Molecular Clocks

[DE&i]

Would anyone have the time or patience to explain to me in layman's terms this new theory please.

I'll understand if no replies as it sounds very complicated but very interesting.

Heres a quote from the paper :

When fossils show a lineage is very old, but molecular clocks suggest it's very young, which date are you going to trust? The answer, according to these authors, is the fossil date. Why?

http://www.evolutionnews.org/2014/12/paper_suggests_092171.html

Regards,

Darren.

[ashcraft]

There are a number of ways of dating organisms. Fossils are "tried and true", rates of erosion and sedimentation, and radiometric methods such as carbon dating, etc. have been shown to be relatively constant across time, so when your arrive at a date for an organism/species, they tend to agree world-wide.

Enter "molecular clocks", they function by measuring the amount of mutations in a genome in a species by comparing a younger individual to one that is considerably older. The thought being that mutations occur at a constant rate, so the difference between the individuals genome represents the time that it took to manufacture that difference.

The problem is that molecular dating and classical geologic dating don't always agree. The question becomes, who is wrong. Classic methods have been cross checked by other methodologies. Examples like carbon dating of charcoal in a firepit and dating of pottery sherds by spectrometric methods from the same site agree very closely. So the problem seems to be with molecular dating. For some reason that we don't presently understand, mutation rates seem to vary across time. It is either that, or our understanding of geology is incorrect.

Molecular dating is a new technology, so stay tuned, this is a learning process.

Brent Ashcraft

[DE&i]

There are a number of ways of dating organisms. Fossils are "tried and true", rates of erosion and sedimentation, and radiometric methods such as carbon dating, etc. have been shown to be relatively constant across time, so when your arrive at a date for an organism/species, they tend to agree world-wide.

Enter "molecular clocks", they function by measuring the amount of mutations in a genome in a species by comparing a younger individual to one that is considerably older. The thought being that mutations occur at a constant rate, so the difference between the individuals genome represents the time that it took to manufacture that difference.

The problem is that molecular dating and classical geologic dating don't always agree. The question becomes, who is wrong. Classic methods have been cross checked by other methodologies. Examples like carbon dating of charcoal in a firepit and dating of pottery sherds by spectrometric methods from the same site agree very closely. So the problem seems to be with molecular dating. For some reason that we don't presently understand, mutation rates seem to vary across time. It is either that, or our understanding of geology is incorrect.

Molecular dating is a new technology, so stay tuned, this is a learning process.

Brent Ashcraft

Well firstly I asked for a question to be answered in the most simplistic form possible for me understand. And your answer fitted the bill perfectly.

Thank you so much for taking some time out for me

Regards,

Darren.

p.s. I'll certainly be watching this space .

Edited December 21, 2014 by DarrenElliot

[Auspex]

The molecular clock idea is a work in progress, based on how we think it works. The degree of uncertainty about the rates of change is still rather high, and there is no good reason to think that the rate is a fixed constant. The tried-and-true methods point to the validity of Punctuated Equilibrium, and the failure of the molecular clock method to synchronize with them is actually evidence in favor of variable rates. So far.

[FossilDAWG]

Interesting question.

First, about molecular clocks. Once two species have separated there is no gene flow between them (pretty much the biological definition of a species). After this time any mutations that appear in one species will not cross over to the other, so each species will, over time, accumulate more and more unique mutations. The basic concept is to compare mutations in the same region of the genome (so you are comparing apples to apples) across two or more species, and quantify the number of mutations that differentiate the species. If you do this with multiple species you can reconstruct phylogenies, as species that have diverged more recently will share a lot of mutations and have fewer unique ones, compared to species that diverged further in the past (and so have had more time to accumulate mutations). If you assume mutations accumulate at a constant rate, you can not only arrange the species in a phylogeny ("family tree") by sorting out who is closest to who by number of different/shared mutations, you can estimate when the species diverged. To do this you have to calibrate your clock. This can be done by using paleontological evidence to date at least a few of the branch points in your tree, and counting how many mutations have accumulated on each branch (where the branches join will be the last common ancestor). This is the most simplistic view of things.

Now for some of the problems. First I mentioned the same part of the genome has to be sequenced for each species you want to have in your analysis. In practice, this has meant sequencing specific genes, genes that will be present in all species. Commonly, especially when work of this sort was first being done, there were a small handfull of genes that people looked at, especially genes for ribosomal proteins and cyclooxygenase. Sequencing genes has certain advantages and disadvantages. First the advantage side. To start the process one has to have a fair amount of the right piece of DNA in hand, and producing this usually is done by a process called PCR (polymerase chain reaction). To carry out this reaction you have to make small synthetic pieces of DNA (called primers) to match the ends of the region you want to amplify. Genes encode proteins, and proteins typically have some parts that are highly conserved across species (usually parts associated with the function of the protein) and other regions that are more variable. It is the variable regions that can accumulate mutations, and so be informative. Mutations don't accumulate in areas that are important for protein function, as they would result in non-functional protein. So, you can design primers that target conserved regions and amplify one or more variable region in between. These primers often will work across many species, as long as the conserved regions are highly conserved across all the species in your analysis. You can only do this for genes, as non-coding regions of the genome are too variable and you have no good conserved domains to start your analysis.

The problem with working with genes is that they encode proteins that have to be functional, so they can't really accumulate mutations totally at random. Some mutations will kill the protein, such as ones that result in non-functional truncated protein. Even in the variable regions, some mutations will change the protein in ways that compromise function. So while one can get a lot of information from sequencing genes, there are some limits to how many mutations can accumulate.

One consequence of this, now well recognized but often overlooked in the early days of such work, is that the rate of accumulation of mutations can vary a lot from gene to gene. One protein may be able to tolerate more change in it's variable domains than another. This means your "clock" must be calibrated for each gene used in your analysis. Other things can also complicate things. For example, a given position (nucleotide) in the gene may, over time, undergo more than one mutation compared to the sequence that was present in the last common ancestor. A given position may even mutate back to the ancestoral state; if you consider that there are only 4 possibilities (G,C,A, or T) for the nucleotide present at each position, the chance of a reversion is fairly high (1 in 3). This means you have to look at a large number of positions in the sequence to get the overall picture, as a small number of positions could give a misleading result. These days, people try to look at several genes, both nuclear and mitochondrial, when reconstructing phylogenies, and the analysis requires complex computer programs that use sophisticated statistical approaches to calculate the most probable phylogenies.

Another issue involves calibrating your molecular clock. As I mentioned this requires making a phylogeny, then assigning dates to as many of the branching points as possible based on fossil evidence. The calibrated clock is then used to estimate the dates of other branch points, for which we do not have fossil evidence. Obviously, fossils that can be assigned to one branch or another with confidence could only come from organisms that lived some time after the actual spit occurred. Also bear in mind that these branches only become obvious over a long time, as lineages diverge. Very close to the actual point where two branches split, the differences may have been very tiny, and not easily seen in fossils. Think of the earliest animals on the path to become whales; when these first began to spend a significant amount of time in the water they would not have appeared very different at all from their close artiodactyl relatives that stayed on land. So, we might have trouble identifying fossils from close to the branch points, so out dates will always be younger than when the split actually occurred.

Add to this the fact that few molecular biologists have much training in geology. This means they might be prone to things like looking up papers on the earliest fossils of some lineage, seeing those fossils dated as (for example) Cretaceous, then looking up Cretaceous and using dates from some geological column. The Cretaceous (again, just an example) lasted as long as the entire time that has passed from the end of the Cretaceous until today. Do you use the beginning, middle, or end of the Cretaceous to calibrate your clock? A somewhat more sophisticated user might use one of the stages (Santonian, or Maastrichtian, for example) which would be better, but still involve time spans of perhaps tens of millions of years. Ideally to calibrate your clock you would nail down the age of the individual oldest specimen as much as possible, using the whole range of geological data available to date the particular part of the particular formation that yielded the specimen, but that would require a lot of geological expertise (more than your typical DNA jocky would even know exists, much less how to access it) and even then could be complicated by geological issues such as heterochronous formations (slightly different ages in different locations).

Geological dating is based on physical phenomena, such as rates of radioactive decay or magnetic reversals. There are technical issues that can complicate this, such as ability to locate layers with radioactive minerals in a well defined stratigraphic context, which makes dating of some formations more precise than others. However if datable rocks are present to be sampled, the error inherent in the actual measurements (amounts of parent and daughter radioactive elements for example) is very small, and as no-one has ever found any evidence that the rate of radioactive decay can be changed by heat, pressure, or any other force acting on it, we can obtain dates that are accurate to within a percent or two.

Considering things that can interfere with the accuracy of our clocks, we can say the following:

Molecular clocks: 1. Need a large sample of genes; 2. mutation rates vary between genes, so clocks must be calibrated for each gene or gene family; 3. mutation rates are statistical estimates, influenced by back mutations and other factors; 4. calibration of the clocks involves parsing of complex geological data, and is always dependent on fossils from after the actual split between branches.

Geological clocks: 1. availability of dateable rocks in a good stratigraphic context to apply to specimens of interest; 2. sensitivity of mass spectrometers to quantify radioactive decay products; 3. informative fossils can only come from after the actual split between branches.

Putting it all together, geological methods are based on physical observations that can be very precise. The main limitation is that we can only see that a split (between branches in a phyogeny) has already occurred, we can't see it as it is actually occurring because the species involved are not yet different enough to classified in one branch or the other. Molecular methods are statistical estimates dependent on calibration by geological methods (and so won't be more precise than those geological dates), and mutation rates can be subtly influenced by a variety of factors.

These days we are sequencing whole genomes of more and more organisms. We have sequences of genomes of a hundred or so species, not really an adequate sample, but as we get more and more of the diversity of living organisms sequences we should be able to use whole genomes instead of a few genes. Ultimately this will remove some (but not all) limitations of sequencing individual genes, so our molecular phylogenies will become more and more precise. In theory this will allow us to estimate when actual branch points occurred (as opposed to when enough differences accumulated to make fossils that can be unambiguously assigned to this or that taxon), but those estimates will always be tied to geological data.

Don

[DeepTimeIsotopes]

Interesting question.

First, about molecular clocks. Once two species have separated there is no gene flow between them (pretty much the biological definition of a species). After this time any mutations that appear in one species will not cross over to the other, so each species will, over time, accumulate more and more unique mutations. The basic concept is to compare mutations in the same region of the genome (so you are comparing apples to apples) across two or more species, and quantify the number of mutations that differentiate the species. If you do this with multiple species you can reconstruct phylogenies, as species that have diverged more recently will share a lot of mutations and have fewer unique ones, compared to species that diverged further in the past (and so have had more time to accumulate mutations). If you assume mutations accumulate at a constant rate, you can not only arrange the species in a phylogeny ("family tree") by sorting out who is closest to who by number of different/shared mutations, you can estimate when the species diverged. To do this you have to calibrate your clock. This can be done by using paleontological evidence to date at least a few of the branch points in your tree, and counting how many mutations have accumulated on each branch (where the branches join will be the last common ancestor). This is the most simplistic view of things.

Now for some of the problems. First I mentioned the same part of the genome has to be sequenced for each species you want to have in your analysis. In practice, this has meant sequencing specific genes, genes that will be present in all species. Commonly, especially when work of this sort was first being done, there were a small handfull of genes that people looked at, especially genes for ribosomal proteins and cyclooxygenase. Sequencing genes has certain advantages and disadvantages. First the advantage side. To start the process one has to have a fair amount of the right piece of DNA in hand, and producing this usually is done by a process called PCR (polymerase chain reaction). To carry out this reaction you have to make small synthetic pieces of DNA (called primers) to match the ends of the region you want to amplify. Genes encode proteins, and proteins typically have some parts that are highly conserved across species (usually parts associated with the function of the protein) and other regions that are more variable. It is the variable regions that can accumulate mutations, and so be informative. Mutations don't accumulate in areas that are important for protein function, as they would result in non-functional protein. So, you can design primers that target conserved regions and amplify one or more variable region in between. These primers often will work across many species, as long as the conserved regions are highly conserved across all the species in your analysis. You can only do this for genes, as non-coding regions of the genome are too variable and you have no good conserved domains to start your analysis.

The problem with working with genes is that they encode proteins that have to be functional, so they can't really accumulate mutations totally at random. Some mutations will kill the protein, such as ones that result in non-functional truncated protein. Even in the variable regions, some mutations will change the protein in ways that compromise function. So while one can get a lot of information from sequencing genes, there are some limits to how many mutations can accumulate.

One consequence of this, now well recognized but often overlooked in the early days of such work, is that the rate of accumulation of mutations can vary a lot from gene to gene. One protein may be able to tolerate more change in it's variable domains than another. This means your "clock" must be calibrated for each gene used in your analysis. Other things can also complicate things. For example, a given position (nucleotide) in the gene may, over time, undergo more than one mutation compared to the sequence that was present in the last common ancestor. A given position may even mutate back to the ancestoral state; if you consider that there are only 4 possibilities (G,C,A, or T) for the nucleotide present at each position, the chance of a reversion is fairly high (1 in 3). This means you have to look at a large number of positions in the sequence to get the overall picture, as a small number of positions could give a misleading result. These days, people try to look at several genes, both nuclear and mitochondrial, when reconstructing phylogenies, and the analysis requires complex computer programs that use sophisticated statistical approaches to calculate the most probable phylogenies.

Another issue involves calibrating your molecular clock. As I mentioned this requires making a phylogeny, then assigning dates to as many of the branching points as possible based on fossil evidence. The calibrated clock is then used to estimate the dates of other branch points, for which we do not have fossil evidence. Obviously, fossils that can be assigned to one branch or another with confidence could only come from organisms that lived some time after the actual spit occurred. Also bear in mind that these branches only become obvious over a long time, as lineages diverge. Very close to the actual point where two branches split, the differences may have been very tiny, and not easily seen in fossils. Think of the earliest animals on the path to become whales; when these first began to spend a significant amount of time in the water they would not have appeared very different at all from their close artiodactyl relatives that stayed on land. So, we might have trouble identifying fossils from close to the branch points, so out dates will always be younger than when the split actually occurred.

Add to this the fact that few molecular biologists have much training in geology. This means they might be prone to things like looking up papers on the earliest fossils of some lineage, seeing those fossils dated as (for example) Cretaceous, then looking up Cretaceous and using dates from some geological column. The Cretaceous (again, just an example) lasted as long as the entire time that has passed from the end of the Cretaceous until today. Do you use the beginning, middle, or end of the Cretaceous to calibrate your clock? A somewhat more sophisticated user might use one of the stages (Santonian, or Maastrichtian, for example) which would be better, but still involve time spans of perhaps tens of millions of years. Ideally to calibrate your clock you would nail down the age of the individual oldest specimen as much as possible, using the whole range of geological data available to date the particular part of the particular formation that yielded the specimen, but that would require a lot of geological expertise (more than your typical DNA jocky would even know exists, much less how to access it) and even then could be complicated by geological issues such as heterochronous formations (slightly different ages in different locations).

Geological dating is based on physical phenomena, such as rates of radioactive decay or magnetic reversals. There are technical issues that can complicate this, such as ability to locate layers with radioactive minerals in a well defined stratigraphic context, which makes dating of some formations more precise than others. However if datable rocks are present to be sampled, the error inherent in the actual measurements (amounts of parent and daughter radioactive elements for example) is very small, and as no-one has ever found any evidence that the rate of radioactive decay can be changed by heat, pressure, or any other force acting on it, we can obtain dates that are accurate to within a percent or two.

Considering things that can interfere with the accuracy of our clocks, we can say the following:

Molecular clocks: 1. Need a large sample of genes; 2. mutation rates vary between genes, so clocks must be calibrated for each gene or gene family; 3. mutation rates are statistical estimates, influenced by back mutations and other factors; 4. calibration of the clocks involves parsing of complex geological data, and is always dependent on fossils from after the actual split between branches.

Geological clocks: 1. availability of dateable rocks in a good stratigraphic context to apply to specimens of interest; 2. sensitivity of mass spectrometers to quantify radioactive decay products; 3. informative fossils can only come from after the actual split between branches.

Putting it all together, geological methods are based on physical observations that can be very precise. The main limitation is that we can only see that a split (between branches in a phyogeny) has already occurred, we can't see it as it is actually occurring because the species involved are not yet different enough to classified in one branch or the other. Molecular methods are statistical estimates dependent on calibration by geological methods (and so won't be more precise than those geological dates), and mutation rates can be subtly influenced by a variety of factors.

These days we are sequencing whole genomes of more and more organisms. We have sequences of genomes of a hundred or so species, not really an adequate sample, but as we get more and more of the diversity of living organisms sequences we should be able to use whole genomes instead of a few genes. Ultimately this will remove some (but not all) limitations of sequencing individual genes, so our molecular phylogenies will become more and more precise. In theory this will allow us to estimate when actual branch points occurred (as opposed to when enough differences accumulated to make fossils that can be unambiguously assigned to this or that taxon), but those estimates will always be tied to geological data.

Don

Now this is impressive, how long did it take you to write this?

[ashcraft]

That answer is the difference between a dabbler, and an expert.

Brent Ashcraft

[verydeadthings]

Great summary, fossildawg. I skimmed the relevant paper (http://onlinelibrary.wiley.com/doi/10.1111/nph.13114/pdf). It seemed to me that two of their main points were explaining why molecular clocks sometimes tend to give younger ages than the actual fossils. First, some of the molecular clock studies ignored fossils that, if used to calibrate the molecular clocks, might have made their molecular clock dates older. Second, and I'll just quote from the paper,, "...the convention of placing calibrations at stem nodes, unless they are explicitly resolved into a crown group, seems to cause significant directional bias. This procedure is methodologically conservative, but it forces crown nodes to be younger than the calibration fossil, whose real evolutionary position was either in the crown or along its subtending branch, not at a stem node."

I'd like to point out that the original link is to an intelligent design website, which uses a common and misleading strategy of suggesting that disagreements within the field of evolutionary biology somehow threaten the entire theory of evolution. Molecular clock dating is a relatively young scientific field, and studies like this use more established geochronological methods to refine the method and theory of molecular clock dating.