I think there must be a gene for it: the innate ability to think about biological cause-and-effect as an abstract pathway instead of as concrete machinery. Classical geneticists do it effortlessly. Their vocabulary is old and rich: synthetic lethal; partial rescue; incomplete penetrance. And these phrases evoke a time when biology was all smooth and wrinkled peapods, ochre- or amber-eyed fruit flies, striped petunias.
In the minds of true geneticists, genes enhance and suppress, and exert their effects upstream and downstream of other genes. And thought processes are unsullied by the need to picture what the gene products are actually doing. Proteins binding, proteins being modified, proteins translocating to the nucleus – it can all be irrelevant, if you just let it wash over you. Of course many geneticists nowadays can and do care about proteins, but they still retain this enviably breezy grasp of the almost ghostly conceptual scaffold that can be placed around them.
If there is a gene for the geneticist mindset, then I am surely a homozygous null mutant. Of course certain thought experiments are easy. In the following example,
A –> B –> C –> Effect
it’s pretty straightforward: A activates B which activates C which leads to a perceptible manifestation. These could be a chain reaction of kinases, say, leading to the onset of mitosis. Knock out B and mitosis no longer occurs: but knock out B while simultaneously adding a dose of exogenous C and mitosis is rescued. If you hadn’t already known that this pathway unfolded in that order, an epistasis experiment like this could help you to sort it out.
It all gets a bit sticky when your hypothesized pathway is more complicated, as in
A –| B –| C –> D –| Effect
that is, when your players are inhibiting as well as activating. And when you start knocking down multiple genes at once to get a feel for the order, it begins to seem, qualitatively, like one of those drunken sentences: “It isn’t that I don’t not dislike you, but…”
Fortunately, it’s nothing that an intense session with pen, paper, salted cashews and a few cocktails down at Henry’s Bar with a good friend can’t clear up. After a few Cosmopolitans, this genetics thing is doddle.
I have a figure in my thesis that is kind of like that, and it was wrong until the last moment. In the version I defended I still had drawn it as “the inhibited B activates C, so let’s draw an arrow here” instead of two inhibity-things from A to B and from B to C. But I passed anyway. It was the only “mistake” in there.
Someone should invent a word for that inhibity-thing.
Someone should invent a word for that inhibity-thing.
Sparrow? (Conjunction of ‘spanner’ — as in ‘spanner in the works’ — and ‘arrow’?
As a confirmed proteinist, I don’t get the geneticist mind-set at all and feel aggrieved when people refer to my little buddies as ‘gene products’. Ugh! I know one crystallographer who tried to fight back, dubbing genes ‘protein products’.
And thought processes are unsullied by the need to picture what the gene products are actually doing. Proteins binding, proteins being modified, proteins translocating to the nucleus – it can all be irrelevant, if you just let it wash over you.
In fact, close study of mutations is the best way of figuring out protein function. It’s not just geneticists that let any thought of protein function “wash over them”, although many don’t in their defense. I still get angry every time I see someone put up a slide with a bunch of expresssion array data (i.e. mRNA levels), then go on in the next slide to show a biochemical pathway (often a kinase signalling pathway) and talk about how the changing levels of message are obviously causing activation/inhibition of said pathway. My bulls#@t meter starts to twitch. Entire pathways are switched on and off in very short time-frames without any change in mRNA levels.
Of course, you can always invoke the magical and mysterious “scaffolding” function for a protein when you don’t yet know what it’s actual function is ;)
Ah but Stephen, either A, B, C or D could be a microRNA or other non-protein type thingy. And where’s your fancy proteinism then, eh?
Ooh. There’s a fight brewing, Stephen.
Jenny, I propose the word ‘Shplop’. Just for entertainment value.
(and what, no pistachios?! Please tell me there were macadamias, at least)
The fact that Darren and I are commenting simultaneously, during work, probably means that the boss should pay more attention ;)
oh well, at least we’re talking about science (Darren more than me).
I’m hiding from a training course on our new FPLC system. I lost patience when one of my fellow attendees took 15 minutes to get the concept of a valve between buffer bottle and pump that would allow simultaneous connection of more than one buffer to the system. It was going to be a looong afternoon.
I’m taking a break from editing a very challenging piece of writing
Even miRNA’s need proteins to make them work… ;)
Shush, I’m trying to
annoy Stephenre-edjumicate a proteinistp.s. BSc (Hons) Genetics, University of Newcastle upon Tyne, 1998 ;)
Giggle
Welcome to the world of NF-kB signaling…
Smacks head
If proteins are gene products, surely genes would be protein quotients? Or do genomic proteionimists not like maths?
Ooh. Math joke. At least we can all peer down our noses at those.
Mike, I love the idea of ‘protein quotients’.
Darren, I just wanted to stress that this post was not a trashing of the geneticist mindset, but an homage. In fact, when planning my RNAi epistasis experiment and trying to predict which combinations of knockdown would lead to which phenotype, it wasn’t til I stopped thinking about the proteins that I got anywhere. It was killing me, imagining protein B being knocked down, so it could no longer bind and presumably
inhibitshplop protein C — but protein C isn’t there either, and knocking down either of these also gives separate phenotypes, so when you do it together — argh. But when I did the geneticists’ trick of saying, when knocking down B and C together, whichever phenotype ‘wins’ is downstream — suddenly it was easy.Now I know it’s not always the case: B and C could do more than one thing (bifurcating pathways). But it’s a great place to start. And then you can use overexpressing rescue to test your hypothesis more cleanly.
By the way, as someone who’s spent a lot of time thinking about phosphorylation, I too get irate at these microarray folks. It’s a good first step, but so many proteins are regulated in milliseconds at the post-translational level, it’s got to be followed up.
And then you can use overexpressing rescue which sounds like something the AA might offer.
Isn’t this just one part of a longer continuum though? Beyond the geneticist you have zoologists and ecologists and Gaia-ologists, getting ever more high-level. Beyond the molecular biologist you have chemists and physicists, getting ever more low-level.
Here I am, stuck in the middle with you.
My Dear Dr Ennis,
we devout proteinists have achieved an advanced superconscious state where annoyance is not possible. If you look very, very closely, you will see that all supposedly non-proteinaceous entities (miRNA and such-like) are merely harmonic energy vibrations in protein space.
Yours amino acidly,
Stephen
Harmonic energy vibrations…
I think Stephen’s been spending too much time on those alternative medicine websites.
I think he’s been [potentially libelous statement redacted].
Stephen, shouldn’t it have been Yours amino acidically?
Centrifugationizationally?
Don’t you dare scoop my future post, Richard.
What, me? innocent look
Frnak, your comment reminded me of xkcd’s excellent "height"http://xkcd.com/482/ and depth comics.
Crap.
height
Oh and by the way, Jennifer, as one who took* her PhD in genetics and remembers with a glowing nostalgia my first lecture on epistasis, I adored this post.
*as discussed last night at the London Science Tweetup, ‘took’ is a far superior verb here than ‘earned’ or ‘received’
Karen, taking the time to preview your posts is like taking the time to smell the roses.
And thanks for your kind words. Maybe you can come round sometime and help me understand how it can be that a gene (A) that is clearly upstream of B by a number of criteria also seems to contribute downstream to B’s phenotype. These complexities are what is really killing me.
@Jenny – Stephen, shouldn’t it have been Yours amino acidically?
No, he hissed. Acidly.
Basically, you’re right.
I’m glad I’m buffered from genetics pathways.
Yes, it’s a litmus test of intellect you might just fail.
I’m suddenly reminded of something that happened in our office yesterday. It’s the office of a biology program, and students need to pick their courses for next year, so they often call to ask if something is still available or how to combine courses. Yesterday, my coworker took a call from a girl who said “My friends say that there is a really easy course in your department. Which one is it?”. I later said: “You should have just told her it’s the genetics course.” =P (Also, why didn’t her friends just tell her which course it was? What kind of friends are those?)
At least you lot actually talk the language of pathways etc etc. Us poor systems types just glaze over completely in biochem / mol biol talks when 5 minites in we see the slide with the multiple intersecting kinase pathways and the endless abbreviations.
The side is always introduced with the comment: “Here is a schematic, very very simplified, of course….”
The only way I ever managed to get my own back was the time someone asked me to do a spoof talk in this vein for a conference. One thing I discovered was that the single letter amino acid code offers many possibilities for juvenile jokes.
My previous favourite protein, when the two domains were aligned, had the letters
FAARTS
WAARSE
in its sequence.
I made a point of subtly highlighting those when I gave talks.
My previous favourite protein
What’s the current favourite?
@Austin: come on, you know you want to…
Actin.
I really like the systems biology fuzzy balls. And I adore the concept of ‘date hubs’ and ‘party hubs’.
What happens when two fuzzy balls go off into a corner by themselves? Or is that pushing the metaphor too far?
Gaaaah. I was never comfortable with the concept of epistasis, that if mutation in gene A suppresses the phenotype of mutation in gene B, then therefore the proteins were doing something or other that was easy to understand. It’s all black magic, witchcraft and voodoo, I tells ya.
For tremendous fun and games, get a statistician and a geneticist arguing over whether an odds ratio that is higher for two mutations together than would be expected from their individual odds ratios alone represents “epistasis”, an “interaction”, or something else. Hours of
fun and gameseye-watering boredom. Really.What happens when two fuzzy balls go off into a corner by themselves?Not sure, but it probably has something to do with this gene
Heh. Reminds me (*Richard W) about the endless arguments signal transdutionists have about additive versus synergistic effects.
Yesterday I overheard one PhD student asking another, ‘Can you help me find a statistics test that will make this difference significant’?
The youth of today.
It’s like getting one person from Montreal and one from New York in the same place, and saying the magic word, “bagels”. The argument could last for weeks, or, indeed, forever.
Wintle’s rule of statistical tests: the more lengthy and obscure the name, the better. I recently used a Kruskal-Wallis nonparametric test with Gaussian approximation of p-value, which made me feel all warm and fuzzy inside. Like those tribbles of which Cath speaks.
You dissing the humble chi square?
Jennifer, I can totally relate to having to “let it go” in order to actually get on with the expts, and try to make sense of it all. You did remind me of my other pet hate at the moment… using only qPCR (or similar) to show knockdown in RNAi expts. What about the protein levels hmmm? Before you jump down my throat, I know it’s just not possible to do western blots on every gene you’re targetting in a screen, but it’s not that hard to do a few westerns on the genes/proteins you follow up?
@Austin: come on, you know you want to…
I can’t remember the ruder ones, as it was a few years ago now and the more vulgar ones didn’t make it into the final talk – family audience and all that.
The talk was a stem cell spoof – had to be on something achingly fashionable – in which stem cells would be driven by my fictional kinase pathway to differentiate to adipocytes and cause obesity. Anyway, I was sorely tempted to give my key dimeric receptor kinase (which was called Kit-Kat) the motif:
Phe-Ala-Thr-Ala-Arg-Ser-Glu
Not very PC, though.
bq- For tremendous fun and games, get a statistician and a geneticist arguing over whether an odds ratio that is higher for two mutations together than would be expected from their individual odds ratios alone represents “epistasis”, an “interaction”, or something else. Hours of
fun and gameseye-watering boredom. Really.As someone who’s both, I’m keeping out of that. Unless you really want me to…
Of the p-value or the statistic?
@Bob – you could have the argument with yourself, you know. ;)
Also – I believe it was approximating the p-value. Does that make any sense at all? GraphPad Prism (yes, I know) says: if your samples are small and no two values are identical (no ties), Prism calculates an exact P value. If your samples are large or if there are ties, it approximates the P value from the chi-square distribution, which doesn’t appear to be a Gaussian approximation, does it? Very puzzling.
@Jenny – not at all. Chi-square is a favourite of mine as it’s really, really easy to calculate. It’s just not useful for a whole lot of things that
people like meuninformed geneticists tend to use it for, is all.but it’s not that hard to do a few westerns on the genes/proteins you follow up?
Unfortunately as I’m involved in screens for the discovery of functions for new genes, there aren’t any antibodies to the proteins I want to follow up. In some ways qPCR is easier than cloning and tagging. Also, qPCR can help you tighten up your shortlist – with about 300 hits, I need all the tightening I can get.
Phe-Ala-Thr-Ala-Arg-Ser-Glu
Giggle.
Also, qPCR can help you tighten up your shortlist – with about 300 hits, I need all the tightening I can get
Ah yes, that’s the trick with screens, we’ve spent the last year trying to narrow down a list of ~500 by various means… some more effective than ohters.
Pinning up the spreadsheet and throwing darts at it is quite effective, I find.
Ah, strictly it is approximating the statistic, and calculating the p-value from the approximated statistic. Oh, and a chi-squared is the sum of squared normal distributions, so you’re right at a deeper level.
I think it’s oddly reassuring that I didn’t understand a word of that.
My biological squishiness is fully intact.
Good. That’s how we statisticians keep our jobs.
It is equally reassuring to see that those identifying with biological squishiness can coexist with those adoring statistical abstactions and algorithms, without throwing too much labgear at one another’s heads…
That’s where the squishy biologists have the advantage, of course.
@Bart: that’s only because we’re so avant-garde here on NN.
It is in our best interests to get along, Bart. After all, one day I will have to go to One of Them and ask for help with my stats.
Aaaah, but any single one of the “One of Them” will require you to fiddle and tinker around in your lab. No squishy stuff, no data. (Secretly, “They” may be referring to you in capital letters as well, Jennifer).
@Richard: Well, supercomputers do make great projectiles as well (although they require a bit more convincing…).
Bart, there’s nothing quote so scary as a virologist quietly saying,
‘Oops.’
Except, possibly, a nuclear physicist saying the same thing.
‘quite’. Quite so scary. Bah. I really do need more coffee.
While it is in the oops that lessons lie…
…it is in the grounds that our coffee languishes.
(that’s a sophisticated pun in case anyone missed it.)
groan
Here I am, stuck in the middle with youStealer’s Wheel. Great pub-rock standby. There’s a funny chord in the middle eight, though, that I can never quite get right.
It’s one of my favorite songs – I’ve been trying to talk the band into doing it.
Bags I don’t get my ear cut off.
Jenny – if you can work out those weird chords at the end of the middle eight, you will let me know, won’t you?
You mean the ‘please help me’ bit?
Jenny – when you have to go to One of Them™ for help, just remember that you will continue to not understand what they’re talking about. This situation can persist for the entire process, from initial analysis through to publication. Don’t ask me how I know.
@Bob – thanks. Wintle’s knowledge of stats has now been exceeded. Please exit brain via left ear. Thank you.
I am perfectly happy to be entirely ignorant when I work with specialist collaborators. Maybe this is lazy, but if I took the months/years to thoroughly understand, I might as well do it myself.
Yes, that bit. Eponymously.
Oooh. Someone’s been listening to the podcast
That’s one of the interesting things that could do with more analysis (and then a blog post). I prefer it if the squish-jockey can do the analysis them selves, with my help, because they should understand better what’s been done. But there comes a point where it’s quicker if I do it. I should think a bit more about how I find out where that point is.
Squish-jockey! Oh, that is brilliant. I want that embroidered on my lab coat.
Bob, I was referring to really really high-tech stuff that takes years to master, like certain kinds of computational modeling, or equipment you need an extra PhD just to look at. I’m happy to do/learn more basic stuff.
@ Bob. The interaction between wet and dry specialists and the ways in which they manage to create knowledge together is something I am very much interested in. I have done the analysis (see here) but I haven’t blogged about it…