Neuroscience Fails Stats 101?

According to a new paper, a full half of neuroscience papers that try to do a (very simple) statistical comparison are getting it wrong: Erroneous analyses of interactions in neuroscience: a problem of significance.

Here's the problem. Suppose you want to know whether a certain 'treatment' has an affect on a certain variable. The treatment could be a drug, an environmental change, a genetic variant, whatever. The target population could be animals, humans, brain cells, or anything else.

So you give the treatment to some targets and give a control treatment to others. You measure the outcome variable. You use a t-test of significance to see whether the effect is large enough that it wouldn't have happened by chance. You find that it was significant.

That's fine. Then you try a different treatment, and it doesn't cause a significant effect against the control. Does that mean the first treatment was more powerful than the second?

No. It just doesn't. The only way to find that out would be to compare the two treatments directly - and that would be very easy to do, because you have all the data to hand. If you just compare the two treatments to control you might end up with this scenario:

Both treatments are very similar but one (B) is slightly better so it's significantly different from control, while A isn't. But they're basically the same. It's probably just fluke that B did slightly better than A. If you compared A and B directly you'd find they were not significantly different.

An analogy: Passing a significance test is like winning a prize. You can only do it if you're much better than the average. But that doesn't mean you're much better than everyone who didn't win the prize, because some of them will have almost been good enough.

Usain Bolt is the fastest man in the world (when he's not false-starting himself out of races). Much faster than me. But he's not much faster than the second fastest man in the world.




Nieuwenhuis S, Forstmann BU, & Wagenmakers EJ (2011). Erroneous analyses of interactions in neuroscience: a problem of significance. Nature neuroscience, 14 (9), 1105-7 PMID: 21878926

Neuroscience Fails Stats 101?

According to a new paper, a full half of neuroscience papers that try to do a (very simple) statistical comparison are getting it wrong: Erroneous analyses of interactions in neuroscience: a problem of significance.

Here's the problem. Suppose you want to know whether a certain 'treatment' has an affect on a certain variable. The treatment could be a drug, an environmental change, a genetic variant, whatever. The target population could be animals, humans, brain cells, or anything else.

So you give the treatment to some targets and give a control treatment to others. You measure the outcome variable. You use a t-test of significance to see whether the effect is large enough that it wouldn't have happened by chance. You find that it was significant.

That's fine. Then you try a different treatment, and it doesn't cause a significant effect against the control. Does that mean the first treatment was more powerful than the second?

No. It just doesn't. The only way to find that out would be to compare the two treatments directly - and that would be very easy to do, because you have all the data to hand. If you just compare the two treatments to control you might end up with this scenario:

Both treatments are very similar but one (B) is slightly better so it's significantly different from control, while A isn't. But they're basically the same. It's probably just fluke that B did slightly better than A. If you compared A and B directly you'd find they were not significantly different.

An analogy: Passing a significance test is like winning a prize. You can only do it if you're much better than the average. But that doesn't mean you're much better than everyone who didn't win the prize, because some of them will have almost been good enough.

Usain Bolt is the fastest man in the world (when he's not false-starting himself out of races). Much faster than me. But he's not much faster than the second fastest man in the world.




Nieuwenhuis S, Forstmann BU, & Wagenmakers EJ (2011). Erroneous analyses of interactions in neuroscience: a problem of significance. Nature neuroscience, 14 (9), 1105-7 PMID: 21878926

Brain Connectivity, Or Head Movement?

"It's pretty painless. Basically you just need to lie there and make sure you don't move your head".


This is what I say to all the girls... who are taking part in my fMRI studies. Head movement is a big problem in fMRI. If your head moves, your brain moves and all fMRI analysis assumes that the brain is perfectly still. Although head movement correction is now a standard part of any analysis software, it's not perfect.

It may be a particular problem in functional connectivity studies, which attempt to measure the degree to which different parts of the brain are "talking" to each other, in terms of correlated neural activity over time. These are extremely popular nowadays. It's even been claimed that this data may help us understand consciousness itself (although we've heard that before).

A new paper offers some important words of caution. It shows that head motion affects estimates of functional connectivity. The more motion, the weaker the measured connectivity in long-range networks, while shorter range connections were stronger. Also, men tended to move more than women.

The effect was small - head movement can't explain more than a small fraction of the variability in connectivity.


The authors looked at 1,000 scans from healthy volunteers. They just had to lie in the scanner at rest. They looked at functional connectivity, using standard "motion correction" methods, and correlated it with head movement (which you can measure very accurately from the MRI images themselves.) Men tended to move more than women. Could this explain why women tend to have higher functional connectivity?

Disconcertingly, head movement was associated with low long range / high short range connections, which is exactly what's been proposed to happen in autism (although in fairness, not all the evidence for this comes from fMRI).

This clearly doesn't prove that the autism studies are all dodgy, but it's an issue. People with autism, and people with almost any mental or physical disorder, on average tend to move more than healthy controls.

One caveat. Could it be that brain activity causes head movement, rather than the reverse? The authors don't consider this. Head movement must come from the brain, of course. Probably from the motor cortex. The fact that motor cortex functional connectivity was positively associated with movement does suggest a possible link.

However, this paper still ought to make anyone who's using functional connectivity worry - at least a little.

Head motion is a particularly insidious confound. It is insidious because it biases between-group studies often in the direction of the hypothesized difference....even though there is considerable variation that is not due to head motion, in any given instance, a between-group difference could be entirely due to motion.

Van Dijk, K., Sabuncu, M., & Buckner, R. (2011). The Influence of Head Motion on Intrinsic Functional Connectivity MRI NeuroImage DOI: 10.1016/j.neuroimage.2011.07.044

Brain Connectivity, Or Head Movement?

"It's pretty painless. Basically you just need to lie there and make sure you don't move your head".


This is what I say to all the girls... who are taking part in my fMRI studies. Head movement is a big problem in fMRI. If your head moves, your brain moves and all fMRI analysis assumes that the brain is perfectly still. Although head movement correction is now a standard part of any analysis software, it's not perfect.

It may be a particular problem in functional connectivity studies, which attempt to measure the degree to which different parts of the brain are "talking" to each other, in terms of correlated neural activity over time. These are extremely popular nowadays. It's even been claimed that this data may help us understand consciousness itself (although we've heard that before).

A new paper offers some important words of caution. It shows that head motion affects estimates of functional connectivity. The more motion, the weaker the measured connectivity in long-range networks, while shorter range connections were stronger. Also, men tended to move more than women.

The effect was small - head movement can't explain more than a small fraction of the variability in connectivity.


The authors looked at 1,000 scans from healthy volunteers. They just had to lie in the scanner at rest. They looked at functional connectivity, using standard "motion correction" methods, and correlated it with head movement (which you can measure very accurately from the MRI images themselves.) Men tended to move more than women. Could this explain why women tend to have higher functional connectivity?

Disconcertingly, head movement was associated with low long range / high short range connections, which is exactly what's been proposed to happen in autism (although in fairness, not all the evidence for this comes from fMRI).

This clearly doesn't prove that the autism studies are all dodgy, but it's an issue. People with autism, and people with almost any mental or physical disorder, on average tend to move more than healthy controls.

One caveat. Could it be that brain activity causes head movement, rather than the reverse? The authors don't consider this. Head movement must come from the brain, of course. Probably from the motor cortex. The fact that motor cortex functional connectivity was positively associated with movement does suggest a possible link.

However, this paper still ought to make anyone who's using functional connectivity worry - at least a little.

Head motion is a particularly insidious confound. It is insidious because it biases between-group studies often in the direction of the hypothesized difference....even though there is considerable variation that is not due to head motion, in any given instance, a between-group difference could be entirely due to motion.

Van Dijk, K., Sabuncu, M., & Buckner, R. (2011). The Influence of Head Motion on Intrinsic Functional Connectivity MRI NeuroImage DOI: 10.1016/j.neuroimage.2011.07.044

The Brain Is Not Made of DNA

A new paper claims to have found A novel functional brain imaging endophenotype of autism.
They used fMRI to show that the brains of teenagers with autism showed no activation differences to looking at smiling happy faces, or afraid faces, compared to unemotional ones. In teens without autism, there was strong activation in many emotional and face-related brain regions. The unaffected brothers and sisters of the autistic people showed intermediate effects.

This is a fine study. The finding that siblings of people with autism have weakened neural responses to emotional faces is quite important as it suggests that this finding correlates (to some degree) with your position on the autism "spectrum".

The abstract of the paper actually downplays this, and says "The response in unaffected siblings did not differ significantly from the response in autism". However, there was a significant linear trend of group, and looking at the graphs, it's clear the siblings were In The Middle, like Malcolm.


There's plenty more nice things you could do with these results, which is an unusally large and rich dataset (120 people - 40 in each group). You could see, for example, whether siblings tend to be similar in terms of neural response. You could see whether the siblings who are most alike in brain response, are closest in symptoms. Or just look a the structural data on brain size and shape to see if there are characteristic differences between siblings that make one of the autistic and the other not.

There are a few problems. Most of the analyses are subject to the non-independence problem, because they defined their regions of interest based on the areas that showed a significant happy vs neutral face effect in the control group. So it's no surprise that when they generated graphs from these areas, the control group showed the strongest effect. However, they also do whole-brain analyses which avoid this problem and I don't think it undermines the main results.

So it's a decent study. But is this a "biomarker", or "endophenotype", as the title of the paper has it?

These are both hot topics in neuroscience at the moment. As the authors put it (emphasis mine):

An endophenotype is a heritable feature associated with a condition, present in affected individuals regardless of whether their condition is manifested, which co-segregates with the condition in families and which is present in unaffected family members at a higher rate than in the general population.

In such family members, endophenotypes represent instances in which genes associated with a particular condition exert measurable effects in individuals in whom they are insufficient to cause the condition itself...

The promise of characterizing endophenotypes lies in their hypothesized intermediate position between genotype and phenotype... the etiology of the endophenotype is likely to be correspondingly simpler: it can be said to be ‘closer to the level of gene action’.

The idea, in other words, is that if we can find a difference in the brains of people with autism, and their unaffected relatives who (presumably) share some of the same genes, we might have found a mechanism by which the genes ultimately cause the symptoms.

It might be easier, then, to find the genes for brain-not-lighting-up-to-happy-faces, than it will be to find genes for autism. Then once we've found those, we can use them to better understand autism.

My concern is that, while in theory endophenotypes seem "closer to the genetics" because they're "biological" rather than "behavioural", this is just a philosophical illusion based on the idea that the mind is not the brain.

We actually have no idea whether brain-not-lighting-up-to-happy-faces is closer to genetics than autistic behaviour. I'd say that our default assumption should be that everything is exactly the same "distance" from DNA, that is to say, everything is the product of complex interactions between genes and environment.

Some things are under the more or less exclusive control of a small number of genes, and these are called "genetic", but it's important not to assume that just because something's "in the brain", it's probably "more genetic" in this sense. The brain is a product of the environment as well.

If you scanned my brain while playing an audio recording of Urda love poetry, not much would happen. I don't know Urdu. In someone who did speak Urdu, all kinds of language and emotional areas would light up. That doesn't mean Urdu-brain-response is genetic. It's exactly as genetic as speaking-Urdu, which isn't genetic.

Spencer, M., Holt, R., Chura, L., Suckling, J., Calder, A., Bullmore, E., & Baron-Cohen, S. (2011). A novel functional brain imaging endophenotype of autism: the neural response to facial expression of emotion Translational Psychiatry, 1 (7) DOI: 10.1038/tp.2011.18

The Brain Is Not Made of DNA

A new paper claims to have found A novel functional brain imaging endophenotype of autism.
They used fMRI to show that the brains of teenagers with autism showed no activation differences to looking at smiling happy faces, or afraid faces, compared to unemotional ones. In teens without autism, there was strong activation in many emotional and face-related brain regions. The unaffected brothers and sisters of the autistic people showed intermediate effects.

This is a fine study. The finding that siblings of people with autism have weakened neural responses to emotional faces is quite important as it suggests that this finding correlates (to some degree) with your position on the autism "spectrum".

The abstract of the paper actually downplays this, and says "The response in unaffected siblings did not differ significantly from the response in autism". However, there was a significant linear trend of group, and looking at the graphs, it's clear the siblings were In The Middle, like Malcolm.


There's plenty more nice things you could do with these results, which is an unusally large and rich dataset (120 people - 40 in each group). You could see, for example, whether siblings tend to be similar in terms of neural response. You could see whether the siblings who are most alike in brain response, are closest in symptoms. Or just look a the structural data on brain size and shape to see if there are characteristic differences between siblings that make one of the autistic and the other not.

There are a few problems. Most of the analyses are subject to the non-independence problem, because they defined their regions of interest based on the areas that showed a significant happy vs neutral face effect in the control group. So it's no surprise that when they generated graphs from these areas, the control group showed the strongest effect. However, they also do whole-brain analyses which avoid this problem and I don't think it undermines the main results.

So it's a decent study. But is this a "biomarker", or "endophenotype", as the title of the paper has it?

These are both hot topics in neuroscience at the moment. As the authors put it (emphasis mine):

An endophenotype is a heritable feature associated with a condition, present in affected individuals regardless of whether their condition is manifested, which co-segregates with the condition in families and which is present in unaffected family members at a higher rate than in the general population.

In such family members, endophenotypes represent instances in which genes associated with a particular condition exert measurable effects in individuals in whom they are insufficient to cause the condition itself...

The promise of characterizing endophenotypes lies in their hypothesized intermediate position between genotype and phenotype... the etiology of the endophenotype is likely to be correspondingly simpler: it can be said to be ‘closer to the level of gene action’.

The idea, in other words, is that if we can find a difference in the brains of people with autism, and their unaffected relatives who (presumably) share some of the same genes, we might have found a mechanism by which the genes ultimately cause the symptoms.

It might be easier, then, to find the genes for brain-not-lighting-up-to-happy-faces, than it will be to find genes for autism. Then once we've found those, we can use them to better understand autism.

My concern is that, while in theory endophenotypes seem "closer to the genetics" because they're "biological" rather than "behavioural", this is just a philosophical illusion based on the idea that the mind is not the brain.

We actually have no idea whether brain-not-lighting-up-to-happy-faces is closer to genetics than autistic behaviour. I'd say that our default assumption should be that everything is exactly the same "distance" from DNA, that is to say, everything is the product of complex interactions between genes and environment.

Some things are under the more or less exclusive control of a small number of genes, and these are called "genetic", but it's important not to assume that just because something's "in the brain", it's probably "more genetic" in this sense. The brain is a product of the environment as well.

If you scanned my brain while playing an audio recording of Urda love poetry, not much would happen. I don't know Urdu. In someone who did speak Urdu, all kinds of language and emotional areas would light up. That doesn't mean Urdu-brain-response is genetic. It's exactly as genetic as speaking-Urdu, which isn't genetic.

Spencer, M., Holt, R., Chura, L., Suckling, J., Calder, A., Bullmore, E., & Baron-Cohen, S. (2011). A novel functional brain imaging endophenotype of autism: the neural response to facial expression of emotion Translational Psychiatry, 1 (7) DOI: 10.1038/tp.2011.18

Eagle-Eyed Autism? No.

An interesting and refreshing paper from Simon Baron-Cohen's autism group from Cambridge. The results themselves are pretty boring - they found that people with autism have normal visual acuity.


But the story behind it is rather spicy.

Back in 2009, a Cambridge group - different authors, but led by "SBC", published a report claiming that people with autism have exceptionally acute vision. Their average visual acuity was claimed to be 2.8

On this scale, 1.0 is defined as normal, and a sharp-eyed young adult with excellent eyesight would get about 1.5. 2.8 means nearly three times as good. Which is, literally, superhuman - a bird of prey would be happy with that. The paper was titled "Eagle-Eyed Visual Acuity In Autism".

However, what followed was straight out of the Book of Obadiah - "Though you soar like the eagle ... from there I will bring you down, sayeth the Lord". Or in this case, sayeth two experts in visual acuity research, Bach and Dakin, whose qualifications included the fact that they wrote the software used in the original study, which is online here.

They wrote a knock-down critique, arguing that the results were a result of using the wrong settings, which meant that the task was extremely easy. In fact, even perfect performance would only correspond to an acuity of less than 1.

You could never make a test so hard that it would require an acuity of 3.0 on a standard computer. Pixels are just too big. A single pixel is easy to spot, for someone of normal-ish vision. The only way to make it harder would be to use a special, extremely high-res monitor, or to get people to sit a long way from the screen.

So how did a result of nearly 3.0 come out? Because they also turned on data extrapolation, basically saying that if you really aced the easy task, you'd probably do quite well on a harder one. This might be sensible in some situations, but it breaks down when the task was so easy. The autistics seemed to have super vision because they got, say, 99% right, as opposed to 98%.

Yet the present paper represents a happy ending as it's written by a combined team of Cambridge people, and Bach and Dakin as well, although the lead authors of the original weren't on it. This time, they used appropriate methods - they got people to sit 4 meters from the screen. To be extra sure, they also gave everyone an eye exam before testing.

And they found no difference at all. The present paper is heartening - rather than grimly sticking to their guns, they admitted their error.


This story should however serve as a cautionary tale; I previously wrote about the fact that in science, a little mistake can cause a lot of problems. This is one of those cases, although arguably there were two seperate mistakes, but one, the extrapolation, was only a problem because of the main mistake, the big pixels.

Tavassoli T, Latham K, Bach M, Dakin SC, & Baron-Cohen S (2011). Psychophysical measures of visual acuity in autism spectrum conditions. Vision research PMID: 21704058

Eagle-Eyed Autism? No.

An interesting and refreshing paper from Simon Baron-Cohen's autism group from Cambridge. The results themselves are pretty boring - they found that people with autism have normal visual acuity.


But the story behind it is rather spicy.

Back in 2009, a Cambridge group - different authors, but led by "SBC", published a report claiming that people with autism have exceptionally acute vision. Their average visual acuity was claimed to be 2.8

On this scale, 1.0 is defined as normal, and a sharp-eyed young adult with excellent eyesight would get about 1.5. 2.8 means nearly three times as good. Which is, literally, superhuman - a bird of prey would be happy with that. The paper was titled "Eagle-Eyed Visual Acuity In Autism".

However, what followed was straight out of the Book of Obadiah - "Though you soar like the eagle ... from there I will bring you down, sayeth the Lord". Or in this case, sayeth two experts in visual acuity research, Bach and Dakin, whose qualifications included the fact that they wrote the software used in the original study, which is online here.

They wrote a knock-down critique, arguing that the results were a result of using the wrong settings, which meant that the task was extremely easy. In fact, even perfect performance would only correspond to an acuity of less than 1.

You could never make a test so hard that it would require an acuity of 3.0 on a standard computer. Pixels are just too big. A single pixel is easy to spot, for someone of normal-ish vision. The only way to make it harder would be to use a special, extremely high-res monitor, or to get people to sit a long way from the screen.

So how did a result of nearly 3.0 come out? Because they also turned on data extrapolation, basically saying that if you really aced the easy task, you'd probably do quite well on a harder one. This might be sensible in some situations, but it breaks down when the task was so easy. The autistics seemed to have super vision because they got, say, 99% right, as opposed to 98%.

Yet the present paper represents a happy ending as it's written by a combined team of Cambridge people, and Bach and Dakin as well, although the lead authors of the original weren't on it. This time, they used appropriate methods - they got people to sit 4 meters from the screen. To be extra sure, they also gave everyone an eye exam before testing.

And they found no difference at all. The present paper is heartening - rather than grimly sticking to their guns, they admitted their error.


This story should however serve as a cautionary tale; I previously wrote about the fact that in science, a little mistake can cause a lot of problems. This is one of those cases, although arguably there were two seperate mistakes, but one, the extrapolation, was only a problem because of the main mistake, the big pixels.

Tavassoli T, Latham K, Bach M, Dakin SC, & Baron-Cohen S (2011). Psychophysical measures of visual acuity in autism spectrum conditions. Vision research PMID: 21704058

Machine-Readable Psychiatry

The idea of trawling the internet to discover what people think about medications is a fascinating one and I've covered some attempts to do this in the past, but it's not easy. And there's something worrying about where it could lead.

A new paper aims to trawl medical records to work out how well depressed patients responded to treatment. The authors used Natural Language Processing or NLP (not that NLP) to interpret electronic medical records from over 5,000 patients treated at hospitals in New England. Each record included notes taken at multiple visits.

A crack team of "three experienced board-certified clinical psychiatrists" reviewed the notes and provided a "Gold Standard" classification as to whether patients were Depressed, Recovered or Intermediate at each visit. The problem here is that they didn't actually see the patients, they just had the notes. If the notes were bad, the result will have been bad too. Garbage In, Garbage Out. Even if you then put a big gold medal on the garbage.

They then found that an NLP algorithm was able to learn how to duplicate the expert opinion, based on the words used in the notes. Using a machine learning approach they were able to teach the computer that if the text contained the word "depressed", it was a sign that the patient was depressed while "much better" was associated with being... guess.

In fairness, it's not a bad attempt to turn text into numbers, and in future it could allow you to do interesting things such as comparing two drugs in terms of which ones make people "much better".

I'm concerned about this though. The essence of the original, narrative notes, is that they contain individual information about that patient's story. You could go through them with a computer and calculate what happens to the average patient given a certain drug. That might be useful information. But if you did that as a replacement for reading about individual patients, you'd be missing the whole point of the narrative notes.

Worse, as this kind of thing becomes feaisble it will feed back on itself and encourage clinicians to write their notes -and therefore to think, inevitably - in machine-readable terms. The authors suggest as much:

As more health care systems move to electronic medical records, there is a unique opportunity to better quantify outcomes. For example, the 16-item patient-rated QIDS-SR [questionairre] has been shown to be highly correlated with clinician-rated measures and sensitive to treatment effects.... At minimum, EMR systems that utilize templates could require clinicians to record a clinical status for example, using the 7-point Clinical Global Impression scale...

Indeed, many say that this already is happening. Now quantification is generally a good thing I think, but only so long as it's an aid to understanding, not a replacement for it.

Yet quantification often does become a replacement for understanding because there's a trap that we face when trying to deal with a complicated set of information. The temptation is to focus on the easiest bit to measure, because that's easy, and then assume that this represents the state of the whole thing. The reason that something's easy to measure is often because it doesn't capture the whole phenomena.

Perlis RH, Iosifescu DV, Castro VM, Murphy SN, Gainer VS, Minnier J, Cai T, Goryachev S, Zeng Q, Gallagher PJ, Fava M, Weilburg JB, Churchill SE, Kohane IS, & Smoller JW (2011). Using electronic medical records to enable large-scale studies in psychiatry: treatment resistant depression as a model. Psychological medicine, 1-10 PMID: 21682950

Machine-Readable Psychiatry

The idea of trawling the internet to discover what people think about medications is a fascinating one and I've covered some attempts to do this in the past, but it's not easy. And there's something worrying about where it could lead.

A new paper aims to trawl medical records to work out how well depressed patients responded to treatment. The authors used Natural Language Processing or NLP (not that NLP) to interpret electronic medical records from over 5,000 patients treated at hospitals in New England. Each record included notes taken at multiple visits.

A crack team of "three experienced board-certified clinical psychiatrists" reviewed the notes and provided a "Gold Standard" classification as to whether patients were Depressed, Recovered or Intermediate at each visit. The problem here is that they didn't actually see the patients, they just had the notes. If the notes were bad, the result will have been bad too. Garbage In, Garbage Out. Even if you then put a big gold medal on the garbage.

They then found that an NLP algorithm was able to learn how to duplicate the expert opinion, based on the words used in the notes. Using a machine learning approach they were able to teach the computer that if the text contained the word "depressed", it was a sign that the patient was depressed while "much better" was associated with being... guess.

In fairness, it's not a bad attempt to turn text into numbers, and in future it could allow you to do interesting things such as comparing two drugs in terms of which ones make people "much better".

I'm concerned about this though. The essence of the original, narrative notes, is that they contain individual information about that patient's story. You could go through them with a computer and calculate what happens to the average patient given a certain drug. That might be useful information. But if you did that as a replacement for reading about individual patients, you'd be missing the whole point of the narrative notes.

Worse, as this kind of thing becomes feaisble it will feed back on itself and encourage clinicians to write their notes -and therefore to think, inevitably - in machine-readable terms. The authors suggest as much:

As more health care systems move to electronic medical records, there is a unique opportunity to better quantify outcomes. For example, the 16-item patient-rated QIDS-SR [questionairre] has been shown to be highly correlated with clinician-rated measures and sensitive to treatment effects.... At minimum, EMR systems that utilize templates could require clinicians to record a clinical status for example, using the 7-point Clinical Global Impression scale...

Indeed, many say that this already is happening. Now quantification is generally a good thing I think, but only so long as it's an aid to understanding, not a replacement for it.

Yet quantification often does become a replacement for understanding because there's a trap that we face when trying to deal with a complicated set of information. The temptation is to focus on the easiest bit to measure, because that's easy, and then assume that this represents the state of the whole thing. The reason that something's easy to measure is often because it doesn't capture the whole phenomena.

Perlis RH, Iosifescu DV, Castro VM, Murphy SN, Gainer VS, Minnier J, Cai T, Goryachev S, Zeng Q, Gallagher PJ, Fava M, Weilburg JB, Churchill SE, Kohane IS, & Smoller JW (2011). Using electronic medical records to enable large-scale studies in psychiatry: treatment resistant depression as a model. Psychological medicine, 1-10 PMID: 21682950

The Holographic Brain

According to the holonomic brain theory,

Cognitive function is guided by a matrix of neurological wave interference patterns situated temporally between holographic Gestalt perception and discrete, affective, quantum vectors derived from reward anticipation potentials.

Well, I don't know about that, but a group of neuroscientists have just reported on using holograms as a tool for studying brain function: Three-dimensional holographic photostimulation of the dendritic arbor.

A while ago, scientists worked out how to "cage" interesting compounds, such as neurotransmitters, inside large, inert molecules. Then, by shining laser light of the right wavelength at the cages, it's possible to break them and release what's inside. This is very useful because it allows you to, say, selectively release neurotransmitters in particular places, just by pointing the laser at them.

There's a problem though. The uncaging doesn't happen immediately: the laser has to be pointing at the same point for a certain fixed time. This makes it very difficult to simultaneously stimulate many different points - which is, ideally, what you'd want to do, because in the real brain, everything happens at the same time: a given cell might be receiving input from dozens of others, and sending output to the same number.

One solution is to simply split and block the beam into several smaller, parallell beams. This allows you to hit several spots simultaneously but it suffers from the problem that all the spots have to lie in the same 2D "slice". A bit like how, if you taped several laser pointers together, you could project a complex series of dots onto the wall, but not a 3D one.

This is where holograms come in. As everyone knows, holograms appear to be 3D images. By adopting the same kind of algorithms as are used in the construction of holograms, the authors were able to use a single laser to generate a series of stimulation spots within 3D space. The image above shows that they were able to stimulate a single dendritic spine of a single neuron by uncaging glutamate.

Then they moved on to a real experiment: stimulating several branches of a single cell. What they found was that if you stimulate several branches simultaneously, the overall excitation produced is less than the sum of the individual stimulations.

The bottom graph shows this: the grey line is what you'd expect if it was simply summed. Interestingly, a drug called 4-AP, which is used to provoke epileptic seizures in experimental animals, blocked this effect and made cells respond in a linear fashion.

This is clearly an extremely promising method. I've previously blogged about how it's possible to visualize individual dendritic branches in the living brain using another laser-based method, two-photon microscopy. In theory, therefore, it might be possible to both see, and manipulate, the brain on a microscopic level, all without physically touching it at all.

Yang S, Papagiakoumou E, Guillon M, de Sars V, Tang CM, & Emiliani V (2011). Three-dimensional holographic photostimulation of the dendritic arbor. Journal of neural engineering, 8 (4) PMID: 21623008

The Holographic Brain

According to the holonomic brain theory,

Cognitive function is guided by a matrix of neurological wave interference patterns situated temporally between holographic Gestalt perception and discrete, affective, quantum vectors derived from reward anticipation potentials.

Well, I don't know about that, but a group of neuroscientists have just reported on using holograms as a tool for studying brain function: Three-dimensional holographic photostimulation of the dendritic arbor.

A while ago, scientists worked out how to "cage" interesting compounds, such as neurotransmitters, inside large, inert molecules. Then, by shining laser light of the right wavelength at the cages, it's possible to break them and release what's inside. This is very useful because it allows you to, say, selectively release neurotransmitters in particular places, just by pointing the laser at them.

There's a problem though. The uncaging doesn't happen immediately: the laser has to be pointing at the same point for a certain fixed time. This makes it very difficult to simultaneously stimulate many different points - which is, ideally, what you'd want to do, because in the real brain, everything happens at the same time: a given cell might be receiving input from dozens of others, and sending output to the same number.

One solution is to simply split and block the beam into several smaller, parallell beams. This allows you to hit several spots simultaneously but it suffers from the problem that all the spots have to lie in the same 2D "slice". A bit like how, if you taped several laser pointers together, you could project a complex series of dots onto the wall, but not a 3D one.

This is where holograms come in. As everyone knows, holograms appear to be 3D images. By adopting the same kind of algorithms as are used in the construction of holograms, the authors were able to use a single laser to generate a series of stimulation spots within 3D space. The image above shows that they were able to stimulate a single dendritic spine of a single neuron by uncaging glutamate.

Then they moved on to a real experiment: stimulating several branches of a single cell. What they found was that if you stimulate several branches simultaneously, the overall excitation produced is less than the sum of the individual stimulations.

The bottom graph shows this: the grey line is what you'd expect if it was simply summed. Interestingly, a drug called 4-AP, which is used to provoke epileptic seizures in experimental animals, blocked this effect and made cells respond in a linear fashion.

This is clearly an extremely promising method. I've previously blogged about how it's possible to visualize individual dendritic branches in the living brain using another laser-based method, two-photon microscopy. In theory, therefore, it might be possible to both see, and manipulate, the brain on a microscopic level, all without physically touching it at all.

Yang S, Papagiakoumou E, Guillon M, de Sars V, Tang CM, & Emiliani V (2011). Three-dimensional holographic photostimulation of the dendritic arbor. Journal of neural engineering, 8 (4) PMID: 21623008

Autistic Brains 'Genes Differ'

The BBC say:

The brains of people with autism are chemically different to those without autism, according to researchers. A study, published in the journal Nature, showed the unique characters of the frontal and temporal lobes had disappeared.

It's not a bad summary, although it doesn't explain quite how interesting the new results are. Here's the paper, from a joint US/British team: Transcriptomic analysis of autistic brain reveals convergent molecular pathology

The authors took 19 brains from people with autism and 17 healthy ones. These came from people who donated their brains to science and then died. The study involved taking samples from three areas of the brain, the superior temporal gyrus, the prefrontal cortex, and the cerebellum. These are regions that have been implicated in autism, although to be honest, so has everywhere else in the brain.

They then looked at gene expression: mRNA levels. This measures the degree to which different genes are "activated" and being used to make proteins. Bear in mind that a gene itself could be completely normal, and yet be abnormally expressed: this was not a study of DNA mutations. So the BBC's headline is a bit misleading. The genes themselves were not the focus of this study.

Anyway, comparing the autistic and control brains, they found 444 genes that were statistically significantly either over- or under-expressed in the cerebral cortex samples from the autistic group. However, in the cerebellum, there were just 2 differences: so the cerebellum was ruled out from further analysis.

They then replicated the study in a different cortical area in 6 new cases and 5 new controls. They found extremely strong overlap with the original cohort, with the same genes being altered in the same direction in almost all cases. This makes me confident that there is something going on here. This scatterplot shows that almost all of the genes that were significantly different in the first batch were also different in same direction in the second one (although not always significantly, as you'd expect.)

However, the authors didn't stop there, and this is where it gets interesting. First, they used a pattern classification algorithm to try to distinguish patients and controls on the basis of gene expression. This is very much like the paper from last year showing that pattern classification could predict autism on the basis of brain structure.

Interestingly, the algorithm correctly " diagnosed" a case of autism who turned out to have a 15q duplication mutation. 15q duplication is a genetic disorder which causes autism, amongst other things, and it may explain up to 1% of cases of autism. This is only one case but it's important because it suggests that "15q autism" is not all that different to other kinds of autism on the neural level.

The authors then looked at what the over- and under- expressed genes actually were. They found that the "up" genes tended to be genes relating to immune and glial function, while the "down" genes tended to be involved in the formation and function of synapses between cells.

Very interestingly, one of the major clusters of genes , "M12", showed strong overlap with genes previously known to be expressed in a type of cell called PV+ GABA interneurons. In mouse models of autism, these are known to be deficient. M12 was underexpressed in autism, and it contains many genes which have previously been found to be mutated in some people with autism, such as CNTNAP2.

Another cluster, "M16", was overexpressed; it contains genes involved in immune and microglial function (microglia are specialized immune system cells inside the brain). However, M16 did not contain overrepresentation of suspected asd genes.

So this all points to something like this: autism is caused by disruption to the function of certain gene networks in the brain involved in synaptic function. This network is a delicate balance and it can be thrown off course by many different mutations and/or environmental factors.

There's no one gene for autism, but all of the genes for autism might be related, or rather, they might form a team that works together. If you want to look at it this way, you could say that autism is a bit like blindness. People can go blind for lots of different reasons: it could be damage to the surface of the eye, or the retina, or the optic nerve which carries information to the brain, or the brain itself. All of these parts depend on all the others to work, and if one of them goes wrong, the whole system suffers.

Also, whatever the abnormality in autism is, it seems to trigger a secondary change in the brain which is immune and/or glial related. By "secondary" I don't mean that it's less important. It might be what causes the symptoms of autism. But it's not the root cause (because if it were, mutations in this network would cause autism, and they don't seem to.)

This study raises many more questions than it answers, but in a good way. It certainly doesn't explain autism, but it's pointed the way towards more focussed research in the future - gene cluster M12.

Voineagu I, Wang X, Johnston P, Lowe JK, Tian Y, Horvath S, Mill J, Cantor RM, Blencowe BJ, & Geschwind DH (2011). Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature PMID: 21614001

Autistic Brains 'Genes Differ'

The BBC say:

The brains of people with autism are chemically different to those without autism, according to researchers. A study, published in the journal Nature, showed the unique characters of the frontal and temporal lobes had disappeared.

It's not a bad summary, although it doesn't explain quite how interesting the new results are. Here's the paper, from a joint US/British team: Transcriptomic analysis of autistic brain reveals convergent molecular pathology

The authors took 19 brains from people with autism and 17 healthy ones. These came from people who donated their brains to science and then died. The study involved taking samples from three areas of the brain, the superior temporal gyrus, the prefrontal cortex, and the cerebellum. These are regions that have been implicated in autism, although to be honest, so has everywhere else in the brain.

They then looked at gene expression: mRNA levels. This measures the degree to which different genes are "activated" and being used to make proteins. Bear in mind that a gene itself could be completely normal, and yet be abnormally expressed: this was not a study of DNA mutations. So the BBC's headline is a bit misleading. The genes themselves were not the focus of this study.

Anyway, comparing the autistic and control brains, they found 444 genes that were statistically significantly either over- or under-expressed in the cerebral cortex samples from the autistic group. However, in the cerebellum, there were just 2 differences: so the cerebellum was ruled out from further analysis.

They then replicated the study in a different cortical area in 6 new cases and 5 new controls. They found extremely strong overlap with the original cohort, with the same genes being altered in the same direction in almost all cases. This makes me confident that there is something going on here. This scatterplot shows that almost all of the genes that were significantly different in the first batch were also different in same direction in the second one (although not always significantly, as you'd expect.)

However, the authors didn't stop there, and this is where it gets interesting. First, they used a pattern classification algorithm to try to distinguish patients and controls on the basis of gene expression. This is very much like the paper from last year showing that pattern classification could predict autism on the basis of brain structure.

Interestingly, the algorithm correctly " diagnosed" a case of autism who turned out to have a 15q duplication mutation. 15q duplication is a genetic disorder which causes autism, amongst other things, and it may explain up to 1% of cases of autism. This is only one case but it's important because it suggests that "15q autism" is not all that different to other kinds of autism on the neural level.

The authors then looked at what the over- and under- expressed genes actually were. They found that the "up" genes tended to be genes relating to immune and glial function, while the "down" genes tended to be involved in the formation and function of synapses between cells.

Very interestingly, one of the major clusters of genes , "M12", showed strong overlap with genes previously known to be expressed in a type of cell called PV+ GABA interneurons. In mouse models of autism, these are known to be deficient. M12 was underexpressed in autism, and it contains many genes which have previously been found to be mutated in some people with autism, such as CNTNAP2.

Another cluster, "M16", was overexpressed; it contains genes involved in immune and microglial function (microglia are specialized immune system cells inside the brain). However, M16 did not contain overrepresentation of suspected asd genes.

So this all points to something like this: autism is caused by disruption to the function of certain gene networks in the brain involved in synaptic function. This network is a delicate balance and it can be thrown off course by many different mutations and/or environmental factors.

There's no one gene for autism, but all of the genes for autism might be related, or rather, they might form a team that works together. If you want to look at it this way, you could say that autism is a bit like blindness. People can go blind for lots of different reasons: it could be damage to the surface of the eye, or the retina, or the optic nerve which carries information to the brain, or the brain itself. All of these parts depend on all the others to work, and if one of them goes wrong, the whole system suffers.

Also, whatever the abnormality in autism is, it seems to trigger a secondary change in the brain which is immune and/or glial related. By "secondary" I don't mean that it's less important. It might be what causes the symptoms of autism. But it's not the root cause (because if it were, mutations in this network would cause autism, and they don't seem to.)

This study raises many more questions than it answers, but in a good way. It certainly doesn't explain autism, but it's pointed the way towards more focussed research in the future - gene cluster M12.

Voineagu I, Wang X, Johnston P, Lowe JK, Tian Y, Horvath S, Mill J, Cantor RM, Blencowe BJ, & Geschwind DH (2011). Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature PMID: 21614001

Secondhand Smoke Goes To Your Head

Secondhand smoking. It's bad for you. But does it get you high?

According to UCLA researchers Arthur Brody et al, it might do, because exposure to secondhand cigarette smoke can cause you to absorb enough nicotine that it has measurable effects in the brain.

That's quite interesting, but the best thing about this study is the methodology. This is the first neuroimaging study I've seen which involved a car. Not a picture of a car. An actual car.

They used PET scanning to measure the binding of nicotine to brain nicotinic acetylcholine receptors (nAChRs), the major target of the drug. They first injected people with a radioactive tracer compound, in this case a nicotine-like molecule which binds to nAChRs. Because nicotine binds to the same target, it displaces the tracer and reduces the radioactive signal from the brain.

Where did the car come in? Well, volunteers were scanned before and after sitting in a car for one hour, next to a smoker who smoked cigarettes over the course of the hour. An average of 3.7 cigarettes to be precise. The windows were closed to keep the car nice and smoky.

The scene was made even more remarkable by the fact that the subject was still being injected with the tracer compound during this period: they were attached to a drip which went through a little gap in the window and outside. Sadly, they don't show us any pictures...


Anyway, they found that secondhand smoking did cause modest but significant binding to the receptors. The graph shows tracer binding in four areas of the brain - the lower the line, the more nicotine. After secondhand smoke, the lines go down.

After sitting in a "placebo car" in which no-one was smoking, however, there was no effect (the empty circles.) Then later on, the participants were able to smoke some cigarettes first-hand: this had a much stronger effect as you'd expect.

The effect of secondhand smoke was pretty large, though. Actual smoking led to nicotine receptor occupancy of about 50%. Secondhand smoke weighed in at about 20%. Interestingly, in the participants who were regular smokers themselves, the secondhand smoke made them report increased cravings for a cigarette - and this correlated with secondhand smoke nicotine binding (though only in one area of the brain.)

Is this a realistic study? I can't imagine many smokers sit smoking inside their car with the windows up for hours on end. If nothing else because it make their car smell all smoky. (Did they have to pay for the car? If so, this might be the only legitimate example of someone buying themselves a new car using their research grant money...) But the authors say that being in a room with multiple smokers would lead to even higher levels of smoke.

Brody AL, Mandelkern MA, London ED, Khan A, Kozman D, Costello MR, Vellios EE, Archie MM, Bascom R, & Mukhin AG (2011). Effect of Secondhand Smoke on Occupancy of Nicotinic Acetylcholine Receptors in Brain. Archives of general psychiatry PMID: 21536968