2016, a slightly perplexed review of neuroscience
In which we countdown 5 stories you may have missed in 2016
Hello, and welcome to the first annual The Spike year-in-review (ah, hubris, my old friend, good to see you again). 2016 has been quite eventful, or so Twitter tells me. Personally, I feel for Charlie Brooker. How the hell is he going to fit everything in to one hour on 2016 Wipe? It’ll be video montage from hell, with occasional cuts to Charlie’s haggard face, alone in his flat, echoing all our fears.
So let’s be strict about our remit, lest we wander down dark paths. We’re here to talk the great, the grin-inducing, and the god-awful in systems neuroscience in 2016. And in the wider world of science that affect us all, regardless of discipline.
We have, for your delectation:
5: Peer review is, sometimes, epically broken
4: Nature reaches a nadir — that maximum lifespan paper
3: The systems neuroscience arms race goes supernova
2: The slow, agonising collapse of Frontier’s reputation
1: Treating Alzheimer’s using a broken television?
So go stick on your favourite chart countdown music, and read on. Not got some handy? Click here [fancy a bit of all-time classic tech-house (yes, it’s a thing) instead? Here. Don’t knock over the plants with your flailing limbs once you hit play]
5: Peer review is, sometimes, epically broken.
Peer review. You submit a paper to a journal. Typically 2 to 4 of your peers review that paper, sending their comments back to the editor. The comments are collated, occasionally edited, and sent to you, for action. Typically that action is: swear like a squaddie who just lost his entire week’s wages on a horse while trapping his fingers in a door and dropping his pint.
Because, sometimes, peer review is stupidly broken. Of the many bits of evidence this year, I offer a mere sample.
Evidence 1. A paper was published in the EMBO Journal after 5 years in review. F-I-V-E. Not all at one journal, thank god. The senior author was so angry that he put out a press release, not to glory in the benefit all humankind will gain from this one paper, but about the review process. And how inhumanely awful it is at dealing with any project that dares cross boundaries between disciplines. We feel his pain.
Evidence 2. A paper was rejected from PLoS Computational Biology after 4 rounds of revision, because it was not “sufficiently of broad enough interest” beyond “computational neuroscience”. Granted, editors are typically busy academics, and are only human. And editorial standards at journals vary according to which editor is assigned to a paper. But one feels compelled to point out that minimal editorial standards have been breached here: (i) its relevance should have been obvious from its first submission; (ii) you’ve got “computational” in the bloody title of your journal.
This kind of treatment makes normally reserved, intelligent, circumspect people go bat-crazy ballistic (I literally just turned on Twitter to grab this latest example). Why? Because years of work have been poured into these individual papers. And people’s careers depend on their being published. It’s no good having your paper published 3 years after you no longer have a job.
Reassuringly, we discovered that even Karel Svoboda gets this incomprehensible crap from the reviewers of his papers. People asking for analysis and experiments that are not relevant to the study. People claiming that “we already knew this”, when we patently don’t. Bad news for him; heartening news for us lesser mortals who also have to deal with this kind of weirdness.
Repeat after me: “I am a responsible referee. I don’t ask for more experiments or analysis unless they are vital — vital — to test the main hypothesis of the paper. I don’t ask for more experiments or analysis just because there is something I’m curious about.”
“I also do not use abusive language in my reviews”. Unbelievably, people have to be told this.
4: Nature reaches a nadir
The journal Nature appears twice in this list. Here’s the bad entry. They published an atrocious paper claiming that humans had reached their maximum possible lifespan. How bad? The Dutch newspaper NRC has run three — one, two, three — investigative reports on how this paper made it into press.
Let me give you a flavour of its unmitigated awfulness. Its conclusion, its title, its whole reason for being published: the maximum possible lifespan of humans is not increasing. That conclusion is in its Figure 2a. And it fails basic statistics. The whole argument is that the age of the oldest person who died in each year is not going up. Here is their plot:
The issue? Their entire conclusion is based on that second, purple line: it does not go up. Which they said means: after about 1990, the oldest age at death has not increased — if anything, it’s going down. But that’s because they plotted two lines instead of one. Why did they plot two lines? No one knows; the paper does not say. If we just plot the data, with no lines, we can see actually the oldest age at death does kinda keep going up:
And if we just fit a single line to that data, we get…
There, one paragraph, and the whole paper is falsified. Using the exact same methods, on the exact same data, we can reach the exact opposite conclusion: maximum human lifespan is increasing.
(So which is true? Neither is true. That’s the point: with only one data-point per year, we can’t make any conclusions strong enough to be published in a quality journal, let alone the career-making, all-powerful, brutally selective journal that is Nature. It should never have been sent to review in the first place, as it is pure crap).
This took about 15 minutes. Most of that was grabbing the data. Thing is, Nature makes a big noise of having a permanent statistician on staff. Did they not see this paper? Worse, did they see this paper?
[Update 21/12/16: others teams reacted to this paper using far more careful, in-depth analyses then mine – but all with the same conclusion: if anything the maximum lifespan is increasing.
See Philipp Berens & Tom Wallis’ re-analysis and code here
Read Adam Lenart’s explanation using an actual model for the data here]
3: The technological arms race in neuroscience is super-heating.
That’s two crap things. Let’s have a happy thing. We want to record as many spikes, from as many neurons at the same time, as is possible within the laws of physics. Why? Because neurons talk to each other using spikes, and those spikes cause everything: seeing, hearing, moving, thinking. So understanding spikes is understanding the brain.
This year we’ve had huge leaps forward in:
* The potential to understand activity across the brain at single neuron level. We’ve had at least four papers on simultaneously imaging — filming — thousands of neurons in different regions of cortex, from: Svoboda; Chen & Helmchen; Sur; and Spencer Smith. With this tech, we can one day watch information flow from one area to another, watch sensation turn to action, thought turn to deed.
* The potential to understand what an entire local region of brain is doing. A different approach is to directly record each neuron’s electrical activity. Which needs electrodes. The problem is, their electrical activity is very weak. So the more electrodes you can cram in to the brain, and the closer they can get to the neurons, the better you will be able to sort the signal from the noise. Enter: Neuropixels. This recording tech, developed by Janelia Farm, and tested by the Allen Brain Institute and UCL, crams nearly 1000 electrodes into 1 centimetre; so each centimetre of the probe is able to capture hundreds of neurons at a time, and every spike from them. As dense a picture of the local activity in cortex as one could hope for.
* The potential to understand what causes neurons to fire. Imaging using calcium does not see spikes, but an indirect measure of their appearance. Recording spikes tells us which neurons were active, but not what caused them to be active. In a dream-world, we would be able to directly image all of a neuron’s changes in voltage: not just the (relatively) massive spike, but all the flickers of voltage preceding it, caused by inputs to that neuron. Then we could actually see what caused a neuron to spike. We took a dramatic step closer to this fairy-tale this year. At the start of December, Mark Schntizer’s lab introduced the world to TEMPO, a method for imaging the voltage changes across all neurons in a small area. With this, they could see when inputs to these neurons were synchronised, and whether these inputs were inhibitory or excitatory. And all in the brain of a moving, thinking mouse.
(For the record, this kind of technology has worked in invertebrates — leeches, sea slugs — for a couple of decades. And these invertebrate guys and gals are almost at the point where they can record the dreamt of flickers of voltage that cause spikes.)
* Knowing where those spikes are going after the neuron has sent them. This is the other key to understanding the brain, knowing which neurons are receiving each message. This year, we’ve had the publication of Tony Zador’s frankly sci-fi RNA barcoding technique, which allows tracing individual connections between untold numbers of neurons. Basically, you inject synthetically sequenced RNA — the barcode — into each neuron, a unique sequence in each one. The barcode RNA is transported along the axons of the neurons, across the synapses, and into the cell bodies of the target neurons. So you take the area of brain at the target end, and sequence the RNA in each cell body: when you find the unique RNA barcode — bingo! The pair of neurons were connected. We can finally know, for certain, that one neuron passes messages to another.
Hang on, hang on, it says “potential” a lot back there. Why just “potential”? Because, these technological advances are only the first step. The second step is to do some science with them. Test an hypothesis. Though it would help if we had some hypotheses to test.
2: The slow, agonising collapse of Frontiers’ reputation.
Oh, Henry. Frontiers in Neuroscience launched in November 2007, with two simple, but thrilling* innovations. (* Thrilling to academic publishing wonks). First, the main journal was supported by a set of online only specialist journals, things like Neural Circuits, and Computational Neuroscience. The best papers in those specialist journals were up-rated to the main journal, ensuring the community decided what was good, and what wasn’t. This, sadly, did not last long.
Second, the review process was — and is — entirely online: each reviewer has their own forum, and the authors, reviewer, and editor can in principle discuss the requested changes before doing them. This cuts out months of wasted time sending written reviews back and forth by email. So when a reviewer writes something you don’t understand, you can directly ask them what they meant. Lovely.
Frontiers was also the first fully open-access neuroscience journal: you pay a fee when you publish your paper, and anyone can download it for free. It appeared to be the future. How badly things have gone wrong.
Alarm bells started ringing a few years ago when overnight something like 40 new journal titles appeared on their website. That immediately says: we’re in it for the money, not the quality.
The amount of unsolicited communications from Frontiers has also sky-rocketed, inviting proposals for special issues on the flimsiest of pretenses. That immediately says: we’re in it for the money, not the quality.
Jeffrey Beall blacklisted Frontiers in 2015, on the grounds that they act like a “predatory” publisher, there to make money not publish quality science. And it’s hard to argue with him.
And this year? Where to begin? A list:
(1) They visited Jeffrey Beall’s institution, and wrote threatening letters to his employers.
(2) Reports of pressure-selling tactics, corporate-sales style
(3) Reports of editorial pressure to wave through terrible papers, for the money.
(4) They published papers on:
(i) how bread can cause mental illness
(ii) incomprehensible gibberish
(iii) more incomprehensible gibberish
(iv) a particularly bizzare conspiracy theory
(v) an anti-vaccination study. One that was based on a tiny online-only survey of mothers with home-schooled children, and asked for their opinion of their child’s cognitive problems. This was pulled within hours of being posted online.
[Hat-tip here to Leonid Schneider’s tireless pursuit of dodgy practices throughout science: read his investigative reportage here].
The core neuroscience journals, which have been there since the start, have had their reputation irrevocably tarnished. Yet another small area of academia has been distorted by money.
1: We could treat Alzheimer’s by looking at a broken television?
A few weeks ago, another high-profile trial of an Alzheimer’s disease treatment failed miserably. Stopped in its tracks because there was no detectable benefit of taking the drug. The drug tried to clear garbage proteins — beta-amyloids — from the brain. But if it did, then that did not alter the decline of memory and of the ability to think.
Bleakness. Then, totally out of left-field, appeared one of those “that’s just bonkers” papers in Nature (lead authored by Hannah Iaccarino & Annabelle Singer, led by Li-Huei Tsai ). Bonkers in a good way: treating Alzheimer’s by looking at a screen. For an hour.
Cortex, when working well, has a strong signature we can detect even with recordings from outside the brain (such as EEG). Neural activity across many areas of cortex briefly goes up and down at 40 times a second, particularly when paying attention to something. In Alzheimer’s disease, this signature is disrupted.
So the authors thought “hey, what if we put the 40 times-a-second oscillation back in? Would that help?”
Yes. Yes it would.
They took some mice with something akin to Alzheimer’s. They directly put the oscillation back in by stimulating specific neurons in cortex, that would transmit the oscillation across all the surrounding cortex. When they did this, the garbage proteins were cut by half. Cool.
(Weird, though. For how on earth does making neurons fire together clear garbage proteins from outside of the neurons? No one knows. But key are likely to be the poor relations in neuroscience, the cells in the brain that are not neurons: glia. Think of the brain as a Victorian country house drama. In the main house, the aristocrats, feasting, gossiping, cavorting; making, to the outside world, all the visible noise and motion. The neurons, hoggers of the limelight. In the pokey attic and dank basement, the servants, cleaning, preparing, clearing the rubbish. The quiet glia. The Alzheimer’s brain is like what happens when the servants go on strike. Very shortly, the aristocrats are wallowing in their own filth, starving, unable to feast, gossip, or cavort. Slowly dying of ineptitude. Presumably, in this metaphor, the 40 times-a-second oscillation is the equivalent of giving the servants the raise they asked for).
Now, it so happens that we can make cortex produce this 40-times-per-second oscillation quite easily: we show it some light flickering at 40-times-per-second. So the authors put their mice in a box with such a light, and left them for an hour. The garbage proteins were cut by half.
If this study is true, then we could treat Alzheimer’s disease using some really low-tech solutions. No pills, no surgery, just a flickering light.
To throw one out there. You may remember that good old cathode ray tubes — televisions and monitors that weighed more than a gorilla — refreshed at 50 or 60 times-a-second. So if we take one of those old televisions, and break it to refresh at 40 times-a-second, then we have our flickering light. Perhaps it may even work at those higher 50 and 60 time-a-second rates.
That’s a big if. There are suggestions already of some problems with this paper. Not least that, as is sadly typical, many experiments only used a handful of animals — so we’ve no idea whether the observed effects are robust or not. Even so, with a cheap treatment with no side effects, there is literally no harm in trying this in a clinical trial. (Except, of course, that there is no money to be made here, because nothing needs to manufactured and sold. So who will fund the trial?)
And, even if it turns out to be hogwash, the paper has put an exciting idea out there: the potential for treating Alzheimer’s without needing to mess with the brain at all — no drugs, no surgery, no electrodes, just simple, everyday technology. And that’s a dream worth pursuing.
There was so much more. Marcus Meister using back-of-the-envelope calculations to show that recent grand claims of controlling neuron activity using magnetism were not physically possible. Optogenetics now working — properly — in primates. Finding out that even systems neuroscience has trolls (don’t you just love prose written entirely in ellipsis…? Gives more insight… into the writer’s psychological… fragmentation… than… a… year of… therapy).
With Obama signing over $1.5 billion to fund the next 10 years of the BRAIN project in one of his final acts as President, there’s a good chance there are more spectacular advances to come. There are now so many national “brain” initiatives that the journal Neuron dedicated an entire issue to discussing them. I’ll leave it to the reader to decide whether that is a good thing. The Golden Age of Systems Neuroscience is showing no signs of abating.
Want more? Follow us at The Spike
Twitter: @markdhumphries
