I looked over the paper pretty quickly, but on a quick read it looks really nice. The data are really impressive, and they seemed to have pulled out really interesting trends for how transcription factors are affecting coding sequence. Furthermore, they map how variants found across the various cell lines they used (81!) to show that these variants actually are causal for transcription factor binding changes. For those saying this is not a big deal, or is obvious; it is, and is not (respectively).
It's been pretty clear for some time that codon choice is not random (hence codon optimization is useful when taking a gene from one organism to another), and over the last few years it's also been clear that codon bias can be evolutionarily constrained. For a while, it mostly thought to be based on tRNA levels (basically anticodons). However, it's been increasingly clear that there are other constraints to protein coding sequences than just the decoded protein sequence (aka the genetic code).
For example, we published a paper a few weeks back [1] showing that in bacteria (and probably higher organisms), the N-terminus of genes has a lot of rare codons and this is due to other constraints such as relieving mRNA structure to allow better translation of proteins. I think in the coming years, we will find that other regulatory elements also shape this code, including sequences that control splicing, small RNAs, mRNA degradation and transport, et cetera
Anyways, it's a pretty fun time in biology. The tools we have now make studies that were ridiculously impossible just a few years ago, a reality for an individual lab. I can't wait to see what the next few years bring.
"over the last few years it's also been clear that codon bias can be evolutionarily constrained. For a while, it mostly thought to be based on tRNA levels (basically anticodons). However, it's been increasingly clear that there are other constraints to protein coding sequences than just the decoded protein sequence (aka the genetic code)."
This is a much older question than the last few years -- it's one of those "evergreen" questions that's been around for almost as long as we've been able to do statistics on codon prevalence. You'll see hand-wavy explanations for differences in codon usage in really old textbooks. The tRNA one is a particularly old chestnut -- I think I read about that one in my first undergrad course on molecular biology.
Agreed. It's an interesting trend. I also just skimmed over it; however, I wonder how significant is their initial TF binding occupancy data. There are a bunch of papers recently that looked at binding sites and found that they don't really correlate well with transcription. The cell is just a stochastic bag of molecules. TFs bind whereever they can. I wonder if they are just seeing noise.
Fig 1E seems promising, as it shows that the bound bases tend to be more conserved. That said, you are right that as far I could tell there were no direct measurements of transcriptional effects. Either way, it was a good read, and brought up a bunch of interesting questions to figure out next.
Is it as comprehensive a map of DNA-encoded transcription factors as can be had? I ask because I can understand the way DNA codes for proteins–as a software-head–but the article definitely gets me with this "second language" and I want to think more about how the overlapping meanings might have developed (and how that might be apply to digital code).
Other than "start," "stop," "intron," "exon," and "suppress" (aka disable), what do the transcription factors do in terms of actual cell proteins?
I admit that a really detailed description of cell protein processes will be beyond my depth. :)
In case you were unaware, another really amazing example of overlapping meainings is in "normal" DNA code, where genes can overlap themselves.
Because a stop codon would kill transcription, these are either on alternate reading frames (i.e. the codons begin on a different one of the three possibilities) or on the antisense side of the gene (the opposite base pairs from those corresponding to the original gene).
It blows my mind that such an incredible amount of stuff could go wrong if there is just a single mutation, and that such an incredibly delicate system arose through evolution and survived billions of years of natural selection.
I know next to nothing about this field, but I just sort of brute-forced my way through the OP paper with the help of wikipedia. It's a painful process. :-)
But it has piqued my curiosity about the field. Are there any good introductory books/websites for the curious layman to get a grounding?
Depends on what level. Wikipedia is actually quite good for this sort of thing. The classic book as a basic overview of molecular biology in general is the Molecular Biology of the Cell [1]. A really cool website to look at on cutting edge studies such as these is the ENCODE consortium's publications in Nature [2]. They simultaneously published 30 papers last year, and this is a really cool meta page that explains the ENCODE project and what they found.
Robert Sapolsky[0] has his Stanford course on human behavioral biology on youtube, almost in its entirety. The whole course is very accessible and much broader than the topic here, but if you wanna go straight for the molecular genetics section, it's parts 4 & 5. Amazing lecturer.
The Machinery of Life [1] is a fantastic book, I was one of those curious laymen and boy did it saturate my thirst for learning about microbiology. I think it's close to the time I read it again.
I'll try to give you a description of what transcription factors do and what the importance of Stamatoyannopoulos finding is. I hope it doesn't sound too weird.
Just noticed, how long my "brief" description is, so here is a
TL;DR:
Comparing cells to computer programs: A program figures out by itself which functions to call based on the byte sequence of the function in memory and changes these preferences based on the environment its running in.
As you know, the DNA in our cells encodes the blueprint for all proteins that exist in the human body. I'm trying to compare this to string encoding, but might be completely wrong, because string encoding is more difficult to me than DNA encoding ;)
The DNA is our hard drive. This hard drive holds the information for all proteins.
A protein will be a string. Our alphabet is made out of 21 characters, they are called amino acids.
Each character (amino acid) is encoded as a 3bit sequence on the DNA, the so called codon. Opposite to the binary system, we have a base4 system, meaning each bit can have 4 different states (A,T,C,G), the nucleic acids.
So, given a byte sequence from the hard drive (gene on the DNA), we always look at codons (three bits) and can then translate those codons into amino acids.
ATG translates to Methionine, GGA translates to Glycine etc....
Having 4^3 (64) different codons but only 21 amino acids, some amino acids can be represented by several different codons. On top of that, some codons have a special meaning, such as "\n". Those special codons are called "stop codon", because it marks the end of a protein and the program stops decoding from this point on. In addition, there is also a "start codon" that marks the begin of a string.
When a cell wants to make a certain protein, it copies the byte-sequence from the DNA and translates it into the actual amino acid sequence and as such forms proteins (prints a string in our example).
The problem we're now facing is that every cell in your body (with a few exceptions, of course) has the exact same DNA sequence. But still, nature is able to somehow make different celltypes, such as a skin cell, blood cell, neuron, you get the point.
The way how cells "know" which proteins to make is that the hard drive doesn't only hold encoded strings but also some byte sequences in between that don't encode any strings. These sequences are called regulatory sequences. These sequences attract so called transcription factors.
Transcription factors are proteins that you could understand as a pointer. They will point to some genes, but not to others and thus instruct the cell which byte sequences should be translated into strings.
These regulatory sequences can not only attract transcription factors, but also suppressive factors. Those are proteins that prevent transcription factors from binding and thus ensure that certain proteins are not made.
There are hundreds of different transcription factors and repressors encoded on the DNA, but only a handful exist as protein in each given cell and they define the difference between celltypes. They come and go as the cell's environment changes as well.
So, going back to our string analogy, say you have saved thousands of strings on your harddrive but only want to show a subset of those to the user (eg language selection). You start your program by passing some information about which strings to choose. The same happens here as well, only that the information which strings to chose is somehow intrinsic but still different between celltypes. And on top of that you not only have distinct languages, but the languages overlap and language A uses string 1,2,3; language B uses string 2,5,6; and language C uses string 1,2,6 etc.
Because of this complexity it is very difficult to understand how the cell "knows" which proteins are right.
Stamatoyannopoulos now for the first time proved that even though two different codons encode the same amino acids they might attract different transcription factors.
To get an understanding of the importance of this finding, imagine having a character encoding, such as DNA, where two different byte sequences encode the exact same string. Now imagine having a program that automatically chooses which strings to display based on which byte sequences encode certain characters.
And now, image that the DNA doesn't only encode strings, but also functions and closures, pointers, variables etc. So each time a part of the DNA is translated, the state of the program changes. Nature is vey complex and we're just at the beginning of understanding the code.
We have methods where we can out-comment some part of the DNA and see how the program changes (mutations). In addition, we recently learned how to see which part of the DNA is being used in certain cells. And in this paper, they refined this method by actually looking at all pointers in a cell and figuring out which part of the DNA they pointed at. And they did this for 81 different cell types to find networks that act together in certain settings.
I may be misinterpreting what you said but it seems as though you consider 'start' 'stop' 'intron' 'exon' 'suppress' as properties of transcription factors. If this is not the case I apologize but to clear any misconception those are all properties of the DNA sequence, transcription factors BIND to DNA sequences but are not part of the gene itself.[0]
I'm also not quite sure what is meant by suppress.
Really all transcription factors do is up-regulate or down-regulate genes, i.e. make more proteins from this gene or make less. However simple that seems this is actually an extremely important function. Other than enabling cells to respond to say, heat, stress, or tension by making proteins that help them cope, gene regulation levels are important in differentiation. This is the process by which cells 'pick' a destiny, i.e. brain cells, kidney cells, etc.
Generally, you can think of every cell in your body (with some limited exceptions) as containing all genes in your body, so how does the cell know what type of cell it should be? One way is by transcription factors, which will tell the cell make more of this kidney-related protein, or other proteins that help determine its fate.
They have also been implicated in memory formation and almost everything else you can think of. But for the actual process of what transcription factors do they just regulate how much protein should be made from this gene.
It's a university press release.[1] I'm sending the link to the published paper from Science kindly shared in an earlier comment here to some geneticists I know locally, who have just been in a seminar together this afternoon, I'm pretty sure. The editors of Science evidently agreed that there is something interesting here, but they have been wrong before.[2]
AFTER EDIT: I heard back from one of my local geneticist friends, a mathematician turned psychologist by higher education who largely does statistical analysis as part of a team of researchers on behavior genetics. He writes, from the perspective of behavior genetics research, "That is fascinating, but if duons are also tagged by SNPs, and especially if they are in the exomic DNA, we've already been studying them and finding very little. In other words, this is huge for molecular genetics and physiology, but I'm not so sure it changes what we do in genotype-phenotype association research." So I take that to say that this could be quite a big deal for molecular genetics and physiology, if this finding is confirmed in follow-up research.
TL;DR for biologists: synonymous codon variation is heavily constrained by the need to maintain transcription factor binding motifs.
While I found the article very interesting as a new postgraduate student, it seems like such a straightforward deduction that I find it difficult to believe no one has ever put it into words until now.
After talking to someone who knows quite a bit about this stuff, apparently most people in the genomics/bioinformatics community kinda already knew about this, just that the researchers who wrote the paper were the first one to officially categorize them.
My first thought on seeing the headline was, "What? Only 2?" Off the top of my head I can think of:
1. Amino acid coding
2. Transcription factor binding and transcriptional regulation
3. Post-transcriptional regulation and RNA degradation
4. Intron/Exon Splice sites
5. Chromosome structure and methylation
6. Origins of replication
All of these have been shown to be controlled at least in part by the DNA's sequence. This story seems to be about a (very interesting) new wrinkle in (2) above.
I agree. I believe I recall my biologist friend saying at one point that there are up to three such "layers" of genetic code, depending on where you start reading the sequence. Am I understanding that correctly?
Well, yeah, if what he's talking about are nucleotide bases, it takes three of them to make a codon. So if you have a sequence ABCABC, then ABC might be a codon (which codes for an amino acid, like a "bead" in a necklace that makes up a protein), BCA might be a codon, and CAB might be a codon. Then at the same time you could go backwards. So really there's six potential different codons that a single basepair might be involved in. But that's probably not what this paper is talking about.
You're correct in that this is not what the paper is talking about. However, there is a third layer (at least) beyond the gene-coding, most-studied area, and the article submitted; epigenetics studies how gene expression is modified by histone modification and DNA methylation, basically different ways of switching gene expression on/off or changing the amount of protein made from a certain gene.
My understanding is that GMO crops have snippets taken from other species' DNA. Those sequences would be internally consistent, and only the patch between the native and foreign sequence would be affected.
Actually I didn't know exonic TF binding was even significant enough to put synonymous codon variation under constraint. I've always assumed it was some sort of metabolic constraint, as in more efficient production of certain tRNAs putting a constraint on codon usage.
Funny how discoveries like this don't surprise me at all anymore. After embracing concepts such as fractal geometry, recursion and the demand for efficiency in nature, it makes sense that in nature everything has evolved in a way in which every layer is a part of the next layer above it (quarks -> electrons, protons -> elements -> compounds -> amino acids -> proteins -> information -> abstract information -> ..., or binary -> hexadecimal -> assembly -> low-level code -> high-level code -> abstractions -> ...), and everything has value in more than one way. DNA, proteins, muscles, neural clusters, even programming are all based around this principle. If you were building two programs that were identical in many ways why wouldn't you abstract out the commonalities, instead of building them twice?
The brain is no exception. I think musicians, for example, are better at math because at the heart of music and at the heart of math, many of the same brain circuits are involved. It is more efficient to have one copy of these circuits and just apply them to both music and math, rather than having two copies of nearly identical circuits. Proteins have more than one use in our bodies, because if each only had only one use we would need an inefficient number of them, possibly more than what exists. Same for neurotransmitters. When nature is limited by constraints it usually finds a way to fold into a higher dimension around that constraint (such as the neocortex wrinkling to increase surface area, or DNA containing information on more than a single level, or even how grass grows over a fallen log instead of "choosing" to just grow somewhere else, or if you are into String Theory, how the universe/multiverse has folded into 11 dimensions, possibly because that was the most efficient way for our universe to exist.)
It reminds me of a river when it is initially forming down the side of a mountain. The water takes the most efficient path at any given instance from the top of the mountain to the bottom. The process is like a greedy algorithm. The water cannot foresee where it will end up, it just flows. The water won't always find the most efficient solution, but give it enough time and will find an efficient-enough solution.
This is sort of a misunderstanding of evolution. It's not building to the most efficient or the greatest form of anything. It's spontaneous change that may or may not survive with no real reference to "the best."
Your comment reminds me of this discussion I had before here on HN and someone made a reply that has stuck with me to this day in its eloquence. I'll quote it here because I want to share what comment I'm talking about but out of context I think it makes little sense and I think you should read the thread [1]–it's very relevant to what we're talking about here.
> Essentially it is trying to say that we are at a global maximum in a space that is probably more nonlinear. In fact our genes may well be suboptimal in some respects even if you think they have "control".
This also seems to misunderstand evolution. You can think of evolution as a hill climbing optimization algorithm, with a complicated and constantly changing fitness metric. While it is true that any given change is random, and even 'good' changes still only succeed with some probability, over a long time period we see a series of incremental improvements. If there is an opportunity to increase efficiency it is likely that evolution would have found it, given how long the process has been running. Of course, this become less likely when we are discussing new features (the thumb is less optimized than DNA because it is so recent), and as the benefit of these optimizations becomes less. They also become less likely if their are fewer evolutionary paths that might lead to them.
That's not how it works. There's no single "global optimum" that lifeforms are evolving towards; there's no final hilltop; there's no intent. This notion of "Evolutionary Teleology" is how we get things like X-Men (which I enjoy for the record, it's not like X-Men is taught in biology class). If there were some global optimum, we'd all have eyesight like Legolas and the color-pallete of a Mantis-Shrimp, instead of a blind spot in the center of our retina.
Evolution just happens, and it is what it is. E.g. Natural Selection didn't "optimize" giraffes so they could reach leaves. Natural Selection is simply a way of explaining that some giraffes had long necks, some had short, and the shorter just happened to die out. Similarly, NBA players didn't "evolve" to play basketball. Some people are tall, and some are short, and the short people just didn't make the cut. "Select" != "Optimize". Evolution is better thought of as a historical accident. We're talking monkeys on a flying rock.
No, he got it exactly right. Evolution is not an algorithm, it doesn't optimize anything, changes are not improvements, it doesn't seek opportunities to improve efficiency.
I think the issue here was assuming that the optimization is the best possible one. It doesn't have to be, it just needs to be good enough to still function given past and present evolutionary pressure.
>Suppose we have a computer program which makes books. All of the commands which tell your computer program how to write a book are stored as sequences of 0's and 1's. Also, all of the letters, punctuation marks, and formatting symbols (line break, indent etc) are stored as specific sequences of 0's and 1's.
The commands which control the computer program and the actual text of the book are both stored in the same file. Up until now, we thought that these were placed side by side in the file, so you would have a segment saying "at 6am every day, print the following text in 12 pt font, and bind it in a hard red cover", followed by a text containing segment, and then another command segment saying "when you finish, print a copy of the book 134 positions further along in the file.
We thought this was the case because we have known the sequences of 0's and 1's that stand for each letter for a long time now. When we looked at a file and attempted to interpret it as If it all stood for letters which made up words' we would see something like "fffffgfgy6- fsjjjjjj the quick brown fox jumped over the lazy dog.bttt68-%jjjjjjfffffffff". After a while, we learned that the 'gibberish' segments were actually full of meaningful statements, but they were in a different language and contained the commands for the program.
Now getting to today's article, this group has found that inside those sections which code for the actual text of the book, there are some commands that run the program. This is accomplished because there is more than one way to write most of the letters. As an example, you could have one segment of text which says "It was the best of times, it was the worst of times", and another segment which says "It was the best of times, it was the worst of times" and also tells the book program to bind the book in gold leaf. And the only difference between the two is that the first one wrote the 's' in 'best' as 00010110, and the second one wrote it as 00010111.
This is very similar to earlier discoveries that the choice of codon can be intimately tied in with protein domains folding; allowing proteins to partially fold while the remainder is "stalled" on the ribosome waiting for the rare codon. It's interesting, but I would hardly call it a "second code".
Honest question... How is this different from what was discovered around 2007?
I ask, because insofar as I'm aware, the moment geneticians parsed the human genome was the moment it occurred to them that most (90%?) of the information was missing. (As in, information to make the remaining proteins that we build in our body.)
I vaguely recollect 2007 being the year where they discoverned and/or described how ARN can influence how ADN is read and turn bits and pieces of information on and off to use the same code to build new proteins. And another team the same year described how injecting ARN could allow to use existing ADN to potentially build proteins that aren't synthesized in the body.
It looks like there is software in DNA, not just firmware. We are just starting to understand that gene expression is very important too, code that express(is executed) or not based on the environment, and that junk DNA is probably "not understood code" .
DNA is a machine. It encodes information, but it also lives and breathes through interaction with various TFs, methylation events, packing, etc. There many types of epigenetic "state information".
We've known this for a looong time, it's just that it's so damned complex that we have trouble grappling with it.
And let's not forget the rest of the cell and all of the information encoded at the protein/messenger level.
I just completed a reasonably advanced course on computational biology and I was still taught that although we have found noncoding DNA performing functions, the majority of our DNA is both noncoding and not used for anything (i.e. evolutionarily obsolete, transposons, etc).
You are correct. Most non-coding DNA is junk, which amounts to around 90% of the genome. Dead transposons alone make up just under half of the human genome.
There is, in fact, a difference between "junk DNA" and "noncoding DNA," now that we see some noncoding DNA performing functions as switches, promoters, TF-binding sites (like the article describes), etc.
I am confused. I have absolutely zero biology background so please, bear with me.
So genes only code how to make proteins? That's IT? What about all the other stuff like what you look like, what diseases you may or may not get, some special functions of your body, your biological strengths and weaknesses, etc etc. Or defining protein generation actually defines all of that? (that would be so fascinating).
And 90% of the DNA is labeled as "non-coding?" ?!?!? Seriously? well it can't be "junk" can it? Or maybe its like a long bitcoin chain with only the latest commit relevant....
Yes, specifying which proteins are made, and how much of each and how that varies with cell type is most of what determines an organism's phenotype. This is largely because enzymes are proteins, and enzymes control all the biochemical reactions in a cell. But proteins can also be transcription factors, which control which other proteins are produced, and they can be important for their mechanical properties, etc. So, not some boring category of biomolecule that gets listed on nutritional information section on food packaging.
I am not a biologist, but this discovery seems ground-braking to me.
What strikes me as an interesting question (I hope someone of you can answer) is: How long will it be, before physicians/gene specialists/biologists/etc. will use the new way of DNA interpretation in cases for regular people? How long before this discovery can be used in mainstream medicine?
My (uneducated) guess is: there has to be a huge reevaluation of the accumulated data (and assumptions) about gene connections with deceases.
>How long before this discovery can be used in mainstream medicine?
Almost immediately. While it won't have therapeutic applications for some time, the first benefit will be helping scientists better understand the cause of certain diseases.
A protein is a chain of amino acids. There are 21 different amino acids. The alphabet of proteins is made of 21 different letters.
DNA strings are sequence of Deoxyribose Nucleic Acid (DNA). There are only 4 of them named A, T, C, G. The alphabet of DNA is thus made of 4 letters.
One needs at least 3 DNA letters to encode one of the 21 amino acids. These 3 DNA letters are called codon. But with 3 DNA letters there exist 4^3 = 64 different codons.
This is similar to a Base64 encoding just by the number of letters in the codon alphabet.
Since 64 is much more than 21, many different codons encode the same amino acid. This provides a liberty degree which is apparently used to encode something else. As suggested by this research result, the secondary encoding would sort of encode the predicate specifying when and maybe how the DNA sequence encoding a protein should be decoded.
So, there was a second code hiding in DNA, with different types of instructions, and those scientists discovered it.
That's nice, that's how science progresses.
Now, let's go to how business works. What about all those "scientists" (nay, engineers doing whatever for a payroll) and all those corporations, cocky and convinced that they had all the knowledge they needed, working as amateur magicians with DNA to create synthetic food, organs, new drugs, etc?
They essentially said "fuck" to a better and timely scientific understanding, in order to greedily rush to market some half baked results -- and consequences (medical or otherwise) be damned.
It might indeed change our understanding of specific point mutations, for example, but it won't negate any of their research, as far as I can tell. On the contrary, it will help them moving forward as we now have some basis on which to understand what is going on a little better in the steps between genotype and phenotype.
I figured I might get downvoted, but when I read the headline the konami code stuck in my head and I couldn't resist. It made me laugh, and I thought it might make others laugh. Which to me has (a little) value.
But now, you are pot and I am kettle (or vice versa). Your comment adds no value or, I'd argue less value, because at least I was trying to be funny. </derail>
"The most dangerous form of stupid comment is not the
long but mistaken argument, but the dumb joke. [...]
Bad comments are like kudzu: they take over rapidly.
Comments have much more effect on new comments than
submissions have on new submissions. If someone submits
a lame article, the other submissions don't all become
lame. But if someone posts a stupid comment on a thread,
that sets the tone for the region around it. People
reply to dumb jokes with dumb jokes."
By writing a dumb joke, you are signaling to new members that it's okay to rehash old, lame memes for nothing more than the lulz.
No, because any joke is a "dumb joke" in the context of a HN discussion. There's no debate about the smartness of them.
The "dumb" part is about it being a joke, as opposed to a serious contribution, argument or commentary. Not about the joke's quality as a joke. That is, even if you're Woody Allen, HN does not want you to write jokes in here -- and especially one liners without any other content.
Is it wrong that I find your anti-joke stance somewhat humorous?
I sometimes find it strange how much this community values orthodoxy. People work really hard to appear to be totally conventional HN readers. To share the same mannerisms, attitudes and opinions of the idealized HN user. I guess that this includes denying humor does not totally surprise me. It seems to be a caricature of itself.
One reason I value humor in my own life is that it reminds everyone of the absurdity which we are all living. I can't escape that this makes an HN user wrapped up in a culture of orthodoxy a bit uncomfortable.
>> HN does not want you to write jokes in here -- and especially one liners without any other content.
But humor is insight, and brevity is the soul of wit. I saw it as commentary on the headline itself, which was ambiguous and vague. I thought it meant scientists had found an easter egg or the Enigma code hidden in DNA.
>Not really. Most of the time is just the mental analogue of fart jokes.
I didn't say all insights were equal. If this was just a fart joke, I wouldn't have thought it was funny. But you can't draw a line between insight and humor, can you?
>No, that's just a cliche.
It's a cliche because it's Shakespeare. That doesn't make it less true.
Your advice sounds a bit pretentious to me. Here I was under the mistaken impression that people should be themselves. (Downvoters in three, two, one...)
RokStdy, don't be discouraged. You've done nothing wrong.
> Your advice sounds a bit pretentious to me. Here I was under the mistaken impression that people should be themselves.
No, no they shouldn't. Not if that means "take whatever action they feel like." One should take purposeful actions which serve to help their fellow HN'er and maintain the tremendously useful tool that this website has become. They should refrain from actions which decrease the signal to noise, or otherwise decrease the utility of HN.
That's not pretentious. That's the just an explicit statement of the implicit communal understanding which made this site what it is.
I hope that one day you can grow enough as a person to realize that authenticity is an important trait. I don't think I've very often heard "no, no you shouldn't" as a direct response to a claim that one should avoid being something you are not.
Why hate? Why wish downvoting? His comment will get buried down while more informative comments bubble to the top. Personally, I found his comment funny and I also learned about Konami codes, which I knew nothing about. So it satisfied my intellectual curiosity :)
HN discourages humor without content, to keep the culture focused around serious discussion. You can have a sense of humor, but it should be in the context of the topic and add something meaningful to the discussion.
There's nothing wrong with the Reddit culture of competing for top joke, but HN strives to be something different. The internet already has plenty of homes for "$nerdReference + $nostalgia == $karma++".
It's been pretty clear for some time that codon choice is not random (hence codon optimization is useful when taking a gene from one organism to another), and over the last few years it's also been clear that codon bias can be evolutionarily constrained. For a while, it mostly thought to be based on tRNA levels (basically anticodons). However, it's been increasingly clear that there are other constraints to protein coding sequences than just the decoded protein sequence (aka the genetic code).
For example, we published a paper a few weeks back [1] showing that in bacteria (and probably higher organisms), the N-terminus of genes has a lot of rare codons and this is due to other constraints such as relieving mRNA structure to allow better translation of proteins. I think in the coming years, we will find that other regulatory elements also shape this code, including sequences that control splicing, small RNAs, mRNA degradation and transport, et cetera
Anyways, it's a pretty fun time in biology. The tools we have now make studies that were ridiculously impossible just a few years ago, a reality for an individual lab. I can't wait to see what the next few years bring.
[1] http://arep.med.harvard.edu/pdf/Goodman_Sci_13.pdf\
EDIT: Since my comment has hijacked the most useful comment linking to the original study, I'll link to the comment here:
https://news.ycombinator.com/item?id=6897122