# How should the popular press compare similarity of genomes?

Note this is a question from a lay reader.

I've read in the popular press that ~1-4% of the genome of non-African Homo sapiens is inherited from Neanderthals, or that Melanesians derive a similar amount from the Denisovans.

This seems like a lot, until you read (elsewhere!) that Neanderthals are 99.7% identical to humans, so most of the fraction inherited is identical in any case.

Or taking it another way, reading that two individuals will typically have 99.9% identical DNA seems pretty similar, until you realise you are only 0.3% 'different' from a Neanderthal, or 15% 'different' from your dog.

It seems to me this use of percentages is confusing, misleading, and open to abuse. When comparing species (for example) I would think that they should speak of differences in terms of base pairs that we could not otherwise have had.

How should the lay (or popular science) press best communicate similarities or differences in genomes when talking about individuals, races, or species?

• Hi, do you happen to have the reference to the news you are talking about? The way to explain the similarities relay on which metric do the articles they refer measure similarity. Also I'm not sure how this is related to bioinformatics and not biology per se or scientific communication.
– llrs
Apr 4 '18 at 6:51
• You're mixing two different concepts, inheritance (or relatedness) and sharing the same sequences. Consider that 2 'unrelated' humans are ~99.9% identical in terms of their DNA but each individual inherits (via direct decent) only (roughly) half their DNA from each parent. Apr 4 '18 at 8:33
• Llopis: I was browsing an old New Scientist article, and thought a little more about what the number meant. But there are many examples, here is one from Nat Geo genographic.nationalgeographic.com/neanderthal Apr 4 '18 at 10:03
• Apr 4 '18 at 10:18
• Llopis: I came here because there was no Anthropology or molecular biology group on Stack Exchange, and because I though that there may be less ambiguous measures of genome difference and similarity in use within the community. Apr 4 '18 at 10:58

1–4% is from an evolution point of view. 99.7% is from a sequence point of view. Because they are measuring different things, they can be compatible with each other. The correct interpretation is: 1–4% of a non-African genome is inherited from Neanderthals and the sequence of this 1–4% differs from modern human sequence by 0.3%. PS: so, for each non-African, Neanderthals contributed 1–4% * 0.3% of base pairs not present in modern human. This sounds tiny, but even a single SNP may lead to severe disease. This tiny fraction has known functional impact to diabetes and high-altitude adaption.

Others pointed out two related questions:

These questions are really about the ambiguity in the definition of "gene". Importantly, traditionally in popular science and population genetics, a gene could mean an allele. This makes differences.

For example, by common uses of "gene" in biology, the answer to "how many genes do we share with our mother" would be nearly 100% – every human has nearly the same set of genes. However, clearly this is not what the asker wants. By gene, he/she really means allele. The expected answer is 50% – we inherit 50% of alleles from our mother.

In the end, what is a "gene" exactly? I don't know. Biologists can often figure out the meaning of word "gene" in the right context, but non-biologists will sometimes have troubles with the ambiguity.

• I think we do a huge disservice to the public when we use the terms 'gene', 'variant' and 'allele' interchangeably in the popular press. I understand the desire to avoid unnecessary jargon, but surely just using the same word for a bunch of different concepts does more harm than good! Apr 4 '18 at 14:22
• @heathobrien Not only just the public but ourselves: I’m convinced that quite a few professional misunderstandings happen because of sloppy misuse of the terms, even though we should know better. At the very least it leads to cognitive overhead to figure out which meaning was intended. Apr 4 '18 at 14:50
• Yes, that's what I'm getting at. So, (unless there are still better metrics), wouldn't it be better for the press to report that up to 0.012% of our relevant (coding) genetic information is 'uniquely Neanderthal', or words to that effect? Apr 5 '18 at 0:01
• @pwray It depends on what you care about. Saying "~1–4% of the genome of non-African Homo sapiens is inherited from Neanderthals" is perfectly correct. I prefer this way, too. Apr 5 '18 at 2:24

I don’t think it’s misleading. Your question seems to be based on the mistaken assumption that the percentages you mention — 1%–4% Neanderthal origin, 99.7% similarity between individuals, etc — should somehow add up to 100%.

As you note yourself, that’s obviously not the case. There was a similar question to the effect, “how can we share 99% of our genes with chimpanzees if we only share 50% of our genes with our mother and 50% with our father?” and the answer is fundamentally the same. Namely, that these numbers refer to different things. And the DNA or gene content we share with these individuals and groups of individuals isn’t additive, or mutually exclusive. Think of it as a Venn diagram:

(Ratios obviously exaggerated, since showing a diagram where mother and father share 99% wouldn’t be very readable.)

• I believe its misleading for the lay reader, because i) percentages are usually used additively in most everyday scenarios and ii) most articles mention (eg) a '2% Neanderthal' figure without further explanation. The obvious conclusion to a lay reader is that the 2% is all foreign DNA added to (or substituting for part of) our original 'african' DNA. Apr 4 '18 at 10:25
• The question doesn't mention gene content, an explanation in terms of number (or %) of shared DNA/nucleotides is sufficient Apr 4 '18 at 14:17
• @Chris_Rands Heng’s answer notwithstanding, in this case (but not in general of course), it doesn’t matter; the same explanation works for both gene content and DNA. Apr 4 '18 at 14:46

I don't think there's any way to summarize data like this in a way that is both comprehensive in the "covering all relevant facts and concepts" meaning and in the "able to be understood" meaning. Relatedness is not a simple quantity that can be summarized in one statistic. There's what percentage of total genome is shared. There's percentage of coding genome shared (I would expect the non-coding portion of the genome to be disproportionately represented in similarities between species). There's in-group versus out-group variation (i.e. how does the variation between Neanderthals and modern humans compare to the variation within modern humans and within Neanderthals). Then there's more abstract concepts such as consanguinity index and metrics of genetic drift (i.e. species that separated more recently have less drift between them).

Ideally, a journalist would pick the statistic that is most relevant to the issue being discussed in the article, give an overview of how the statistic is calculated and what it represents, and explain why they think it is the most relevant statistic, and perhaps explain what some other statistics are and whey the journalist chose not to focus on those. But this would be a lot to expect from a journalist, and for a reader to go through.

Fact is, 87.6% of percentages are largely or completely meaningless. If someone claims that teenagers get into accidents 20% more often than the general public, does that mean that the number of accidents per teenagers is 20% more than accidents per general person? Accidents per mile driven is 20% more? Accidents per hour? Is it comparing all teenagers to the general public? Teenagers with driver's licenses to the general public with driver's licenses? Is number of accidents, or dollar amount of damages? Then you have things like Simpson's Paradox and selection bias. A percentage is supposed to be one number divided by another, and giving just that number is implicitly asserting that it summarizes the data. But generally speaking, knowing those two original numbers, or at the very least where those numbers come from, is an important part of interpreting the data.

• I’m not sure this answers the question or helps clarify. Also I don’t know why you expect “the non-coding portion of the genome to be disproportionately represented in similarities between species“- clearly coding sequences are more conserved across species Apr 4 '18 at 18:15
• I agree with Chris, that non-coding portion of the genome to be disproportionately represented in similarities between species is a very strange statement. Besides that there are animals virtually covered by genes (even a fruitfly has like half of the genome coding), there is no reason to think that non-coding parts of genomes do not play a role in speciation - relatedness - distance between two species. However +1, it's a very nice map of considerations while searching for a good metric. Apr 4 '18 at 18:45
• @Chris_Rands Random drift would affect non-coding more than coding, but evolutionary pressure would affect coding more than non-coding. So it depends on whether the speciation was driven by mere isolation, or the two populations being in different ecological niches. Apr 4 '18 at 19:00
• My initial comment was a bit harsh, sorry. But on this point, drift would impact non-coding and coding sequences the same no? Anyway, theory aside, coding sequences are often preserved over large distances (like human to yeast) but non-coding much less often Apr 4 '18 at 20:04
• @Chris_Rands Coding sequences have been selected for by evolution, so random mutations tend to result in decrease in fitness, and are thus removed. Mutations in non-coding sequences are neutral to fitness, and survive at the same rate as the original. Apr 4 '18 at 20:15