pipe to `efetch -db nucleotide -format fasta`. The RVFV workflow used exactly this pattern per segment, with length windows matched to each segment's expected size [4]. ### Does NCBI Taxonomy show evolutionary relationships between viruses? It shows classification relationships agreed on by taxonomy curators, which frequently align with evolutionary evidence but are not computed from an alignment you control. For evolutionary claims about your specific dataset, build your own maximum-likelihood or Bayesian phylogeny from the sequences and use the Taxonomy database as the labeling framework, as the viral sub-species classification workshop describes for the evolving space of clade and lineage naming below the species level [3]. ### Which alignment tool should I use for viral genome evolution studies? MAFFT and MUSCLE are both standard and both appear in published viral workflows. MUSCLE was used for complete Lassa virus S and L genomes before RAxML tree building [13]. MAFFT's auto mode handles small and large viral datasets well. The tool matters less than checking the alignment for truncated sequences and reading-frame integrity before tree building. ### How many sequences do I need for a reliable molecular clock estimate? There is no fixed number, but published studies commonly use dozens to over a thousand dated sequences, and the requirement is adequate temporal breadth as much as count. The Getah virus analysis used 82 genome-wide sequences [15], the CSFV analysis used 1,053 full-length E2 sequences [2], and the RVFV [genomic epidemiology](/knowledge/bioinformatics/genomic-epidemiology-integrating-pathogen-genomics-into-outbreak-investigations) study combined newly generated data with public GenBank records across 13 countries [4]. What matters is that sampling dates span enough of the virus's history to produce detectable temporal signal, which you verify with a root-to-tip analysis before trusting any rate. ### How do I find a virus Taxonomy ID and use it in a search? Search the virus name at the NCBI Taxonomy website, or run `esearch -db taxonomy -query "Virus name"` and read the taxid from the `efetch -db taxonomy -format docsum` output. Then use the taxid in NCBI Virus filters or as `[Organism]` with `esearch -db nucleotide` to restrict Nucleotide searches to that exact taxon rather than free-text matches. ### Should I submit my viral sequences to NCBI, GISAID, or both? Both, when the pathogen's community norms include GISAID, because the two archives serve different user bases and both support outbreak response. The Thai monkeypox clade Ib team submitted their genome sequences to GISAID and NCBI together to maximize public health utility [19]. GenBank accession numbers are what most evolutionary software pipelines and literature citations reference, so NCBI submission is the baseline requirement. ### Why does my BLAST result not match the NCBI taxonomy classification? Two common causes exist. First, mislabeled records exist in sequence archives, so a BLAST hit can reveal that a sequence filed under one species actually belongs elsewhere. Second, genuinely novel viruses have no close relative in the database, as demonstrated by a newly isolated phage with no match among more than 15 million viral genomes in NCBI and IMG/VR [9]. In either case, report the discrepancy, verify the record's original publication, and avoid forcing an assignment the similarity data do not support. ## Sources 1. [Global Transmission and Evolution of Chikungunya Virus: Origins, Adaptive Mutations, and Intercontinental Spread of the Three Genotypes.](https://pubmed.ncbi.nlm.nih.gov/41346561/) 2. [Molecular evolution and antigenic mapping of classical swine fever virus: a comprehensive analysis of E2 genomic variability and selection dynamics.](https://pubmed.ncbi.nlm.nih.gov/42115469/) 3. [Report from the BV-BRC, CDC, NCBI, and NIAID Viral Sub-Species Classification Workshop.](https://pubmed.ncbi.nlm.nih.gov/42283465/) 4. [Spatiotemporal patterns of Rift Valley fever virus in Africa: a retrospective genomic epidemiology and phylodynamic modelling study.](https://pubmed.ncbi.nlm.nih.gov/41722611/) 5. [Dengue virus genetic diversity and lineage dynamics in Thailand: implications for surveillance and vaccine implementation.](https://pubmed.ncbi.nlm.nih.gov/42135747/) 6. [Evolutionary dynamics and molecular adaptation of Rift Valley fever virus across human and non-human outbreaks in Africa.](https://pubmed.ncbi.nlm.nih.gov/42289653/) 7. [Comparison of the Genotypes of the West Nile Virus Identified in Iran and Neighboring Countries.](https://pubmed.ncbi.nlm.nih.gov/42990883/) 8. [Exploration and Retrieval of Virus-Related Molecular Data Using ExTaxsI: The Monkeypox Use Case.](https://pubmed.ncbi.nlm.nih.gov/38060123/) 9. [Identification and Genomic Analysis of a New Viral Species With Low Similarity to Existing Viruses.](https://pubmed.ncbi.nlm.nih.gov/41189414/) 10. [Comprehensive Evolutionary and Structural Analysis of the H5N1 Clade 2.4.3.4b Influenza a Virus Based on the Sequences and Data Mining of the Hemagglutinin, Nucleoprotein and Neuraminidase Genes Across Multiple Hosts.](https://pubmed.ncbi.nlm.nih.gov/41011764/) 11. [Genotype Distribution and Migration Patterns of Hepatitis C Virus in Shandong Province, China: Molecular Epidemiology and Phylogenetic Study.](https://pubmed.ncbi.nlm.nih.gov/40825211/) 12. [Avian Influenza Virus: Comparative Evolution as the Key for Predicting Host Tropism Expansion.](https://pubmed.ncbi.nlm.nih.gov/40732656/) 13. [Tracing the Evolution and Recombination Events of Neurotropic Arenaviridae Viruses: a Bioinformatics Approach.](https://pubmed.ncbi.nlm.nih.gov/41376332/) 14. [Increasing evolution, prevalence, and outbreaks for rift valley fever virus in the process of breaking geographical barriers.](https://pubmed.ncbi.nlm.nih.gov/38272089/) 15. [From transmission to adaptive evolution: genomic surveillance of Getah virus.](https://pubmed.ncbi.nlm.nih.gov/40535537/) 16. [Amphipoda and the Riboviria: viral diversity on a global scale.](https://pubmed.ncbi.nlm.nih.gov/41833796/) 17. [Discovery of novel dsRNA viral sequences by in silico cloning and implications for viral diversity, host range and evolution.](https://pubmed.ncbi.nlm.nih.gov/22848734/) 18. [Metaviromics: a tectonic shift in understanding virus evolution.](https://pubmed.ncbi.nlm.nih.gov/29433650/) 19. [Genomic characteristics of the first five monkeypox virus clade Ib cases in Thailand: August 2024-March 2025.](https://pubmed.ncbi.nlm.nih.gov/42782136/)