29

It's widely claimed that we share 50% of our DNA with bananas. Does this have actual basis or is it a myth?

Example claims The Mirror (UK), NHGRI

60

Finally, a question covering my nominal area of expertise. To answer this question meaningfully we'll need to define some concepts but first.

Yes, sort of. The statement is factually correct for reasonable interpretations.

So, on to the terms.

I'll link to a more specific stackexchange to support the definitions

Homology.

To draw an analogy: If someone said "humans share 90% of their skeleton with birds" would it be a reasonable statement? the overall structure is the same, most bones have an equivalent that's longer, shorter, thicker, thiner. The bones might have adaptions for strength or weight but stretch and squish things a little and you'll get something that looks the same.

homology

Orthology

Orthologs are genes in different species that evolved from a common ancestor. Orthologs normally retain the same function.

Paralogy

Paralogs are genes which have been duplicated. One copy may end up doing something very different to the original but retain some things in common. enter image description here

When do 2 genes count as the same?

Lets look at an example

Histone H1

Histone is found in most species and it's highly conserved. It's not identical across species but it's mostly the same.

histone

In fact a lot of the genes which we need to live are common across many species. Genes to copy DNA, genes to repair cell walls, genes to control temperature, genes to metabolize various sugars. No matter whether you're a human or a banana plant you need a lot of the same basic machinery to live.

The genes aren't perfectly identical but since they mostly have to do the same job and you'll likely die of fail to breed without them, they tend to be highly conserved, with most of the differences at less important points in the genes.

The genes get moved around, they get flipped back to front or swap chromosomes or chromosomes get merged or split, but they're generally there somewhere.

For example, here's a mouse genome colored by what sections have homologs in the human genome.

mouse genome

It's all about how strict you are with what you count as "sharing" DNA

If you count genes sequences with a single base different as not the same, then most humans wouldn't count as very similar to other humans.

If you allow for 1, 2, 3, 4 or more mutations per 100 bases while still counting something as "the same", then you can get almost any percentage you want.

That's why I said "sort of" above.

I could link to some paper where they give some number, but that wouldn't be terribly informative. I could point to 3 more which give different numbers for the same thing because it's all about where you set the cutoff points when deciding if something counts as the same.

  • 2
    This is a really good answer - it is informative, referenced, and explains well the subject. A few people will note like it since it's not really the usual format (one or two papers referencing the subject), but I still like it. +1 from me! – T. Sar Sep 22 '16 at 14:09
  • 1
    That diagram you show of the color-coded mouse/human homologs; does that type of diagram have a name? – Jason C Sep 22 '16 at 15:08
  • 3
    Oh wait, nevermind; cinteny.cchmc.org generates them and seems to refer to them as "synteny maps". They don't have bananas in their database, sadly, but they do have rice. I made this but I'm not sure what parameters they used to generate the mouse / human one. – Jason C Sep 22 '16 at 15:15
  • 1
    Yep, that sounds about right. With very distantly related species you need to limit it to exons and accept a lot more mismatches to get a reasonable degree of synteny. Checking out their human-mouse synteny diagram/map it looks like they have slightly stricter default parameters than were used to generate the diagram I included in my post. – Murphy Sep 22 '16 at 15:57
  • 2
    Won't this answer be better phrased as "humans share 50% of genes with bananas"? From the synteny map with rice, it's pretty clear that the "50% DNA" claim is false. – March Ho Sep 25 '16 at 16:57
11

The claim that we share 50% of our DNA is likely a misquote of an older claim, that we share 50% of our genes with bananas. Both claims, as best I can tell, are false. I'll address each claim in turn.

50% of our DNA?

The idea that we share 50% of our DNA is, by most obvious definitions of what that might mean, completely false, and trivially so. Per Wikipedia, the human genome is roughly 3 giga base pairs long. The banana (Musa Acuminata) genome, on the other hand, is only around one fifth of that length - 600 Mb according to ProMusa, or just 520 Mb according to the first publication of a reference genome in 2012.

Clearly, then, we cannot align slices of a single copy of the banana genome against the human genome and have them cover 50% of its length, because there aren't even enough base pairs for this to be possible!

Where, then, does this claim come from? Bioinformatician Neil Saunders tried to track down the answer and wrote up his findings in a blog post entitled 50% bananas; the oldest source he can find is this interview with Steve Jones in episode Almost Like A Whale of The Science Show on ABC radio on 12th January 2002), a full decade before the publication of the first banana genome.

Steve doesn't substantiate the claim in the interview, and it comes in the context of a wisecrack about how genomically similar organisms can still be very different from each other:

... we also share about 50% of our DNA with bananas and that doesn't make us half bananas, either from the waist up or the waist down. So there are limits in what genetics can tell us about what it means to be human ...

What's more, a little while before this interview, other sources were claiming we share 50% of our genes with bananas. Here is a source for that claim from April 2001: https://www.pbs.org/wgbh/nova/genome/deco_lander.html. Therefore, while we can't be sure what Steve's basis was, we can speculate that he simply accidentally misquoted that similar claim, in the course of cracking a joke in a radio interview.

But what about that alternative claim, then?


50% of our genes?

Murphy's answer suggests that it's reasonable to instead take the claim that we "share 50% of our DNA" with bananas to mean that 50% of humans' tens of thousands of genes have homologs in banana genes. What's more, the oldest source I can track down for any variant of the "50% banana" claim - the previously linked 2001 interview with Dr Eric Lander, in which the interviewer, science journalist Robert Krulwich, brings it up - says that we share 50% of our genes, not 50% of our DNA. So, is this variant of the claim true?

As best I can tell, no. At the very least, it's dubious and I cannot find any substantiation for it.

As previously noted, the banana genome was not sequenced until 2012, yet this claim dates back to at least 2001 - and so it was made without access to the data we now have. And Neil Saunders, in his "50% bananas" blog post, tried using the Orthologous Matrix Browser - a public tool for finding orthologs of genes - to look for orthologs of human genes in the banana genome. He reports that it only finds about 3500 human genes with such orthologs. (Specifically, 3440 - I get a slightly different number, perhaps due to changes to the OMB's algorithm.)

You can replicate Neil's result yourself by entering HUMAN and MUSAM (corresponding to Musa acuminata, the scientific name for bananas) at https://omabrowser.org/oma/genomePW/ and clicking "get pairs". The OMB returns a TSV of ortholog pairs, with the ID of a human gene on the left and its corresponding banana gene on the right. As Neil notes, if we count the number of unique human gene IDs listed, then we get the total number of human genes that OMB's algorithm thinks have at least one ortholog in the banana genome. Such counting is shown below in Python:

>>> import requests
>>> result_tsv = requests.get('https://omabrowser.org/cgi-bin/gateway.pl?f=PairwiseOrthologs&p1=HUMAN&p2=MUSAM&p3=OMA').text
>>> unique_human_gene_names = {line.split('\t')[0] for line in result_tsv.splitlines()}
>>> len(unique_human_gene_names)
3484

He notes that this,

given that there are twenty thousand-ish human protein coding genes, equates to around “17% banana”.

Of course, the OMB is just using an algorithm to attempt to find orthologs by looking at reference sequences. There may be some defensible alternative approach that yields a different result. But if so, I have not been able to find it; the only attempt I know of to fact-check the 50% claim in light of published reference genomes is Neil's, and his method, at least, clearly finds that the claim is false.

  • 2
    "a result you can trivially confirm" It is trivial to run your code. It is not trivial to confirmt hat your code does anything meaningful, or that the fetched data is meaningful. – Oddthinking Jun 16 at 17:00
  • @Oddthinking If you don't trust that particular code (or me, or the cited source) then just count the unique genes in some other way. All it's doing is counting the number of unique items in one column of a TSV. I agree that confirming the fetched data is meaningful is a trickier problem. – Mark Amery Jun 16 at 17:20
  • 1
    Counting the unique items in one column of a TSV is trivial (to a Python developer). But you haven't demonstrated that counting unique items in that column is a meaningful way to count genes, and doing so shouldn't dismissed with a hand-waved "trivial". – Oddthinking Jun 16 at 23:34
  • 1
    @Oddthinking That's a reasonable criticism. I will tweak the answer to provide more justification for the method. – Mark Amery Jun 17 at 9:50
  • @Oddthinking I've tweaked the post a little in light of your criticism. In particular, I've done two things: slightly fleshed out my explanation of the approach to ortholog counting, and slightly softened the assertion that the "50% of genes" variant of the claim is false (since all I can really say defensibly is that OMB's algorithm that attempts to identify orthologs yields a very different result; I don't know that there's no defensible methodology that would yield a different result). – Mark Amery Aug 31 at 13:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .