Coming from a bioinformatics background this is really, really surprising to me. I knew there was always a chance of individual cells diverging from the shared genome but we always considered that a hallmark of cancerous, abnormal growth.
Thinking that you could have a trillion unique variations of the genome significantly ups the computational complexity of simulating an organism by a frightful order of magnitude. We are so much further from understanding biological systems than we ever thought.
That's been the main lesson from the modern era of sequencing, genomics, and bioinformatics - we haven't learned nearly as much as we have unlearned.
Agreed. That's like orders of magnitude orders of magnitude.
Definitely surprising, although in one of those hindsight realizations -- c.elegans has a specific consistent function for each cell and each cell under normal development.
If 300+ cells of a worm can be that well orchestrated, of course slightly more complex creatures could. Evolution is a series of slight successes.
What if the cells of the brain have their own genome because each one of the billions of cells has a near specific specialized function.
Could cellular function be that precise? Each micron of tissue the result of specific evolutionary pressures?
I suspect that it's much like the immune system. For example, antibody diversity is "generated by DNA rearrangements during B-cell development".[0] And:[1]
> In another parallel with the immune system, cadherin-related neuronal receptors (CNRs) are diversified synaptic proteins. The CNR genes belong to protocadherin (Pcdh) gene clusters. Genomic organizations of CNR/Pcdh genes are similar to that of the Ig and TCR genes. Somatic mutations in and combinatorial gene regulation of CNR/Pcdh transcripts during neurogenesis have been reported.
So as with evolution, more-or-less random variation generates diversity in the developing nervous system. And then there's selection.
> What if the cells of the brain have their own genome because each one of the billions of cells has a near specific specialized function.
That's the opposite of my interpretation -- it goes to show that even in the face of inevitable statistical noise, biological systems are incredibly fault-tolerant and durable. And furthermore, a neuron is just a neuron. It's unlikely that the genomic differences in a genius's neurons are significant compared to the overall genetic average which gave rise to a certain structure, certain proteins being more common, etc.
An interesting hypothesis, but difficult to reconcile with the surprisingly successful history of hemispherectomy [1]. Unless, perhaps, many unique neurons still serve nearly identical functions, despite their differences? Perhaps their uniqueness is a defense against disease, rather than functional specialization?
I don't mean "functions" in the sense of classic neuroscience, like specialized regions of the brain; I am using it as a very generic term, for lack of a better one.
It's a pretty fascinating result, and it almost seems obvious in hindsight -- a statistical result of entropy which has profound implications for anti-aging research.
The biggest takeaway for me is that physically isolated genetic diseases and asymmetries which manifest during an organism's growth may not be contained whatsoever in the organism's zygotes / passed along to its offspring (not that this was an impossible result before, but I suspect there is a higher probability of genetic independence than previously thought.)
I would love to see comparisons of the standard deviations of "healthful" individuals to less healthy individuals. I would imagine that some genomes are far better at protecting themselves than others.
My own hypothesis is that maybe our memories are stored in DNA in our brain cells. This article made my hypothesis a little more likely... Also, read the article "Microsoft experiments with DNA storage: 1,000,000,000 TB in a gram" at Ars Technica. https://arstechnica.com/information-technology/2016/04/micro...
You hypothesis physically implausible - for starters, the speed at which memories are made and recalled is much, much faster than the timescale in which "writing" and "reading" of DNA happens.
While you can encode a lot of data in arbitrary arrangements of a molecule (just as you can encode a lot of data in e.g. arbitrary arrangements of magnetic polarity in a dense material) if you have the right tools to do that, our cells have different tools specialized to do different processing on DNA molecules. There's no evidence to suppose that these changes in DNA are somehow controllable by external factors as opposed to random variations, and there's no evidence to suppose that these changes in DNA are readable in any other way than the usual creation of proteins required for functioning of the cell.
I suspect that we'll find this wherever there's a process involving massive generation of somatic diversity, plus functional selection. The immune system. The nervous system. Where else might that be useful?
This doesn't necessarily change anything about computing neural nets. You need to view the genomic variation as representing the complexity of each neural unit: its connections (in and out), it's propensity to fire, and function it operates by. The genome is just the code representing this information.
Basically, each node in a neural nets is unique akin to each neuron.
> you could have a trillion unique variations of the genome
It's not really as bad as it sounds. The vast majority of those small mutations must be in the non-coding and non-regulating zone, with no effects on the genome's functionality.
As the article mentions[1], it's an interesting line of study, but let's not jump to conclusions.
[1]: "[...] each neuron may harbor hundreds of somatic mutations. Given the long life span of neurons and their central role in neural circuits and behavior, somatic mosaicism represents a potential mechanism that may contribute to neuronal diversity and the etiology of numerous neuropsychiatric disorders." - http://science.sciencemag.org/content/356/6336/eaal1641
One major trend over the past 20 years has also been the realization that "non-coding" regions tend to be anything but, their effects are subtle and still being discovered.
Maybe? It's certainly the dominant theory but I wouldn't ascribe real truth to it. There are certainly species today that operate this way but there's no way to prove they represent the origin of multicellularity. There's no reason to even believe it happened only once.
Mystics have been talking about this sort of thing for ages. Perhaps you should look into some of those assertions to develop future investigative paradigms.
One of the fundamental techniques used by most varieties of mysticism is to make proclamations vague enough that they can be plausibly associated with any fact that is discovered later on. This is how mystics trick gullible people into believing that they have hidden knowledge. The trick is that these proclamations can't be used to make useful predictions; they only fit the data in hindsight.
No two cells are ever genetically identical. Within a common hyper-mutable region, tandem repeats, the mutation rate is 10^-3 to 10^-5. There are over 10^5 tandem repeats in the genome, and therefore at least one mutation is expected for every cell division. Many other types of hyper-mutable regions exist in the human genome.
In this research they only examine one form of genetic variation, SNPs. These findings only reflect a small proportion of the somatic variation present in the body.
There is no real surprise in these results, but the data may nevertheless be useful!
(while most here understands CPU RAM RAII DRY etc., you might want to mentions that SNP means "single nucleotide polymorphism", which are mutations where just a single base pair differs.)
"A primary cause of somatic mutations has to do with errors during the DNA replication that occurs when cells divide—neural progenitor cells undergo tens of billions of cell divisions during brain development, proliferating rapidly to produce the 80 billion neurons in a mature brain."
Certainly there must be tens of billions of cell divisions to create all the neurons, but each neural progenitor cell would only divide 30-40 times, right?
I'm not surprised that there are mutations, but the number of mutations is remarkable to me, and seems like yet another evolutionary check on brain size that I hadn't considered (energy use, difficulty of birth, and difficulty of childhood being the more obvious ones).
>"Certainly there must be tens of billions of cell divisions to create all the neurons, but each neural progenitor cell would only divide 30-40 times, right?"
Also worth noting that there's a difference between mutation and cancer, although of course they're related. Just as mice and people get cancer at similar rates, so do elephants, which have many more cell divisions, but have more elaborate anticancer mechanisms (presumably because elephants lacking them died of cancer).
Well, DNA is kind of Big Data. And everyone who worked with Big Data knows that there will always be all kinds of inconsistencies and errors.
Personally, I was always skeptical that all non-sex human cells share the same DNA, it's just statistically unbelievable that billions of cells each having billions of DNA pairs would have then equal. I expected something like 1% of cells to have mutations.
Now they say that it's 100% for neurons. Exact 100% number is also pretty sketchy from statistical standpoint.
Nobody ever said that all cells share 100% of DNA in the sense you're implying. It's obvious that DNA undergoes dynamic processes, some of which will result in lasting differences. The accumulation of errors is the basic mechanism for cancer, which has been recognised for decades.
But it was previously thought that such differences were detrimental to function (and health). There were some mechanisms for modifying cells' DNA, collectively known as "epigenetics", but these were reversible, at least in principle.
This adds another dimension of complexity, which biology already had plenty of. My favourite is the double- or triple-coding DNA: because DNA is read in triplets, there are three possible reading frames. Some organisms have evolved DNA that produces entirely different, but independently useful proteins on two or even all three reading frames of the same DNA. Talk about efficiency...
Do you have any links for the last paragraph? I know that there can be two gene encoded in reverse in bacteria but have no knowledge of any frame-shifted translations?
About the second paragraph: well, it was not really thought that differences are always detrimental to health, and there is quite a number of variant prioritization tools out there that try to find those variants that are causal compared to all the silent/no-effect mutations (mostly for comparing humans but also somatic).
Look at it from the other end - if there are on average a 1000 difference between two neurons, and a difference can be encapsulated by a single bit you'd have 2^1000 variants - you'd expect to have to look at 2^500 neurons to find a duplicate. And a difference is probably more than one bit.
Now they say that it's 100% for neurons. Exact 100% number is also pretty sketchy from statistical standpoint.
What class of problem is searching for duplicates or uniqueness among a set of billions of items? If it is a tractable problem, what is the most effective algorithm?
Sometimes I wonder how so much complexity and co-dependency has evolved on earth in such little time. Most estimates I've read say that we're likely within an order of magnitude of 100 trillion generations deep from the universal common ancestor at this point. That sounds like a lot, but not really. If you took 30 four sided die, you'd have to roll them a million trillion times to have a good shot at getting any specific permutation.
How many 'die rolls' does it take to get a selective feature to emerge in an organism? If you do a google image search for 'camouflage bugs', you'll find some brain-bending examples. There's clearly a selective advantage for some of those 'configurations', but how many generations would it take for each genetic mutation required to make a lichen katydid or an orchid mantis to converge?
I don't think dice are a good comparison, as each roll is a new "state", knowing nothing about the previous. Vs. evolution which has kind of a feedback loop which influences future output.
Why do you assume that progress only happens at the interface between generations? If that were the case, human beings would be at a huge disadvantage, given that we wait ~20 years or so before reproducing.
The longer we wait before reproducing, the more intelligent the decision to reproduce will be -- which is what determines the genome of the resulting offspring.
> There's clearly a selective advantage for some of those 'configurations', but how many generations would it take for each genetic mutation required to make a lichen katydid or an orchid mantis to converge?
If you pose the question like that with the posteriori knowledge in mind, then yes, these configurations are highly improbable. I think another question that could be asked is: How many generations would it take for some mutations to produce some camouflage effect in some of the millions of existing species?
Surely some mutations that produce camouflage effects will happen.
Some changes can happen really quickly too. If you look at breeders of fancy pigeons or fish, they can do incredible things in a relatively small number of generations. Of course, breeders are generally much more selective than nature, but it gives you a lower bound on what is possible.
Wait, the headline says no two alike, while the body text seems to go no further than that every cell is potentially different. What is it? If the somatic mutation rate is >1/neuron, that's a real surprise to me, but if it's "there's plenty of mosaicism and it makes a real difference", then not.
the more closely you look, the more obvious it becomes that this is the rule rather than the exception. I would be somewhat surprised if children have functional mutations rampant between neurons, and I suspect that some fraction of this is artifactual. But I have no doubt (does anyone?) that some degree of somatic mosaicism is the rule. About the only cells that tend to hang around much longer than neurons are blood stem cells, and as soon as you look closely at those, it's all but unavoidable.
God damn it, that Science paper is piss weak even for a Science paper. It's yet another consortium advertisement. Meanwhile mapping of somatic mutations in blood progenitors has been happening for a decade.
This exactly. Somatic variants are one of the most common things around.
I always wonder why HN will almost never upvote a general informatics research article instead of a good review and with other fields of research it is the other way around where a review would be much better suited.
Isn't this old news? The reason is vaguely similar to the DNA modification that happens in the immune system: recognition of self. In this case, the goal is to avoid loopback connections. Nerve cells that touch themselves are bad. By DNA modification, the cells get different surface protein and are thus able to avoid connecting to themselves.
Readit News
Thinking that you could have a trillion unique variations of the genome significantly ups the computational complexity of simulating an organism by a frightful order of magnitude. We are so much further from understanding biological systems than we ever thought.
That's been the main lesson from the modern era of sequencing, genomics, and bioinformatics - we haven't learned nearly as much as we have unlearned.
Definitely surprising, although in one of those hindsight realizations -- c.elegans has a specific consistent function for each cell and each cell under normal development.
If 300+ cells of a worm can be that well orchestrated, of course slightly more complex creatures could. Evolution is a series of slight successes.
What if the cells of the brain have their own genome because each one of the billions of cells has a near specific specialized function.
Could cellular function be that precise? Each micron of tissue the result of specific evolutionary pressures?
> In another parallel with the immune system, cadherin-related neuronal receptors (CNRs) are diversified synaptic proteins. The CNR genes belong to protocadherin (Pcdh) gene clusters. Genomic organizations of CNR/Pcdh genes are similar to that of the Ig and TCR genes. Somatic mutations in and combinatorial gene regulation of CNR/Pcdh transcripts during neurogenesis have been reported.
So as with evolution, more-or-less random variation generates diversity in the developing nervous system. And then there's selection.
0) https://www.ncbi.nlm.nih.gov/books/NBK27140/
1) https://www.ncbi.nlm.nih.gov/pubmed/12558794
That's the opposite of my interpretation -- it goes to show that even in the face of inevitable statistical noise, biological systems are incredibly fault-tolerant and durable. And furthermore, a neuron is just a neuron. It's unlikely that the genomic differences in a genius's neurons are significant compared to the overall genetic average which gave rise to a certain structure, certain proteins being more common, etc.
I don't mean "functions" in the sense of classic neuroscience, like specialized regions of the brain; I am using it as a very generic term, for lack of a better one.
[1] https://en.wikipedia.org/wiki/Hemispherectomy
If that's what the data says.
It's probably an emergent quality from a set of simple rules being iterated. You can do a lot with a very very little if you're procedural about it.
z_(n+1)=z_n^2+C
The biggest takeaway for me is that physically isolated genetic diseases and asymmetries which manifest during an organism's growth may not be contained whatsoever in the organism's zygotes / passed along to its offspring (not that this was an impossible result before, but I suspect there is a higher probability of genetic independence than previously thought.)
I would love to see comparisons of the standard deviations of "healthful" individuals to less healthy individuals. I would imagine that some genomes are far better at protecting themselves than others.
While you can encode a lot of data in arbitrary arrangements of a molecule (just as you can encode a lot of data in e.g. arbitrary arrangements of magnetic polarity in a dense material) if you have the right tools to do that, our cells have different tools specialized to do different processing on DNA molecules. There's no evidence to suppose that these changes in DNA are somehow controllable by external factors as opposed to random variations, and there's no evidence to suppose that these changes in DNA are readable in any other way than the usual creation of proteins required for functioning of the cell.
Mine doesn't.
Basically, each node in a neural nets is unique akin to each neuron.
It's not really as bad as it sounds. The vast majority of those small mutations must be in the non-coding and non-regulating zone, with no effects on the genome's functionality.
As the article mentions[1], it's an interesting line of study, but let's not jump to conclusions.
[1]: "[...] each neuron may harbor hundreds of somatic mutations. Given the long life span of neurons and their central role in neural circuits and behavior, somatic mosaicism represents a potential mechanism that may contribute to neuronal diversity and the etiology of numerous neuropsychiatric disorders." - http://science.sciencemag.org/content/356/6336/eaal1641
Dead Comment
In this research they only examine one form of genetic variation, SNPs. These findings only reflect a small proportion of the somatic variation present in the body.
There is no real surprise in these results, but the data may nevertheless be useful!
Certainly there must be tens of billions of cell divisions to create all the neurons, but each neural progenitor cell would only divide 30-40 times, right?
I'm not surprised that there are mutations, but the number of mutations is remarkable to me, and seems like yet another evolutionary check on brain size that I hadn't considered (energy use, difficulty of birth, and difficulty of childhood being the more obvious ones).
Surprisingly to most, this is not a mainstream opinion: https://www.ncbi.nlm.nih.gov/pubmed/25459141
I would guess the main reason is that it leads to major problems with the current model of cancer.
Also worth noting that there's a difference between mutation and cancer, although of course they're related. Just as mice and people get cancer at similar rates, so do elephants, which have many more cell divisions, but have more elaborate anticancer mechanisms (presumably because elephants lacking them died of cancer).
Personally, I was always skeptical that all non-sex human cells share the same DNA, it's just statistically unbelievable that billions of cells each having billions of DNA pairs would have then equal. I expected something like 1% of cells to have mutations.
Now they say that it's 100% for neurons. Exact 100% number is also pretty sketchy from statistical standpoint.
But it was previously thought that such differences were detrimental to function (and health). There were some mechanisms for modifying cells' DNA, collectively known as "epigenetics", but these were reversible, at least in principle.
This adds another dimension of complexity, which biology already had plenty of. My favourite is the double- or triple-coding DNA: because DNA is read in triplets, there are three possible reading frames. Some organisms have evolved DNA that produces entirely different, but independently useful proteins on two or even all three reading frames of the same DNA. Talk about efficiency...
About the second paragraph: well, it was not really thought that differences are always detrimental to health, and there is quite a number of variant prioritization tools out there that try to find those variants that are causal compared to all the silent/no-effect mutations (mostly for comparing humans but also somatic).
What class of problem is searching for duplicates or uniqueness among a set of billions of items? If it is a tractable problem, what is the most effective algorithm?
How many 'die rolls' does it take to get a selective feature to emerge in an organism? If you do a google image search for 'camouflage bugs', you'll find some brain-bending examples. There's clearly a selective advantage for some of those 'configurations', but how many generations would it take for each genetic mutation required to make a lichen katydid or an orchid mantis to converge?
The longer we wait before reproducing, the more intelligent the decision to reproduce will be -- which is what determines the genome of the resulting offspring.
If you pose the question like that with the posteriori knowledge in mind, then yes, these configurations are highly improbable. I think another question that could be asked is: How many generations would it take for some mutations to produce some camouflage effect in some of the millions of existing species? Surely some mutations that produce camouflage effects will happen.
the more closely you look, the more obvious it becomes that this is the rule rather than the exception. I would be somewhat surprised if children have functional mutations rampant between neurons, and I suspect that some fraction of this is artifactual. But I have no doubt (does anyone?) that some degree of somatic mosaicism is the rule. About the only cells that tend to hang around much longer than neurons are blood stem cells, and as soon as you look closely at those, it's all but unavoidable.
I always wonder why HN will almost never upvote a general informatics research article instead of a good review and with other fields of research it is the other way around where a review would be much better suited.