Uncertainty in Science — Mistaken Assumptions Regarding Genetic Mutation Rates May Have Significantly Overestimated the Rate of Evolutionary Change — and Four General Points about Public Confusion regarding the Scientific Process

30 October 2012

Overview — using a specific science article to make four more general points

This essay uses mistaken ideas about an evolutionary DNA “clock” to illustrate:

unwarranted assumptions’ necessary role in advancing testable science

and

the public’s misunderstanding of the varying magnitudes of scientific uncertainty.

Non-biologists will be able to follow what I say because I speak in logic’s terms, rather than on the specifics of DNA mutation.

Four basic points

These are:

(1) Making unwarranted assumptions is necessary to the scientific process, but not recognizing these assumptions overtly leads to miscommunications with the public.

(2) In regard to public understanding, science is both:

more certain than widely acknowledged — regarding well proven theories and/or data accumulation

and

more uncertain than widely perceived — regarding difficult to test hypotheses.

(3) Occasionally, the scientific process fools us by pretending to be comparatively sure of its data and inferences, under circumstances in which its actual ambiguity should arguably be specifically highlighted.

(4) In deciphering tentative pictures of reality (built from scientific hypotheses), thoughtful people should recognize the points where analytical and data pitfalls are most likely to lie as booby traps.

Disclaimer — this essay is not intended to bolster anti-scientific sentiment

The crux of the science-to-public communication problem is that science takes uncertainty and ambiguity for granted, while the public generally assumes that most things are comparatively certain. Witness the recent prison conviction of Italian natural disaster experts for allegedly misleading the public in regard to the possibility of the 2009 L’Aquila earthquake.

Most of us do not have a clue about the probabilistic nature of our own lives. Even when we do, psychological defense mechanisms usually raise barriers against actively understanding the point.

Citation — to the Science article that I use to illustrate my four points

Ann Gibbons, Turning Back the Clock: Slowing the Pace of Prehistory, Science 338(6104): 189-191 (12 October 2012)

Regarding the article that I chose to illustrate these points

Ann Gibbons’ article is about DNA mutation and its relationship to dating evolutionary events.

Her essay serves as a good example of how making assumptions is necessary to creating scientifically testable hypotheses. But it serves equally well to indicate how the process can confuse the public and even the scientists who are engaged in it.

In short, we tend to put both too much, and sometimes too little, stock in as yet unproven scientific findings. For example, disputes over the validity of climate change causation have led to foolishly polarizing disputes.

By misunderstanding where we are on the spectrum of scientific proof, we tend to make interpretative errors.

Let’s follow Ms. Gibbons’ review of the DNA mutation clock — how long ago did key human evolutionary events take place?

People are familiar with the use of ape and hominid fossils to date forks in the branches of the evolutionary road that led to Homo sapiens. But because the geological dating of the strata in which these fossils are found usually comes with wide error bars, pinning dates down is difficult.

With the advent of DNA technology, some researchers wondered whether we could find a standardized mutation clock that would allow us to extrapolate the rate of changes in human and ape DNA backward, so as to more precisely date these forks in evolutionary branching:

For the past 15 years, researchers have estimated the speed of the molecular clock by counting the mutational differences between humans and primates in matching segments of DNA, then using different species' first appearances in the fossil record to estimate how long it took those mutations to accumulate.

For example, the fossils of the oldest known orangutan ancestor are about 13 million years old, so DNA differences between humans and orangutans had about that long to accumulate. By doing similar calculations in many segments of DNA in various primates, researchers calculated an average rate of about one mutation per billion base pairs per year for humans and other apes.

Hidden assumptions in the DNA clock technique

Ms. Gibbons highlighted the obvious caveats in this too-simple clock technique:

But this method of calculating the mutation rate has drawbacks.

For starters, it assumes that the fossil dates accurately record the first appearance of a species, but that can change with a new find.

Second, there are no fossils of our closest living relatives: chimps and gorillas.

Third, the method assumes that species split at the same time as their genes diverged, but in fact, genetic separation can be millions of years earlier than species divergence.

Finally, the method assumes that mutation rates are similar across apes, although factors such as generation time—the average number of years between generations—affect the rate.

The first sign that something was amiss came in 2003, when a study tracking genes that cause hemophilia, muscular dystrophy, and other diseases in parents and children found slower mutation rates than expected.

And there are two more significant caveats

These are:

(1) What says that, even within one species, the mutation rate stays equal in all places and all times?

(2) What indicates that we actually understand the mechanisms by which DNA, epigenetic, and mitochondrial differences lead to species differentiation?

If we do not, we are probably not detecting or tracking all of the pertinent changes.

The problem of making unwarranted (statistically based) smoothing assumptions

From a medical perspective, the clock technique assumes that the mutation rate is statistically evened out across a very wide sample of patients and subjects.

In other words, if for some reason a child in Place A was subjected to high levels of cosmic radiation that caused substantially more mutations than another child living in Safer Place B, then a large across the geographic board sample of tested people would subsume Person A into the larger population.

Sounds good, until you think about it:

First, as a practical matter, none of these studies can afford to check the DNA of large numbers of people — much less the numbers that would be representative of the species at large.

Second, none of this testing goes back in time through conditions that may have been environmentally different, thus leading to higher or lower mutation rates. We are assuming that what is happening now is indicative of what was happening then. Given biological and planetary complexity this is a questionable assumption.

In light of these caveats, let’s look at what new data shows in regard to the mutation clock

Underlying the newly derived mutation clock, Ann Gibbons highlighted an Icelandic study of 78 parent and child genomes. It found that the average newborn had 36 non-inherited mutations. These mutations serve as the basis for evolution.

The Icelandic research team found that the mutation rate was only about half of the old DNA mutation calibration. Seven more studies agreed with the new and much slower rate.

If the slower clock is accurate, we would have to move critical evolutionary dates significantly backward in time:

For example, the slow clock suggests that the ancestors of modern humans and Neandertals diverged about 400,000 to 600,000 years ago, rather than 272,000 to 435,000 years ago. This fits nicely with fossils of Homo heidelbergensis, which date between 350,000 and 600,000 years ago and are thought to be ancestral to Neandertals.

Scally and Durbin also revised the dates for modern human evolution, such as pushing back the timing of a dramatic population bottleneck from 100,000 to 120,000 (rather than 60,000 to 80,000) years ago, and the emergence of modern humans out of Africa to 90,000 to 130,000 (rather than less than 70,000) years ago.

But all is not well with this new clock.

It puts the split between humans and chimpanzees too early, moving it from 4-7 million years ago to 8-10 million. The same is true for the split from orangutans, moving the earlier clock date of 13-14 million to the new clock’s 34-46 million.

“So, Pete, what’s going on?”

Think back to the earlier caveats. We have apparently made some standardization assumptions that Reality turns out not to support.

As a fix, two researchers have suggested that mutations accumulated faster during early primate evolution and more slowly in human beings. As bodies enlarge, generation times tend to slow and, therefore, the generational mutation rate does too.

But critics of this variable generation time idea do not think that accounting for its effect can remove the entirety of the error (relative to the actual fossil record).

So, another proposed fix is to recognize that the new clock probably misses some mutations. And because it does, the clock is too slow.

One research team looked at mutations in microsatellite DNA (short tandem repeats). These locations are more prone to mutations. Once their faster rate of change is incorporated into the new clock, it speeds up enough to accommodate fossil evidence of the human chimpanzee split, by centering it somewhere between 3.7 and 6.6 million years ago. (As opposed to the unmodified new clock’s 8-10 million.)

But all is still not well.

Even this modified new clock has apparent problems with appropriately incorporating the supposedly oldest members of the Hominini family (after it split from chimps).

Unwarranted assumptions — the entrance of previously unaccounted for biological complexity

I frequently harp on biological complexity and our ignorance regarding it. Too many scientific hypotheses discount already known complexity, without overtly acknowledging the simplifying assumptions that they have made. This leads to confusion, both in regard to interpreting the validity of the findings and in communicating those to the public.

For example, my medical brain never bought into the idea of a standardized rate of DNA mutation across the entirety of hominid time. Apparently, my skepticism is shared:

All these complicating factors show that researchers are still exploring what alters the mutation rate, and by how much.

Yet another twist may be the age of fathers at conception. In a landmark study of 78 trios of parents and offspring in Iceland, Stefánsson and Augustine Kong . . . found that older fathers pass on more mutations to their offspring, because mistakes are made as aging DNA is copied to produce sperm over men's lifetimes.

(Older mothers don't pass on more mutations, because eggs are all produced before birth.)

The effect is large, causing two extra mutations per year of father's age.

That means that 36-year-old dads pass on twice as many mutations as 20-year-old dads, according to a report in Nature in August. So an increase in the average age of fathers might have sped up the molecular clock.

“It's difficult to extrapolate back like this,” Stefánsson says. “You have to recognize that there are going to be flawed estimates.”

Problems with the clock idea get still more complicated:

DNA mutates at varying rates, depending on where it is in the genome. So which parts do you pick?

Nor do we know much about the effects of population sizes on mutation rates. What happens to the mutation rate, for example, when populations are isolated or significantly reduced?

Neither of these questions address even more elemental problems regarding our fundamental ignorance regarding:

(i) the mechanics of DNA’s relationship to speciation

(ii) its potential variability across environmentally wide expanses of time.

We can draw four general operating rules from the DNA clock example

(1) Science is about making unwarranted, but testable assumptions.

(2) People need to remain conscious of what they are assuming, so that they can test their assumption(s).

(3) To avoid misleading the public, scientists need to overtly acknowledge their assumptions.

(4) Maintaining scientific credibility requires that communicators explicitly consider the differences in the magnitude of evidence that supports some ideas and not others.

Rule One — science is about making unwarranted, but testable assumptions

In a sea of ignorance, we have to make simplifying assumptions, so as to generate testable hypotheses.

In the mutation clock example, we can understand that scientists had to start somewhere. It was easiest (and most testable) to begin with the two ideas that:

(a) mutations drive speciation

and

(b) a standardized mutation rate might allow us to work the investigation of evolution backward in time as a parallel to the fossil record.

By attempting to harmonize the results obtained via these premises with the actual fossil record, science learned that something had gone awry with its initial assumptions about the DNA clock. Progress.

Rule Two — know what you are assuming

Having followed these kinds of issues for decades, I am convinced that most of us forget what we are assuming in our haste to have our assumptions generate testable models.

Rule Three — to communicate effectively and without alienating the public, overtly acknowledge scientific assumptions

We might know what we’re thinking, but how is anybody else supposed to, unless we tell them?

Rule Four — alert the public to relative magnitude(s) of evidence that supports varying theories and hypotheses

This is where our sensationalized culture most often lets us down.

Too often today, hypotheses for which there is little to no scientifically persuasive evidence are trumpeted as fact. Doing so gets grant money, profit and fame. But our interpretations of Reality suffer.

The moral? — Science’s hypothesis model, which necessarily depends on pretending that something might be true (until we can prove that it is not), inherently confuses the public — and even some scientists

Knowing what we are assuming keeps us on scientific track and helps us communicate more credibly with people who do not share the ambiguity tolerant mindsets that we do.

I chose the DNA clock idea for this discussion because it has always seemed to me to be so obviously a mistaken model. Yet, at the same time, I recognized that complicating the invariant clock hypothesis would have hindered our ability to test it. Sometimes you knowingly make probable mistakes, so as to discover the limits of what you really don’t know.

This is paradoxical enough an approach to confuse lots of people.

Science is about uncertainty. And even at its most certain, it is usually about probability. Probability limits are usually wide enough to make specific predictions in individual cases difficult. Hence the science community’s upset regarding the L’Aquila earthquake convictions.

One prevalent mistake that I have noticed in the majority of science communications is that they mention their error bars less overtly than they should for the purpose of accurate communication.

So, read and listen with rational skepticism.

PeteFree.com

Uncertainty in Science — Mistaken Assumptions Regarding Genetic Mutation Rates May Have Significantly Overestimated the Rate of Evolutionary Change — and Four General Points about Public Confusion regarding the Scientific Process