6
$\begingroup$

The May 24, 2022 Bloomberg opinion piece Monkeypox Isn’t Looking Like a Covid-Sized Threat; It’s still early, but contact-tracing efforts and analysis of the virus’s genome offer hope that this outbreak can be contained. includes the following:

Why now, monkeypox?

Scientists are scrambling to answer this key question. Two strains, or clades, of monkeypox are known to exist, and the one currently circulating seems to be the milder West Africa one. (For more on that, see this explainer from Therese Raphael and Sam Fazeli.)

One challenge is that researchers lack a good baseline measure on the transmissibility of this clade of the virus. Over the weekend, Maimuna Majumder, a computational epidemiologist at Harvard Medical School, provided an initial analysis of the monkeypox virus’s R0 (R-naught) — the number of people expected to be infected by a single case. These findings, which have yet to be peer reviewed and are based on very limited data, suggest an R0 of 1.15 to 1.26 — low enough to imply that the virus can be kept in check by contact tracing, vaccination and isolation of infected people.

While the present chain of monkeypox transmission differs from earlier patterns, this clade of the virus is generally mild for most infected people (albeit with a long period of isolation and recovery). And, critically, vaccines and antivirals are available to address infections.

I'd never heard the term "clade" before. I've found News Medical/Life Science's (Last Updated: Jan 11, 2022) What are Viral Clades? which which introduces the concept of the cladogram and basal clade and uses the RNA virus SARS-CoV-2 as an example is somewhat helpful but it's more of an explainer than a procedure, consistent with the site's mission to be

a tight-knit community of scientific, medical, and life sciences experts that produce and share the latest information, in a readable, understandable way.

So I checked Wikipedia's Cladogram which has a general discussion but doesn't go into great detail.

So I'd like to ask:

How are DNA virus cladograms actually calculated in practice? Is the procedure different for RNA viruses? Are these processes somewhat subjective?

By "subjective" I mean do scientists producing cladograms have to make any decisions as to what genetic differences are important for the diagram, or would several groups working independently always arrive at identical diagrams from identical data sets because they apply a standardized methodology?

$\endgroup$
6
  • $\begingroup$ I understand that there are three question marks in the title and one might at first suggest this is multiple questions, but in this particular case I don't think it's better to break this up into three separate question posts because they are so closely related and answers will likely come from the same sources and require the same expertise to write and future readers will find them equally interesting. But if a potential answer author feels strongly otherwise I am certainly willing to consider a breakup. $\endgroup$
    – uhoh
    May 23 at 20:58
  • $\begingroup$ potentially related: How can I classify the 3 clades(S, G, V) of the coronavirus that are found on GISAID? and in bioinformatics How can I classify the 3 clades(S, G, V) of the coronavirus without using protein data? As I'm not a bioinformaticist and not likely to understand answers written in bioinformatic-ese, I've chosen to post this here rather than there. $\endgroup$
    – uhoh
    May 23 at 21:05
  • 9
    $\begingroup$ For what it's worth, clades are not something specific to viruses, but all biology. A clade is a group of organisms with a common ancestor; cladistics is the study of the ancestral relationships of organisms and classification by those relationships. $\endgroup$
    – Bryan Krause
    May 23 at 22:19
  • 1
    $\begingroup$ This question is slightly different than the suggested similar questions about the S,G,V clades. This question is about phylogenetic reconstruction generally, while the others were asking more how to identify an unknown sample within an existing phylogeny. $\endgroup$ May 24 at 5:38
  • 1
    $\begingroup$ Too short for an answer, but consider that some people construct trees with BEAST2. $\endgroup$
    – Galen
    May 25 at 0:15

2 Answers 2

11
$\begingroup$

How are DNA virus cladograms actually calculated in practice?

A cladogram, or phylogenetic tree1, is constructed by comparing similarities and differences between organisms, and placing those within an evolutionary model.2 In practice, especially for viruses and microorganisms, the analysis is performed on the organisms' nucleic acid or protein sequences.3,4 For example, a simple model might assume that all nucleic acid changes are equally likely; others might treat transitions and transversions differently; still others may have a unique substitution rate for each possibility, with the overall likelihood depending on the gene under analysis, position within the gene, frequency of each nucleotide, etc. The choice of model may be directly selected by the researcher, but is often selected from an analysis that balances model fit against over-parameterization. The researcher selects the tree that optimizes some aspect of the model, for example the likelihood that the selected tree could produce the observed data under the chosen evolutionary model.5,6

Is the procedure different for RNA viruses?

No, the procedure outlined above applies to both DNA and RNA viruses, although aspects of the model parameters may differ.

Are these processes somewhat subjective?

The same model, applied to the same data, will produce the same result (assuming a thorough tree search was performed), so that aspect is not subjective. However, model choice implies assumptions about the nature of biological evolution, and researchers may disagree on the validity of those assumptions.

That being said, the methods for analyzing viral outbreaks are very well characterized and it's unlikely two competent researchers will produce greatly conflicting trees. In reference to the monkeypox example, the two clades mentioned are strongly supported, to the point where it would be innaccurate to call them subjective.

Extra

It should be noted that for any given tree, the identification of clades is not at all subjective. Every node in a phylogeny represents an ancestor of a clade. However, researchers find some clades especially relevant to the study system, so they will call them out with a name or label. For example in monkeypox the West African/Central African clades are very relevant to most discussions.

The above generally applies to any group of organisms. Viruses and other microbes may differ in that we have access to specimens that were the actual ancestors of the current generation. For example, apart from what we learn from fossils, we can only infer the characteristics of the common ancestor of chimps and humans. However, for SARS-CoV-2 we have samples of the original isolate, even if viruses with that exact sequence are no longer present in the wild. When constructing a tree with this sample we would place that specimen at the node representing the ancestor of all current SARS-CoV-2, rather than at one of the tips of the tree.

Finally the phrase "basal clade" is nonsensical. All clades, by definition, contain an ancestor and all of its descendants. There are clades where the ancestor existed earlier in time and has more descendants than others, but then, that clade is itself just a part of a larger clade. The linked article states, "a basal clade is one closest to the root common ancestor", but of course, that ancestor had more than one successful descendant, and so both descendants are equally close to the root. The article also states, "The S clade... mutated to produce the V clade," but if the term is being used properly the V clade is a part of, and a subclade within, the S clade.7 Here is a longer discussion of why "basal" is usually an inappropriate descriptor of phylogenetic placement. (However, as mentioned above, for some microbes we do actually have samples of the ancestral strains, so "basal is a bit more meaningful there.)


1A phylogenetic tree is a broader category of analysis where the lengths of the branches in the tree have meaning. A cladogram only shows topological relationships, while a phylogenetic tree shows both topology and some magnitude of change (usually either evolutionary or temporal).

2A tree can also be produced by simple similarity clustering, which is fast, but can be positively misled compared to the true phylogeny.

3Phylogenies can also be constructed from other information, such as morphological similarities/differences, as in fossils, for example. Still, some evolutionary model is used to transform the similarity matrix into the phylogeny.

4Of course, the protein sequence can be derived from the nucleic acid sequence. Which is used usually depends on the nature of the question and the breadth of the organisms included in the analysis.

5This is the maximum likelihood method, Bayesian methods are also common.

6Selection of the optimal tree is NP-hard, and there exist many methods to tractably estimate the solution.

7SARS-CoV-2 has its clades named after specific mutations that "define" (are synapomorphies for) the clade. However, this isn't strictly informative, since there's no reason a member of that clade couldn't mutate back to the previous state, or that a member of another clade couldn't pickup an identical mutation.

$\endgroup$
2
  • $\begingroup$ This is very helpful, as are the linked papers, thanks! "...the procedure...applies to both DNA and RNA viruses, although aspects of the model parameters may differ" The nature of that difference could be the basis of a new question at some point, but I'll need to give it some careful thought. $\endgroup$
    – uhoh
    May 24 at 6:08
  • 1
    $\begingroup$ Very nice answer. $\endgroup$
    – bob1
    May 24 at 8:59
6
$\begingroup$

Disclaimer: I'm not a bioinformatician, though I occasionally dabble in it, so I may well have some things incorrect. Please feel free to correct or make your own answer.

As I see it, you are asking how cladograms/dendrograms are generated for viruses.

The answer is basically that you take the genetic sequences and compare them for changes in sequence and conserved regions. The more similar the sequences are, the greater the likelihood that they are closely related and they should come out on the same branch or leaf of the tree. Conversely, the more changes in a sequence (compared to a reference) should have greater divergence and come out in a different branch of the tree.

Edited to add regarding "more similar": Similarity in sequences is based on a few parameters, but is largely based around conservation of energy. So, if you had 3 DNA sequences (AA codes underneath)

AATTCCGGG
AsnSerGly

AAATCCGGG
LysSerGly

AACTCCGGG
AsnSerGly

Compared to 1 as reference, 1 and 2 are more similar because they conserve the A/T pairing (lower energy), whereas 3 substitutes a C, which results in a C/G substitution pairing (higher energy). Of course, if you were looking at this from the possibility of translation into protein (no frame given), then 3 actually makes more sense given that it retains the coded amino-acid sequence. Substitution matrices for these sorts of things have been worked out that produce odds-ratios for a particular substitution type. The commonly used ones are PAM (Point Accepted Mutation) and BLOSUM (BLOcks SUbstitution Matrix).

The procedure to generate a cladogram is essentially the same for any organism that you care to analyze and follows these basic steps:

  • Sequence
  • Assembly
  • Alignment
  • Tree building

Now, while this all sounds good and easy, it really really isn't. The first two of those steps are more or less unambiguous, though with the advent of single-molecule sequencing these have changed a bit too in the past 10ish years. The latter two are dependent on exactly which alignment algorithm you use. There are quite a few different alignment algorithms out there, many of which are now considered out-dated or have been superseded by a newer one.

If you want to have a go at some basic alignment and tree-building, I recommend using Clustal Omega, which is a web-server where you can put your sequences (in DNA, RNA or protein) in FASTA format, and hit the button. Grab a few sequences for the mRNA/CDS (because this cuts out the introns) for a gene from different species from GenBank and have a go. It should have an alignment output and a tab on the output with a cladogram.

It also depends on exactly what sequence and what form you use for analysis. For instance, to take a recent example - The trees and all of the variant names you see for SARS-CoV-2 (e.g. these ones at nextstrain.org) are based off the genomic sequences for the spike protein only. If you used the sequence for the RNA dependent RNA polymerase (RdRp) you would get a different picture (almost no divergence because these are highly conserved, so highly you can use these to find novel viruses that infect all clades of life), and these trees would also be different to those generated if you used the whole genome. In addition, you can convert the genomic sequences into protein, and you might well get a more conservative tree out - because of the redundancy in the codons for each amino-acid.

As with much in biology it also has to do with exactly what you use as reference sequence(s) and how divergent those are from your target. For instance if you were to attempt to build a tree of all Humans (Homo sp.), you might use Chimpanzee (Pan troglodytes) as an out-group (reference), but you could also use Orangutan (Pongo pygmaeus) or Gorilla (Gorilla gorilla), or a combination of these, and you likely would get slightly different results each time. Appropriate choice of out-group is one of the hard questions in this sort of analysis.

$\endgroup$
5
  • 1
    $\begingroup$ Accidentally posted before finished... $\endgroup$
    – bob1
    May 23 at 22:55
  • 1
    $\begingroup$ @uhoh If you have three sequences 1: AATTCCGG 2: AAATCCGG and 3: AACTCCGG then 1 and 2 are more similar (conserved base pairing) than 1 and 3 or 2 and 3, though this might not hold-up if you went to the protein sequence because of the codon redundancy... Just one of the problems... $\endgroup$
    – bob1
    May 23 at 23:25
  • $\begingroup$ yep, in addition to substitutions there are also additions and deletions of single base pairs and complete codons, and those can happen in areas that vary in the amount of impact on expression rate and on the nature and activity of the resulting protein (if that's what the stretch is doing). It still seems quite hard to unambiguously define or decide what "more similar" means with all this going on, which is why I ask about the subjectivity. $\endgroup$
    – uhoh
    May 23 at 23:29
  • 2
    $\begingroup$ The similarities have been worked out over time and basically it uses an odds-ratio for the positional changes. Substitution matrices are available. The common ones are PAM and BLOSUM $\endgroup$
    – bob1
    May 23 at 23:35
  • $\begingroup$ thanks for the edit! $\endgroup$
    – uhoh
    May 24 at 6:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.