Category Archives: Neurology & Psychiatry

Functional MRI research is a scientific sewer

First a primer about the science underlying functional Magnetic Resonance Imaging (fMRI). Chemists use MRI all the time, but they call it Nuclear Magnetic Resonance. Docs and researchers quickly changed the name to MRI because no one would put their head in something with Nuclear in the name.

There are now noninvasive methods to study brain activity in man. The most prominent one is called BOLD (Blood Oxygen Level Dependent), and is based on the fact that blood flow increases way past what is needed with increased brain activity. This was actually noted by Wilder Penfield operating on the brain for epilepsy in the 1930s. When a patient had a seizure on the operating table (they could keep things under control by partially paralyzing the patient with curare) the veins in the area producing the seizure turned red. Recall that oxygenated blood is red while the deoxygenated blood in veins is darker and somewhat blue. This implied that more blood was getting to the convulsing area than it could use.

BOLD depends on slight differences in the way oxygenated hemoglobin and deoxygenated hemoglobin interact with the magnetic field used in magnetic resonance imaging (MRI). The technique has had a rather checkered history, because very small differences must be measured, and there is lots of manipulation of the raw data (never seen in papers) to be done. 10 years ago functional magnetic imaging (fMRI) was called pseudocolor phrenology.

Some sort of task or sensory stimulus is given and the parts of the brain showing increased hemoglobin + oxygen are mapped out. As a neurologist as far back as the 90s, I was naturally interested in this work. Very quickly, I smelled a rat. The authors of all the papers always seemed to confirm their initial hunch about which areas of the brain were involved in whatever they were studying. Science just isn’t like that. Look at any issue of Nature or Science and see how many results were unexpected. Results were largely unreproducible. It got so bad that an article in Science 2 August ’02 p. 749 stated that neuroimaging (e.g. functional MRI) has a reputation for producing “pretty pictures” but not replicable data. It has been characterized as pseudocolor phrenology (or words to that effect). Keep reading you’re about to find out just why this was.

What was going on? The data was never actually shown, just the authors’ manipulation of it. Acquiring the data is quite tricky — the slightest head movement alters the MRI pattern. Also the difference in NMR signal between hemoglobin without oxygen and hemoglobin with oxygen is small (only 1 – 2%). Since the technique involves subtracting two data sets for the same brain region, this doubles the error.

Under two years ago, it was shown that 70% of people having functional MRIs (fMRIs) were asleep during the test, and that until then fMRI researchers hadn’t checked for it. For details please see
https://luysii.wordpress.com/2014/05/18/how-badly-are-thy-researchers-o-default-mode-network/. You don’t have to go to med school, to know that the brain functions quite differently in wake and sleep.

Recent work shows that functional MRI work is even worse. A devastating report in [ Proc. Natl. Acad. Sci. vol. 113 pp. 7699 – 7600, 7900 – 7905 ’16 ] showed that certain common settings in 3 software pacakages (SPM, FSL, AFNI) used to analyze fMRI data gave false positive results ‘up to’ 70% of the time. Some 3,500 of the 40,000 fMRI studies in the literature over the past 20 years used these settings. The paper also notes that a bug (now corrected after being used for 15 years) in one of them also led to false positive results.

Here’s a bit more detail on what they did. It turns out that analyzing one voxel (essentially a single MRI pixel) at a time produces valid results. The problem comes when multiple voxels (clusters) are analyzed together. Clusterwise inference considers both the strength of activity at spots throughout the brain as well as the size of the spots. When a parameter called the cluster defining threshold (CDT) is set too low, the analysis is more likely to be false positive. This was true for all 3 packages tested. Parametric statistical methods produce the problem (not for voxels but for clusters). It relies on Gaussian Random Field Theory (RFT) for clusters , which depends on two other assumptions (1) the spatial autocorrelation function has a squared exponential shape — e.g. Gaussian (2) the spatial smoothness of the fMRI signal is constant over the brain. Neither of these assumptions is correct. Those of you who’ve read Nassim Nicholas Taleb about the stock market know about ‘fat tails’. It turns out that the spatial correlation function has them. Here’s what a fat tail is all about. Human height goes fall quite nicely into a Gaussian distribution. There are 7 and 8 footers about but they are rare. If the human height distribution wasn’t Gaussian but had a fat tail, we’d see 12 and 15 footers.

If that wasn’t bad enough,the following is even worse (in my opinion). 40% of 241 recent fMRI studies didn’t report using well known methods for correcting for multiple testing. They may have done so, but every biomedical paper, and drug study says so explicitly. Not only that but drug studies are required to explicitly state the hypothesis (or hypotheses) they are testing.

This is probably why in the early days, fMRI researchers always confirmed their original hypothesis. They could test the massive fMRI for statistical rarity, and since the data was so large, find it and post hoc propter hoc publish it. Possibly they did so out of ignorance, but even so this is inexcusable

Science proves cognitive training will raise your IQ 5 – 10 points

Who among you doesn’t want to be smarter? A placebo controlled study with 25 people in each group showed that cognitive training raised IQ 5 – 10 points [ Proc. Natl. Acad. Sci. vol. 113 pp. 7470 – 7474 ’16 ].

You know that there has to be a catch and there is. The catch points to a problem with every placebo controlled trial ever done, particularly those with drugs, so drug chemists pay attention.

What was the placebo? It was the way subjects are recruited for these studies. Of 19 previous studies in the literature, 17 recruited patients using terms like ‘cognition’ or ‘brain training’, so the authors put out two ads for subjects.

Here are the two ads they used

Ad #1

Brain Training and Cognitive Enhancement
Numerous studies of ahown that working memory training can increase fluid intelligences (several references cited)
Participate in a study today !
EMail for more information GMUBrainTraining@Gmail.com

Ad #2

EMail Today and Participate in a study
Need SONA credits? (I have no idea what they are)
Sign up for a study today and earn up to 5 credits
Participate in a study today !
cforough@masonlive.gmu.edu

I might mention that the two ads were identical in total size, font sizes, coloration used etc. etc.

” Two individual difference metrics regarding beliefs about cognition and intelligence were also collected as potential moderators. The researchers who interacted with participants were blind to the goal of the experiment and to the experimental condition”  Not bad. Not bad at all.

The results: those recruited with ad #1 showed the increase in IQ, those recruited with ad #2 showed no improvement.

It was an expectancy effect. Those who thought intelligence could be raised by training, showed the greatest IQ improvement.   Every sick patient wants to get better, and any drug trial simply must mention what it is for, the risks and rewards, so this effect is impossible to avoid. It probably explains the high placebo response rate for migraine and depression (over 30% usually).

What is really impressive (to me at least) is that the improvement was not in a subjective rating scale (such as is used for depression), but in something as objective as it gets. IQ questions have a right and wrong answers. You can argue about whether they ‘really’ measure intelligence, but they measure what they measure and fluid intelligence is one of them.

Medicine is full of fads and fashions, sugar is poison, fat is bad (no it’s good) etc. etc. and this is true in spades for treatments, particularly those touted in the press. Next time you’re in a supermarket, look at the various nostrums mentioned in the magazines at the checkout stand.

When I first started out in practice, one particular headache remedy was getting great results. The rationale behind it seemed bizarre, so I asked a very smart  old GP about it — his advice — “use it while it works”. Rest in peace, Herb

Why you do and don’t need chemistry to understand why we have big brains

You need some serious molecular biological chops to understand why primates such as ourselves have large brains. For this you need organic chemistry. Or do you? Yes and no. Yes to understand how the players are built and how they interact. No because it can be explained without any chemistry at all. In fact, the mechanism is even clearer that way.

It’s an exercise in pure logic. David Hilbert, one of the major mathematicians at the dawn of the 20th century famously said about geometry — “One must be able to say at all times–instead of points, straight lines, and planes–tables, chairs, and beer mugs”. The relationships between the objects of geometry were far more crucial to him than the objects themselves. We’ll take the same tack here.

So instead of the nucleotides Uridine (U), Adenine (A), Guanine (G), Cytosine (C), we’re going to talk about lock and key and hook and eye.

We’re going to talk about long chains of these four items. The order is crucial Two long chains of them can pair up only only if there are segments on each where the locks on one pair with the keys on the other and the hooks with the eyes. How many possible combinations of the four are there on a chain of 20 — just 4^20 or 2^40 = 1,099,511,621,776. So to get two randomly chosen chains to pair up exactly is pretty unlikely, unless in some way you or the blind Watchmaker chose them to do so.

Now you need a Turing machine to take a long string of these 4 items and turn it into a protein. In the case of the crucial Notch protein the string of locks, keys, hooks and eyes contains at least 5,000 of them, and their order is important, just as the order of letters in a word is crucial for its meaning (consider united and untied).

The cell has tons of such Turing machines (called ribosomes) and lots of copies of strings coding for Notch (called Notch mRNAs).

The more Notch protein around in the developing brain, the more the proliferating precursors to neurons proliferate before differentiating into neurons, resulting in a bigger brain.

The Notch string doesn’t all code for protein, at one end is a stretch of locks, keys, hooks and eyes which bind other strings, which when bound cause the Notch string to be degraded, mean less Notch and a smaller brain. The other strings are about 20 long and are called microRNAs.

So to get more Notch and a bigger brain, you need to decrease the number of microRNAs specifically binding to the Notch string. One particular microRNA (called miR-143-3p) has it in for the Notch string. So how did primates get rid of miR-143-3p they have an insert (unique to them) in another string which contains 16 binding sites for miR-143-3p. So this string called lincND essentially acts as a sponge for miR-143-3p meaning it can’t get to the Notch string, meaning that neuronal precursor cells proliferate more, and primate brains get bigger.

So can you forget organic chemistry if you want to understand why we have big brains? In the above sense you can. Your understanding won’t be particularly rich, but it will be at a level where chemical explanation is powerless.

No amount of understanding of polyribonucleotide double helices will tell you why a particular choice out of the 1,099,511,621,776 possible strings of 20 will be important. Literally we have moved from physicality to the realm of pure ideas, crossing the Cartesian dichotomy in the process.

Here’s a copy of the original post with lots of chemistry in it and all the references you need to get the molecular biological chops you’ll need.

Why our brains are large: the elegance of its molecular biology

Primates have much larger brains in proportion to their body size than other mammals. Here’s why. The mechanism is incredibly elegant. Unfortunately, you must put a sizable chunk of recent molecular biology under your belt before you can comprehend it. Anyone can listen to Mozart without knowing how to read or write music. Not so here.

I doubt that anyone can start from ground zero and climb all the way up, but here is all the background you need to comprehend what follows. Start here — https://luysii.wordpress.com/2010/07/07/molecular-biology-survival-guide-for-chemists-i-dna-and-protein-coding-gene-structure/
and follow the links (there are 5 more articles).

Also you should be conversant with competitive endogenous RNA (ceRNA) — here’s a link — https://luysii.wordpress.com/2014/01/20/why-drug-discovery-is-so-hard-reason-24-is-the-3-untranslated-region-of-every-protein-a-cerna/

Also you should understand what microRNAs are — we’re still discovering all the things they do — here’s the background you need — https://luysii.wordpress.com/2015/03/22/why-drug-discovery-is-so-hard-reason-26-were-discovering-new-players-all-the-time/weith.

Still game?

Now we must delve into the embryology of the brain, something few chemists or nonbiological type scientists have dealt with.

You’ve probably heard of the term ‘water on the brain’. This refers to enlargement of the ventricular system, a series of cavities in all our brains. In the fetus, all nearly all our neurons are formed from cells called neuronal precursor cells (NPCs) lining the fetal ventricle. Once formed they migrate to their final positions.

Each NPC has two choices — Choice #1 –divide into two NPCs, or Choice #2 — divide into an NPC and a daughter cell which will divide no further, but which will mature, migrate and become an adult neuron. So to get a big brain make NPCs adopt choice #1.

This is essentially a choice between proliferation and maturation. It doesn’t take many doublings of a NPC to eventually make a lot of neurons. Naturally cancer biologists are very interested in the mechanism of this choice.

Well to make a long story short, there is a protein called NOTCH — vitally important in embryology and in cancer biology which, when present, causes NPCs to make choice #1. So to make a big brain keep Notch around.

Well we know that some microRNAs bind to the mRNA for NOTCH which helps speed its degradation, meaning less NOTCH protein. One such microRNA is called miR-143-3p.

We also know that the brain contains a lncRNA called lncND (ND for Neural Development). The incredible elegance is that there is a primate specific insert in lncND which contains 16 (yes 16) binding sites for miR-143-3p. So lncND acts as a sponge for miR-143-3p meaning it can’t bind to the mRNA for NOTCH, meaning that there is more NOTCH around. Is this elegant or what. Let’s hear it for the Blind Watchmaker, assuming you have the faith to believe in such things.

Fortunately lncND is confined to the brain, otherwise we’d all be dead of cancer.

Should you want to read about this, here’s the reference [ Neuron vol. 90 pp. 1141 – 1143, 1255 – 1262 ’16 ] where there’s a lot more.

Historically, this was one of the criticisms of the Star Wars Missile Defense — the Russians wouldn’t send over a few missles, they’d send hundreds which would act as sponges to our defense. Whether or not attempting to put Star Wars in place led to Russia’s demise is debatable, but a society where it was a crime to own a copying machine, could never compete technically to produce such a thing.

Why our brains are large: the elegance of its molecular biology

Primates have much larger brains in proportion to their body size than other mammals. Here’s why. The mechanism is incredibly elegant. Unfortunately, you must put a sizable chunk of recent molecular biology under your belt before you can comprehend it. Anyone can listen to Mozart without knowing how to read or write music. Not so here.

I doubt that anyone can start from ground zero and climb all the way up, but here is all the background you need to comprehend what follows. Start here — https://luysii.wordpress.com/2010/07/07/molecular-biology-survival-guide-for-chemists-i-dna-and-protein-coding-gene-structure/
and follow the links (there are 5 more articles).

Also you should be conversant with competitive endogenous RNA (ceRNA) — here’s a link — https://luysii.wordpress.com/2014/01/20/why-drug-discovery-is-so-hard-reason-24-is-the-3-untranslated-region-of-every-protein-a-cerna/

Also you should understand what microRNAs are — we’re still discovering all the things they do — here’s the background you need — https://luysii.wordpress.com/2015/03/22/why-drug-discovery-is-so-hard-reason-26-were-discovering-new-players-all-the-time/weith.

Still game?

Now we must delve into the embryology of the brain, something few chemists or nonbiological type scientists have dealt with.

You’ve probably heard of the term ‘water on the brain’. This refers to enlargement of the ventricular system, a series of cavities in all our brains. In the fetus, all nearly all our neurons are formed from cells called neuronal precursor cells (NPCs) lining the fetal ventricle. Once formed they migrate to their final positions.

Each NPC has two choices — Choice #1 –divide into two NPCs, or Choice #2 — divide into an NPC and a daughter cell which will divide no further, but which will mature, migrate and become an adult neuron. So to get a big brain make NPCs adopt choice #1.

This is essentially a choice between proliferation and maturation. It doesn’t take many doublings of a NPC to eventually make a lot of neurons. Naturally cancer biologists are very interested in the mechanism of this choice.

Well to make a long story short, there is a protein called NOTCH — vitally important in embryology and in cancer biology which, when present, causes NPCs to make choice #1. So to make a big brain keep Notch around.

Well we know that some microRNAs bind to the mRNA for NOTCH which helps speed its degradation, meaning less NOTCH protein. One such microRNA is called miR-143-3p.

We also know that the brain contains a lncRNA called lncND (ND for Neural Development). The incredible elegance is that there is a primate specific insert in lncND which contains 16 (yes 16) binding sites for miR-143-3p. So lncND acts as a sponge for miR-143-3p meaning it can’t bind to the mRNA for NOTCH, meaning that there is more NOTCH around. Is this elegant or what. Let’s hear it for the Blind Watchmaker, assuming you have the faith to believe in such things.

Fortunately lncND is confined to the brain, otherwise we’d all be dead of cancer.

Should you want to read about this, here’s the reference [ Neuron vol. 90 pp. 1141 – 1143, 1255 – 1262 ’16 ] where there’s a lot more.

Historically, this was one of the criticisms of the Star Wars Missile Defense — the Russians wouldn’t send over a few missles, they’d send hundreds which would act as sponges to our defense. Whether or not attempting to put Star Wars in place led to Russia’s demise is debatable, but a society where it was a crime to own a copying machine, could never compete technically to produce such a thing.

Reproducibility and its discontents

“Since the launch of the clinicaltrials.gov registry in 2000, which forced researchers to preregister their methods and outcome measures, the percentage of large heart-disease clinical trials reporting significant positive results plummeted from 57% to a mere 8%”. I leave it to you to speculate why this happened, but my guess is that probably the data were sliced and diced until something of significance was found. I’d love to know what the comparable data is on anti-depressant trials. The above direct quote is from Proc. Natl. Acad. Sci. vol. 113 pp. 6454 – 6459 ’16. The article looked at the 100 papers published in ‘top’ psychology journals, about which much has been written — here’s the reference to the actual paper — Open Science Collaboration (2015) Psychology. Estimating the reproducibility of psychological science. Science 349(6251):aac4716.

The sad news is that only 39% of these studies were reproducible. So why beat a dead horse? The authors came up with something quite useful — they looked at how sensitive to context each of the 100 studies actually was. By context they mean the time of the study (e.g., pre- vs. post-Recession), culture (e.g., individualistic vs. collectivistic culture), the location (e.g., rural vs. urban setting), or the population (e.g., a racially diverse population vs. a predominantly White or Black or Latino population). Their conclusions were that the contextual sensitivity of the research topic was associated with replication success (e.g. the more context sensitive, the less likely it was that the study could be reproduced). This was even after statistically adjusting for several methodological characteristics (e.g., statistical power, effect size, etc. etc). The association between contextual sensitivity and replication success did not differ across psychological subdisciplines.

Addendum 15 June ’16 — Sadly, the best way to say this is — The more likely a study is to be true (replicable) the more likely it is to be not generally applicable (e.g. useful).

So this is good. Up to now the results of psychology studies have been reported in the press as of general applicability (particularly those which enforce the writer’s preferred narrative). Caveat emptor is two millenia old. Carl Sagan said it best — “Extraordinary claims require extraordinary evidence.”

For an example data slicing and dicing, please see — https://luysii.wordpress.com/2009/10/05/low-socioeconomic-status-in-the-first-5-years-of-life-doubles-your-chance-of-coronary-artery-disease-at-50-even-if-you-became-a-doc-or-why-i-hated-reading-the-medical-literature-when-i-had-to/

Mind the gap (junction that is)

Gap junctions don’t get much play in pharmacology, or even in neurology, where they are widespread in the central nervous system, linking neurons to neurons, astrocytes to astrocytes. They may get quite a bit more if blocking them is a way of treating metastatic disease (see later).

A bit of background if you’re unfamiliar with them. This is from my notes Molecular Biology of the Cell 4th Edition p. 1074

The gap junction is a cylindrical oligomer composed of 6 identical rod shaped subunits (called connexins). They have 4 transmembrane segments and two extracellular loops which contain a beta-strand structure (and which are an essential structural basis for the docking of the two connexons). Multiple connexons in a membrane tend to form hexagonal arrays.

The gap junction spans the lipid bilayer creating a channel along the central axis. The pore is made of two such protein hexamers one from each cell (called a hemichannel or a connexon) arranged end to end. Different tissues have different specific gap junction proteins (connexins). Man has 14 distinct connexins each encoded by a separate gene (20 homologous proteins in man PNAS 103 pp. 5213 – 5218 ’06). Most cell types express more than one. Connexins are capable of assembling into a heteromeric connexon Adjacent cells expressing different connexins can form intercellular channels in which the two aligned dihalf-channels are different. Each gap junction can contain a cluster of a few to MANY THOUSANDS of CONNEXONs.

Neuroscientists should be interested in them as they form a functional ‘synapse’ between cells, e.g. a way of transferring information between them. For the afficienado there will be much more at the end. To flog a nearly dead horse, this is yet another way a wiring diagram of the brain won’t help you understand it — gap junctions don’t show up when you’re looking at classic synapses. For details see https://luysii.wordpress.com/2011/04/10/would-a-wiring-diagram-of-the-brain-help-you-understand-it/

A recent paper in Nature implied that cancer cells can form gap junctions with astrocytes (a glial cell of the brain). Usually we think of gap junctions being of the same cell type, but not here apparently.

Then they describe a mechanism for the cancer cell tweak the astrocyte so it produces something enabling the cancer cell to survive. Here’s whqt they claim

[ Nature vol. 533 pp. 493 – 498 ’16 ] Human and mouse breast and lung cancer cells express protocadherin7 (PCDH7) whicboth promotes (how?) the assembly of carcinoma – astrocyte gap junctions made of connexin43. PCDH7 normally is only expressed in brain. It joints the stialyl transferase ST6GALNAC5 and neuroserpin as brain restricted proteins which metastastic cells from breast and lung cancer use to colonize the brain.

Metastastic cells then uswe the channels to transfer cGAMP to astrocytes activating the STING pathway, which results in InterferonAlpha (IFNalpha) and Tumor Necrosis Factor (TNF), paracrine signals. These activate STAT1 and NFkappaB in the metastatic cells, supporting tumor growth and chemoresistance.

Meclofenamate and tonabersat are ‘modulators’ of gap junctions, breaking the loop between metastatic cancer cell and the astrocyte. Adding them to the tissue culture studied in the paper, inhibited tumor growth. So here might be a way treat metastatic cancer — particularly since meclofenamate is an FDA approved generic drug available without a prescription.

I think the mechanism described above is incomplete — why should a tumor cell transfer something to another cell to have it secrete something which makes the original cell use something it already had.

Now for a few of the things gap junctions are doing in the brain.
****

[ Neuron vol. 90 pp. 810 – 823 ’16 ] ManhyGABAeric interneurons (are there other kinds?) IN VITRO are coupled by gap junctions. This work used dual patch clamp recordings of interneurons IN VIVO. They studied coupled cerebellar Golgi cells, and showed that, in the presence of spontaneous background synaptic activity, electrically coupled cerebellar Golgi cells showed robust milliSecond precision correlated activity. This was further enhanced by sensory stimulation.

The electrical coupling equlized membrane potential fluctuations, so that coupled neurons approach action potential threshold together. They say that something called spike triggered spikelets transmitted through gap junctions conditionally triggered postJunctional spikes, if both neurons were close to threshold.

Spikelets are brief low amplitude potentials which look like action potentials but which are much smaller. A spike cannot be generated without a much larger potential change than provided by a spikelet, because the spikelet voltage is too small to activate the ion channels of electrically excitable membranes.

So gap junctions controls the temporal precision and degree of both spontaneous and sensory evoked correlated activity betwen interneurons, by the cooperative effects of shared synaptic depolarization and spikelet transmission.

[ Neuron vol. 90 pp. 912 – 913, 1043 – 1056 ’16 ] It has been found that the strength of electrical coupling between neurons in a network is highly variable (even in the same neuron, so it could be coupled at different strengths with each of its partners). Site specific modulation of electrical coupling quickly reconfigures networks of electrically coupled neurons in the retina. Phosphorylation of connexin36 alters its conductivity.

The number of gap junctions determines the strength of ele tical coupling between cerebellar Golgi cells. Ultrastructural analysis shows that gap junctions vary widely in size, which also influences coupling strength (according to a computer simulation). These are dendro-dendritic electrical synapses (widespread in the brain between inhibitory interneurons).

Only 18% or so of the channels present at the gap junctions account for the boserved strength of electrical transmission between cerebellar golgi cells.

Somato-somatic junctions occur in the mammalian trigeminal mesencephalic nucleus. Could the excess junctions be acting as adhesion molecules.

In one system, the turnover of gap junction channel proteins is rapid and comparable with that of glutamic acid receptors.

Gap junctions are ‘low pass filters’ (they pass slow fluctuations of membrane potential better than they pass rapid fluctuations). This is why the electrical synapses are inhibitory — each action potential from a Golgi cell consists of a rapid (but brief) depolarizing spike followed by a relatively deep and protracted afterhyperpolarization — which is 200 times longer than the spike — and transmitted much more effectively.

Inhibition by sparse excitatory input breaks up Golgi network synchronization, because the coupling to adjacent cells is different for each one, causing dispersion of the spikes.

In quietly attentive animals cerebellar Golgi cells generate rhythmic synchronous activity at 8 Hertz. The same behavior is seen in cerebellar slices. The hyperpolarizing electrical post-synaptic potentials (PSPs) are the only synchronizing force. This is the default state, but it can be disrupted by a variety of sensory stimuli (or by movements) which reduce spiking frequency and rhythmicity.

Golgi cells can inhibit thousands of granule cells, and every granule cell gets inhibitory input from 4 – 8 Golgi cells. The transient nature of network desynchronization ‘could’ allow the cerebellar input layer to act as a timing device over the 10 milliSecond to 1 second timescale.

Spot the flaw

Mathematical talent varies widely. It was a humbling thing a few years ago to sit in an upper level college math class on Abstract Algebra with a 16 year old high school student taking the course, listening with one ear while he did his German homework. He was later a double summa in both math and physics at Yale. So do mathematicians think differently? A fascinating paper implies that they use different parts of their brain doing math than when not doing it. The paper has one methodological flaw — see if you can find it.

[ Proc. Natl. Acad. Sci. vol. 113 pp. 4887 – 4889, 4909 – 4917 ’16 ] 15 expert mathematicians and 15 nonMathematicians with comparable academic qualifications were studied (3 literature, 3 history 1 philosophy 2 linguistics, 1 antiquity, 3 graphic arts and theater 1 communcation, 1 heritage conservation — fortunately no feminist studies). They had to listen to mathematical and nonMathematical statements and decide true false or meaningless. The nonMathematical statements referred to general knowledge of nature and history. All this while they were embedded in a Magnetic Resonance Imager, so that functional MRI (fMRI) could be performed.

In mathematicians there was no overlap of the math responsive network (e.g. the areas of the brain activated by doing math) with the areas activated by sentence comprehension and general semantic knowledge.

The same brain networks were activated by all types of mathematical statement (geometry, analysis, algebra and topology) as opposed to nonMathematical statement. The areas activated were the dorsal parietal, ventrolateral temporal and bilateral frontal. This was only present in the expert mathematicians (and only to mathematical statements) These areas are outside those associated with language (inferior frontal gyrus of the left hemisphere). The activated areas are also involved in visual processing of arabic numbers and simple calculation. The activated areas in mathematicians were NOT those related to language or general knowledge.

So what’s wrong with the conclusion? The editorialist (pp. 4887 – 4889) pointed this out but I thought of it independently.

All you can say is that experts working in their field of expertise use different parts of their brain than they use for general knowledge. The nonMathematicians should have been tested in their field of expertise. Shouldn’t be hard to do.

NonAlgorithmic Intelligence

Penrose was right. Human intelligence is nonAlgorithmic. But that doesn’t mean that our physical brains produce consciousness and intelligence using quantum mechanics (although all matter is what it is because of quantum mechanics). The parts (even small ones like neurotubules) contain so much mass that their associated wavefunction is too small to exhibit quantum mechanical effects. Here Penrose got roped in by Kauffman thinking that neurotubules were the carriers of the quantum mechanical indeterminacy. They aren’t, they are just too big. The dimer of alpha and beta tubulin contains 900 amino acids — a mass of around 90,000 Daltons (or 90,000 hydrogen atoms — which are small enough to show quantum mechanical effects).

So why was Penrose right? Because neural nets which are inherently nonAlgorithmic are showing intelligent behavior. AlphaGo which beat the world champion is the most recent example, but others include facial recognition and image classification [ Nature vol. 529 pp. 484 – 489 ’16 ].

Nets are trained on real world images and told whether they are right or wrong. I suppose this is programming of a sort, but it is certainly nonAlgorithmic. As the net learns from experience it adjusts the strength of the connections between its neurons (synapses if you will).

So it should be a simple matter to find out just how AlphaGo did it — just get a list of the neurons it contains, and the number and strengths of the synapses between them. I can’t find out just how many neurons and connections there are, but I do know that thousands of CPUs and graphics processors were used. I doubt that there were 80 billion neurons or a trillion connections between them (which is what our brains are currently thought to have).

Just print out the above list (assuming you have enough paper) and look at it. Will you understand how AlphaGo won? I seriously doubt it. You will understand it less well than looking at a list of the positions and momenta of 80 billion gas molecules will tell you its pressure and temperature. Why? Because in statistical mechanics you assume that the particles making up an ideal gas are featureless, identical and do not interact with each other. This isn’t true for neural nets.

It also isn’t true for the brain. Efforts are underway to find a wiring diagram of a small area of the cerebral cortex. The following will get you started — https://www.quantamagazine.org/20160406-brain-maps-micron-program-iarpa/

Here’s a quote from the article to whet your appetite.

“By the end of the five-year IARPA project, dubbed Machine Intelligence from Cortical Networks (Microns), researchers aim to map a cubic millimeter of cortex. That tiny portion houses about 100,000 neurons, 3 to 15 million neuronal connections, or synapses, and enough neural wiring to span the width of Manhattan, were it all untangled and laid end-to-end.”

I don’t think this will help us understand how the brain works any more than the above list of neurons and connections from AlphaGo. There are even more problems with such a list. Connections (synapses) between neurons come and go (and they increase and decrease in strength as in the neural net). Some connections turn on the receiving neuron, some turn it off. I don’t think there is a good way to tell what a given connection is doing just by looking a a slice of it under the electron microscope. Lastly, some of our most human attributes (emotion) are due not to connections between neurons but due to release of neurotransmitters generally into the brain, not at the very localized synapse, so it won’t show up on a wiring diagram. This is called volume neurotransmission, and the transmitters are serotonin, norepinephrine and dopamine. Not convinced? Among agents modifying volume neurotransmission are cocaine, amphetamine, antidepressants, antipsychotics. Fairly important.

So I don’t think we’ll ever truly understand how the neural net inside our head does what it does.

Is a rational treatment for Multiple Sclerosis in our future?

Two very recent papers taken together point the way to a rational treatment of multiple sclerosis (and probably all autoimmune disease). The short story:
Paper #1 found a way to find the antigen or antigens patients with MS are reacting to
Paper #2 found a way to selectively impair the response to an inciting antigen without clobbering the whole immune system

Some history: Some evening in 1966 or 1967 a fellow neurology resident and I were sitting on the ward having dealt with the complications of high doses corticosteroids for a case of optic neuritis (often the first sign of MS). I said, some day they’ll look at what we’re doing the way we look at docs of 200 years ago using leeches (and bloodletting). As a kid, I remember my parents driving into Philly. Shortly after getting over the Ben Franklin bridge we’d pass a pharmacy offering leeches on its sign.

It was obvious even back then that MS in some way was an attack by the immune system on the brain. Finding the particular antigen the system was reacting to would lead us to the cause and hopefully less simplistic treatment than clobbering the immune system. We didn’t know all the proteins we had or even how many, so people would look for antibodies to a variety causes (which they’d arrived at by reasoning, not data). Increased antibody titers to a variety of viruses were found, but that led nowhere. No one ever isolated a virus from MS brain, although sightings on electron microscopy were eagerly reported. Eventually it became obvious that the immune system was on high alert with increased antibodies to lots of things.

This leads to paper #1 [ Proc. Natl. Acad. Sci. vol. 113 pp. 2188 – 2193 ’16 ] To make a long story short they used something called the Human Protein Atlas Program to find what proteins the antibodies in MS patients were reacting to. So rather than having a theory about what MS patients might be reacting to and testing it, they looked at all proteins and watched. It’s the difference between being a Greek philosopher reasoning things out from first principles and collecting data. Only when the technology is available can you stop a priori theorizing and just look. Don’t be too hard on the earlier researchers, they didn’t have the tools.

The found that MS patients were reacting to a protein called anoctamin2, which actually showed increased expression near and inside the demyelinating plaques of MS.

For the gory details keep reading, otherwise skip to **** where I’ll discuss paper #2

Gory details — The Human Protein Atlas produces human protein fragments, selected on the basis of their low similarity to other proteins in the proteome. [ Science vol. 347 1260419 (23 Jan) ’15 ] The atlas hopes to find out where and how much of each our proteins is at the tissue and cellular level. It is based on antibody based profiling on tissue microarrays (of proteins?). This based on transcript expression (RNA-Seq), and immunohistochemistry (24,028 antibodies coresponding to 16,975 protein coding genes). 44 tissues were studied. The antibodies produced more than 13 million tissue based mmunohistochemistry images. They also report subproteomes (secreted proteins n = 3,171, and membrane bound proteins n = 5,570). Interstingly there was an overall concurrence between mRNA and protein levels for a given gene product across various tissues.

The PNAS paper profiled 2,169 plasma samples from MS cases and population based controls (with neurologic disease) using bead arrays built with 384 human protein fragments seleted from an initial screening with 11,520 antigens. There was increased reactivity to anoctamin2 (aka TMEM16B) in MS vs. controls (by how much?). This was corroborated in independent assay with alternative protein constructs and by epitope mapping with peptides covering the identified region of anoctamin 2.

ImmunoFLuorescence in human MS brain tissue showed increased anoctamin2 expression as small cellular aggregates near and inside MS lesions. The controls had other neurologic disease. There was a 5.3 fold change in fluorescence intensity in the MS group. The antibodies are directed against the amino terminal region.

*****

*****

Paper #2 — [ Nature vol. 530 pp. 422 – 423, 434 – 440 ’16 ] basically found a way to knock out the immune system’s response to a single antigen — not all of them. The point is that just an antigen by itself isn’t enough to turn the immune system on. A costimulatory molecule must also be present on the antigen presenting cell. If it isn’t there the immune system is actually turned off by forming regulatory T cells (which even though they are part of the immune system they actually turn it off).

One can form models of human autoimmune disease in mice. Two such are EAE (Experimental Allergic Encephalomyelitis) formed by giving the animal myelin basic protein (a constituent of myelin which is attacked in MS), and rheumatoid arthritis (formed by giving collagen to the animals). What is so great about this paper is that MHC II carrying peptides from collagen suppress disease in a mouse model of rheumatoid arthritis, but NOT in mice with EAE. MHC-II carrying CNS antigen peptides control EAE but not collagen induced arthritis.. In addition neither treatment impaired the immune response to infection — something that almost always happens when you clobber the immune system.

Well it’s a long way from the lab to the bedside, but imagine finding what the immune system is reacting to and stopping it (without stopping the immune system). That’s what these two papers portend. Exciting times.

Nicastrin the gatekeeper of gamma secretase

Once a year some hapless trucker from out of town gets stuck trying to drive under a nearby railroad bridge with a low clearance. This is exactly the function of nicastrin in the gamma secretase complex which produces the main component of the senile plaque, the aBeta peptide.

Gamma secretase is a 4 protein complex which functions as an enzyme which can cut the transmembrane segment of proteins embedded in the cell membrane. This was not understood for years, as cutting a protein here means hydrolyzing the amide bond of the protein, (e.g. adding water) and there is precious little water in the cell membrane which is nearly all lipid.

Big pharma has been attacking gamma secretase for years, as inhibiting it should stop production of the Abeta peptide and (hopefully) help Alzheimer’s disease. However the paper to be discussed [ Proc. Natl. Acad. Sci. vol. 113 p.n E509 – E518 ’16 ] notes that gamma secretase processes ‘scores’ of cell membrane proteins, so blanket inhibition might be dangerous.

The idea that Nicastrin is the gatekeeper for gamma secretase is at least a decade old [ Cell vol. 122 pp. 318 – 320 ’05 ], but back then people were looking for specific binding of nicastrin to gamma secretase targets.

The new paper provides a much simpler explanation. It won’t let any transmembrane segment of a protein near the active site of gamma secretase unless the extracellular part is lopped off. The answer is simple mechanics. Nicastrin is large (709 amino acids) but with just one transmembrane domain. Most of it is extracellular forming a blob extending out 25 Angstroms from the membrane, directly over the substrate binding pocket of gamma secretase. Only substrates with small portions outside the membrane (ectodomains) can pass through it. It’s the railroad bridge mentioned above. Take a look at the picture — https://en.wikipedia.org/wiki/Nicastrin

This is why a preliminary cleavage of the Amyloid Precursor Peptide (APP) is required for gamma secretase to work.

So all you had to do was write down the wavefunction for Nicastrin (all 709 amino acids) and solve it (assuming you even write it down) and you’d have the same answer — NOT. Only the totally macroscopic world explanation (railroad bridge) is of any use. What keeps proteins from moving through each other? Van der Waals forces. What help explain them. The Pauli exclusion principle, as pure quantum mechanics as it gets.

Follow

Get every new post delivered to your Inbox.

Join 93 other followers