Genomic Snippets

Robert Frost

A Mendelian Randomization (MR) study published in Nature Medicine reports a causal association between blood levels of OAS1 and susceptibility to COVID-19 infection. I liked many things in this study, of which I’d like to highlight three.

First, I liked that the authors performed the MR analysis using not just pQTLs, but also eQTLs and sQTLs as genetic instruments. In doing so, the authors reveal the mechanism underlying the association between the genetic variant and the OAS1 protein level in the blood, which, in turn, offers protection against COVID-19 infection.

In a typical scenario, one would expect a genetic variant to increase the protein level by increasing the mRNA level, considering the central dogma of biology (i.e. we expect an increased translation as a consequence of an increased transcription). But often the scenario is not that simple. Many of the pQTL studies published so far have shown that often pQTLs are not eQTLs. That is, often genetic variants influence protein levels without influencing the mRNA levels. The current study’s finding is one such example.

Here, the genetic variant increases the OAS1 protein level not by increasing the OAS1 transcription, but by favoring the OAS1 splicing towards one particular isoform. This is a scenario where a pQTL is not an eQTL but an sQTL. So, this finding teaches us that genetic variants can influence protein levels without influencing the transcription, but instead by influencing the splicing.  The authors have provided a table of posterior probabilities for pairwise colocalizations across eQTLs, sQTLs, pQTLs, and COVID-19 genetic associations, which clearly shows that there is no colocalization between eQTLs and pQTLs or between eQTLs and COVID-19 genetic associations.

The second thing I liked about this study is it draws a clear causal pathway from genetic variant to mRNA to protein to disease risk. This means it is an opportunity to explore how the effect sizes change as we move from one end of a causal pathway to its other end. For e.g. I’d expect that the effect size of the genetic variant’s association with splicing should be larger than its association with protein levels, which in turn should be larger than its association with disease risk (assuming that the genetic variant’s effect on disease risk is mediated entirely via its effect on the protein levels). However, I wasn’t able to retrieve the eQTL’s and pQTL’s effect sizes, which were disappointingly not reported in the study, and the reference pQTL study has not released the summary statistics publicly.

Thirdly, the authors report that the phenotypic association between OAS1 blood level and COVID-19 risk reverses between infectious and non-infectious states. While the non-infectious state OAS1 levels decrease the risk, the infectious state OAS1 levels increase the risk for COVID-19. This is expected as the viral infection itself causally increases the OAS1 levels in the blood. So, here, by using MR, we are able to dissociate the effect of OAS1 levels on COVID-19 risk from the effect of COVID-19 infection on the blood OAS1 levels, which, I think, is a great example of the wonderful things that we can do using human genetics.

Gabriel García Márquez

A brain imaging study published in Human Brain Mapping by the ENIGMA consortium beautifully captures the thinning of brain cortex as humans age. The study presents data spanning almost the full human lifespan from as early as 3 years of age to 90 years of age in an impressive sample size of 17,075 healthy individuals, which is the largest to date.

The rate of decline in the cortical thickness seems to be not uniform throughout the life. The cortex is at its thickest during childhood, then there is a steep decline until 30 years of age, and thereafter, the decline is gradual. This is an impressive finding, and it sparked a multitude of thoughts in me.

I am surprised that the authors didn’t write anything about the biological processes that drive the cortical thinning. Of course, they are due to loss of neurons. But why this happens very early in life? It is likely that the early steep decline in the cortical thickness is the outcome of synaptic pruning, a process through which our brain gets rid of unwanted neurons. Our brain development happens in such a way that first it produces as many neurons as possible (the process called neurogenesis, which happens predominantly in-utero,) then, it keeps the ones it wants and gets rid of the rest. It is a process of molding the brain to perfection like an artist sculpting a statue to its perfect shape by chipping away the unwanted parts bit by bit. The synaptic pruning is carried out by glial cells such as microglia, which are produced predominantly after birth (the process called gliogenesis).  The peaking of the prenatal neurogenesis and postnatal gliogenesis have been beautifully demonstrated in vivo, and more importantly, in vitro using brain organoids, which I have tweeted just a few days ago.

When I looked at the plot from the ENIGMA study illustrating the cortical thinning across lifespan, I wondered what happens to the cognitive abilities during the early steep decline in the cortical thickness. Disappointingly, the authors didn’t discuss that either in the paper. I remembered a great review article published in Annual Review of Developmental Psychology by Prof. Elliot M. Tucker Drob. I have superficially glanced through the paper many months ago.  Particularly, a plot from the article stayed fresh in my memory; it illustrates the age related change in the fluid and crystallised cognitive abilities. In the article, Prof. Tucker-Drob writes

Cognitive abilities that require predominantly effortful processing at the time of assessment (e.g., fluid reasoning, visuospatial ability, episodic memory, and processing speed) typically peak in early adulthood (e.g., the twenties) and decline monotonically throughout middle and late adulthood, whereas cognitive abilities that rely predominantly on recital or rote application of previously acquired knowledge (e.g., crystallized knowledge, procedural knowledge, and specialized professional skills) typically peak in late adulthood (e.g., the sixties) …

From the article, it’s clear that during the adolescence and early adulthood all our cognitive abilities are in the rise, and it amazes me that, at the same time, our cortex is thinning out swiftly due to synaptic pruning. So, it is sensible to assume that the outcome of synaptic pruning is increase in the cognitive abilities. We grow wiser and wiser as our brain gets sculpted to perfection during our adolescence and early adulthood. But I wonder if the cognitive effects of the cortical thinning differ between the fluid and crystallised abilities. It is possible, as both follow different trajectories. While the fluid abilities peak at 20s, the crystallised abilities peak at 60s.

One way to deduce the cognitive associations of synaptic pruning is to find out the cognitive associations of disorders characterised by disrupted synaptic pruning. Yes, you guessed it right. Schizophrenia. One of the strongest GWAS associations of schizophrenia sits in the MHC locus where the alleles corresponding to a higher C4 expression poses an increased schizophrenia risk. C4 codes for complement factor 4, whose deposition over neurons sends eat-me signals to microglia resulting in the neuronal death.  The C4 schizophrenia risk allele leading to an accelerated synaptic pruning has been demonstrated in mice models recently, and I have tweeted about it.

People often tend to quickly equate schizophrenia with poor cognition. But it is much more complicated than that. Clinically, schizophrenia patients exhibit poor cognitive functioning, particularly during the first psychotic attack (which is often considered to mark the disease onset). But as we know, phenotypic associations are affected by multiple confounding factors. If you look at the genetic correlations, schizophrenia exhibits a puzzling relationship with educational attainment and intelligence (two main cognitive phenotypes for which large scale GWASs exist.) Schizophrenia shows a positive genetic correlation with educational attainment, but a negative genetic correlation with intelligence. This might be a reflection of schizophrenia’s differential correlations with crystallised and fluid abilities. Educational attainment is a measure of crystallised abilities, and intelligence is a measure of fluid abilities (at least the one used in the past GWASs, which were powered mainly by the UK Biobank sample). In line with this assumption, I have also observed similar findings in my own work.

In my PhD project, I found that individuals with schizophrenia exhibit poor cognitive functioning in secondary school at around 15 years of age, long before the disease onset. Interestingly, the poor cognition was reflected only in their mathematics grades (a measure of fluid ability), but not in English or Danish grades (measures of crystallised abilities). Genetic correlation analysis revealed positive correlations with language grades, but negative correlations with mathematics grades. More interestingly, even individuals who never had schizophrenia exhibited differential math and language performances when stratified based on their polygenic risk for schizophrenia. Those with higher polygenic risk performed better in language, but poorer in math, and those with lower polygenic risk did the opposite.

Assuming that math grades and intelligence measurements reflect fluid abilities, and language grades and educational attainment reflect crystallised abilities, it makes sense to assume the differential cognitive correlations of schizophrenia should be--at least partly--driven by the synaptic pruning disruption and its likely consequence on the cortical thinning. To test these hypotheses, we need large scale genetic studies based on longitudinal brain imaging measures.

I’ll conclude by listing some research questions that need to be answered by the future studies.

Does the early steep decline in cortex thickness is driven by synaptic pruning?

What are the cognitive effects of cortex thinning during adolescence and early adulthood? Does the effects differ across cognitive domains?

Is the cortical thinning more accelerated in individuals with schizophrenia or in those with increased schizophrenia polygenic risk?

How much influence does genetics has on the early life cortical thinning ?

E. L. Doctorow

A GWAS of peptic ulcer disease (PUD) published in Nature Communications reports an interesting drug target locus. The index variant at the locus is located upstream of gene CCKBR that codes for cholecystokinin receptor B, which is a known drug target. Proglumide, an old peptic ulcer drug, is a cholecystokinin antagonist, and it acts by blocking cholecystokinin receptors A and B sub-types. This a nice example of a GWAS locus supporting existing drug targets.I have tweeted similar findings before (ex1, ex2)

A quick lookup in GTEx database revealed an eQTL association between the index variant rs10500661 and CCKBR, but only in the testis tissue. However, I couldn’t stop noticing that the direction of the eQTL association align with rs10500661’s direction of association with PUD risk and proglumide’s mechanism of action. The effect allele T, which decreases the CCKBR expression, has a protective effect on PUD (OR-0.90). And this aligns with the fact that proglumide treats PUD by blocking the receptor protein.

If you think about it, what we are doing here is reverse engineering. We are leveraging the knowledge about the drug proglumide and pinpointing the causal gene (CCKBR) and the disease variant’s functional consequence on the causal gene (decreasing the CCKBR expression). Though this has no translational benefit, it is kind of cool, isn’t it? Perhaps, all such loci with known drug targets could be curated and used as a training set for machine learning methods that are recently being developed to identify GWAS causal genes.

Learning about this CCKBR locus branched out two streams of thoughts in me—one, about eQTLs and the other, about GWAS loci leading to drug discoveries.

I have been learning about the role of eQTLs in human diseases for quite some time now, and my understanding so far is this: disease relevant eQTLs are a drop of true signals mixed in a big bowl of noise. The noise doesn’t necessarily mean that the eQTLs themselves are false positive signals. Most of the eQTLs in fact represent true associations between the genetic variant and the mRNA level. However, how many of those genetic effects are propagated to protein, to protein’s molecular function, and finally, to the end phenotype is not clear. Hence, I am often inclined to trust eQTLs when they are presented not as the lead evidence, but as a supportive evidence such as the one I discuss in this post. I am sure I have seen many such examples before, and have tweeted about it. But for now, I am able to recollect one, which remains fresh in my memory because of a peculiar characteristic of the finding. In this paper, the authors found a significant association between a loss of function mutation in gene GIGYF1 and type 2 diabetes (OR - 4.15) in the UK biobank, and interestingly, they also found a GWAS locus containing GIGYF1 close to the index variant. And it turns out that the index variant is an eQTL for GIGYF1, and the effect allele increases the GIGYF1 expression, and has a protective OR (0.96). This aligns with the rare variant association, where loss of function of GIGYF1 increases the diabetes risk. So, what we are seeing here is that human genetics is offering evidence that if you knockout the gene, the risk of T2D increases, and if you increase the gene expression, the risk decreases. That’s something we don’t get to witness often. So, it stayed in my memory. The point is, in such scenarios, it’s hard to discard the eQTL evidence as irrelevant, isn’t it ? And that’s why I tend to trust eQTLs when it is presented as a supportive evidence, but not when when it is presented as the lead evidence. This is also the reason why my enthusiasm for transcriptome wide association studies (TWASs) that was once in the peak has started declining lately. And recently, my enthusiasm for proteome wide association studies (PWASs) is in the rise. I have tweeted about one recently. But it looks like PWASs too might follow the same trajectory as the TWASs.

The second train of thought that I had while reading this paper is about turning GWAS discoveries into drug discoveries. One question that lingers in my mind whenever I stumble upon a GWAS locus containing a known drug target is this: if this drug has not been discovered already, would this GWAS finding lead to a successful drug discovery ? Finding the answer for this question is very important. People who firmly believe that GWAS discoveries will lead to drug discoveries often base their beliefs on examples such as the ones that I mentioned earlier. Particularly, these examples are used to justify the importance of minuscule effect sized loci. Often, a comparison is made between the GWAS locus’s effects size and the existing drug’s effect size. The GWAS locus represents a trivial disruption of the gene hence the small effect size. The same gene can be disrupted strongly using drugs, which will lead to a large effect size. And, one particular paper is often being referenced in such scenarios--a Plos Genetics paper that retrospectively shows that, compared to drugs without genetic evidence, drugs with genetic evidence are more likely to succeed in clinical trials and get FDA approval. All these are convincing points, but my only worry is all these evidences make sense only in the hindsight.

A paper published in the journal Endocrinology reports an interesting Mendelian Randomisation (MR) study on the causal association of thyroid hormones—TSH, T4—on blood lipid profiles. A specific aspect of this study grasped my interest. The authors studied the causal association of thyroid hormones on blood lipid profiles within the physiological range. What is special about studying a causal association within the physiological range, you might wonder.

Whenever I cross paths with an MR paper, my first thought will be: is it possible to answer the same question using the conventional approach, a randomized controlled trial (RCT)? Often, the scenario will be that it’s not possible, and this makes the study interesting as it gives me an opportunity to appreciate the fact that human genetics is enabling scientists to find answers for biological questions that otherwise cannot be found. And that’s why I often hand pick such studies and tweet to my followers. This study is one such example.

The authors of this study wanted to answer the question: do  TSH and T4, within the physiological range, causally influence blood lipid levels? It is not possible to conduct an RCT for this. Because, one, it is not ethical to intentionally administer thyroid hormone to an otherwise healthy person, and two, it is not easy—perhaps, not possible—to trivially increase or decrease the TSH or T4 levels in the blood without bouncing them off the physiological range because even slight changes in their levels will provoke the hormonal feedback systems.

The MR analysis indeed showed that an increase in TSH level, within the physiological range, causally leads to increases in the LDL and total cholesterol levels in the blood. Similarly, a decrease in T4 level, within the physiological range, causally leads to increases in the LDL and total cholesterol levels in the blood. It’s amazing that we can find these effects by leveraging human genetics, isn’t it?

Another reason--in fact, the main reason—why I found this study interesting is it kickstarted a train of thought in me about human cognition.

Although I work mainly on the genetics of brain-related phenotypes, being trained as a medical doctor, it’s in my nature to be interested in the genetics of all human diseases and traits.  Often, I find that learning about the genetics of traits and diseases whose biological mechanisms are simple and straightforward helps to improve my understanding of the genetics of brain-related diseases and traits. Or at least, it helps me to appreciate their complexities on a deeper level. When I read this paper, I couldn’t resist comparing the relationship between thyroid hormones and lipids with the relationship between head circumference and cognition (intelligence, if you like). And the differences are indeed fascinating.

The positive relationship between thyroid hormones, T4, for example, and cholesterol levels is observed throughout the distribution spectrum, and not just within the physiological range. Extremely low levels of T4 (i.e. clinical hypothyroidism) lead to hypercholesterolemia and similarly, extremely high levels of T4 (i.e. clinical hyperthyroidism) lead to hypocholesterolemia. On the other hand, if you consider the head circumference and cognition, within the physiological range (i.e. within plus or minus 2 SDs of the population average) there exists a positive relationship between the two (many will frown upon this, however, the association has been robustly demonstrated using MR using strong genetic instruments). However, outside the physiological range, extremely low (microcephaly) and extremely high head circumference (macrocephaly,) both lead to extremely poor cognition (i.e. intellectual disability). This is due to a phenomenon called haploinsufficiency.

Hold tight, now I am dragging you through a rabbit hole through which I was dragged through by my train of thought.

Most of the brain-specific genes are haploinsufficient or dosage-sensitive, that is, either decreased (e.g. loss of one allele) or increased dosages (e.g. duplication of the gene) of these genes lead to loss of function. This dosage sensitivity can be well appreciated if we look at the effect of copy number variations (CNVs) on cognition, where both deletions, as well as duplications, lead to poor cognition (see these threads mine: thread1, thread2). This is, however, not the case for head circumference. There are many examples of CNVs where deletion leads to microcephaly and duplication leads to macrocephaly, and the phenomenon is called the mirroring effect. I’ve written twitter threads on this (thread1, thread2). Note that the dosage sensitivity works only on the brain function (cognition), but not on the brain structure (head circumference or brain volume). When we take a slightly deeper look, we will realize that the molecular functions of the dosage-sensitive genes themselves are not affected. The gene is, of course, transcribed to mRNA, the mRNA, to protein, which indeed performs its molecular function, e.g. brain cells proliferation in the case of head circumference. But the final phenotype, the higher-level function (i.e. cognition) is what is sensitive to the gene dosage. So, a foolproof system is in place, beyond the boundaries of the central dogma, to protect the brain from going awry. This is, of course, a design that evolution has perfected over millions and millions of years.

This foolproof system is why there exist no rare variants that boost cognition. Even a recent rare variants association analysis in almost 500,000 individuals in the UK Biobank didn’t yield any large effect positive association for intelligence. It is, of course, unrealistic to expect a large effect negative association in UKBB as the participants here are cognitively healthy.

Likewise, this foolproof system is why there exists no evidence of positive selection of genetic variants for cognition or any brain-related phenotypes, for that matter, in any of the human populations. And we will not find any in the future either. This means it’s theoretically impossible to find biological differences in cognition between human populations, at least due to evolutionary adaptation. Because no random rare mutations can ever increase cognition beyond a threshold as all brain-specific genes are dosage sensitive. They can only do so below the threshold, which will be outside the natural selection’s radar. And that explains why common variants additively have such a huge effect on human cognition. All those tiny sneaky bastards who have escaped natural selection and managed to survive in the population.

In his autobiography “On the move,” Dr. Oliver Sacks writes about his experiences with patients with Tourette syndrome. For the first time, I am learning about the psychiatric features of Tourette syndrome.  I’ve always imagined Tourette’s as predominantly a neurological condition characterized by involuntary movements (motor tics) and sounds (vocal tics). I find the psychiatric symptoms associated with Tourette’s fascinating, but also terrifying. I was shocked to read that the neurologist Gilles de la Tourette who first described this condition (hence the name), was shot in the head by his own patient (there is a picture in the Wikipedia page depicting the shooting).

In 1971, Dr. Sacks was approached by a The New York Times journalist, who wanted to write about the patients with sleeping sickness at Beth Abraham hospital in New York who were awakened by the magical drug L-dopa. One of the side effects of L-dopa that these patients developed is tics, which reminded Dr. Sacks of Tourette syndrome, which he admits he had never seen before, but only read about it. Following the publication of The New York Times article, Dr. Sacks receives letters asking for medical opinion from many who suffer from tics, similar to what was described in the article.  He develops interest in Tourette syndrome and starts meeting patients with Tourette’s. Dr. Sacks’ descriptions of his Tourette’s patients will make you feel as if you were physically present in his clinic, and have witnessed his patients with your own eyes.

The tics and abnormal movements seen in Tourette’s patients are involuntary, often triggered by external stimuli, and sometimes they manifest as imitations of the stimuli themselves. Dr. Sacks describes one such experience.

People with Tourette’s are often unusually open to hypnosis and suggestion and disposed to involuntary repetition and imitation. I saw this at that first TSA meeting when, at one point, a pigeon flew onto a windowsill outside the conference room. It opened and closed its winds, fluttered, and then settled down. There were seven or eight people with Tourette’s sitting in front of me, and I could see several of them making fluttering movements with their arms and shoulder blades, echoing the pigeon or each other

In 1976, John, a Tourette’s patient, presents himself as a research subject to Dr. Sacks.  When John describes his own condition, Dr. Sacks senses a strange mixture of grandiosity and self-deprecation in his description.  

I am the greatest Touretter in the world. I have the most complex Tourette’s you will ever see. I can teach you things about Tourette’s that nobody else knows. Would you like me as a specimen to study?

Later, during one of the sessions with John, Dr. Sacks spots a peculiar characteristic in John’s vocal tics. By recording and playing it back at slow speed, Dr. Sacks discovers something that blew my mind. I was at the edge of the seat for this whole paragraph.

I was struck in particular by a strange sound that John often uttered along with his tics. When I recorded this and slowed down the playback, elongating the sound, I discovered that it was in fact a German word—“verboten!”—crushed into a single unintelligible noise by its ticcish rapidity.  When I mentioned this to John, he said that was how his German-speaking father would admonish him whenever he ticced as a child. I sent a copy of this tape to Luria, who was fascinated by what he called “the introjection of the father’s voice as a tic”

The mind of a Tourette’s patient, it seems, is extremely vulnerable, to the point that one can easily implant a hallucination in the patient just by uttering it.

One summer day, while John was in my office, a butterfly flew in through the open window. John followed its soaring, zigzag flight with sudden, erratic jerks of his head and eyes as he poured out a stream of endearments and imprecations: “I want to kiss you, I want to kill you,” he repeated, and then abbreviated this to “kiss you, kill you, kiss you, kill you.” After two or three minutes of this—he seemed unable to stop as long as the butterfly was fluttering around—I said, jokingly, “If you were really concentrating, you could disregard the butterfly, even if it landed on your nose.

The moment I said this, he clutched the end of his nose and tore at it, as if to dislodge an enormous butterfly which had attached itself there. I wondered if his excessively vivid Tourettic imagination had crossed over into hallucination, conjuring up a phantom butterfly as real, perceptually, as an actual one. It was like a little nightmare being enacted in full consciousness before me.

A nature paper reports a discovery of the mechanism behind glutamate transporters’ dual transport function. This is the first time I am learning about this interesting characteristic of the glutamate transporters. The glutamate transporters transport both glutamate and chloride ions across the cell membrane, but both these transport processes seem to be “thermodynamically uncoupled.” That is, the glutamate flow and the chloride flow are independent of each other. One could experimentally disrupt one flow without disrupting the other. Scientists have already guessed that there should exist two pores, one for the glutamate and the other for chloride as it’s impossible for both to flow through the same pore. However, the existence of a separate channel for chloride has not been shown at the molecular level till now.  By freezing the glutamate transporter in its open configuration and imaging it using cryo-electron microscopy, the authors reveal to the world, the hidden chloride channel in the glutamate transporter, for the first time. The glutamate transporter protein seems to undergo a configuration such a way that a bunch of its hydrophobic amino acids come together to create pockets on the inner and outer sides of the cell membrane, and create a channel for the chloride ions to pass through. This mechanism seems to be highly conserved across species, from archaea (single cellular organisms) to mammals, which means this has been happening for more than a billion (?) years, and here we are witnessing it for the first time. Fascinating isn’t it ?

I got interested in this paper when I learnt that the chloride transporting function of the glutamate transporters is more important in brain than in other tissues, and a missense mutation in the glutamate transporter, which specifically affects the chloride flow (but not the glutamate flow), causes cerebellar ataxia. Reading this triggered a train of thoughts in my brain. Now the scientists have precisely figured out the amino acid residues in the protein that forms the chloride channel, I wondered, if there will be individuals in the population who carry missense mutations affecting any of these amino acid residues. If so, what kind of phenotypes they express? I also wondered, why the authors of this paper didn’t approach a human geneticist and asked “Hey, we found that these amino acids are essential for the chloride ions to flow through the glutamate transporter, and there is already one missense mutation reported in the literature. Could you look up if other mutations affecting any of these amino acids exist in the population ?”

As I mentioned in my earlier post, among the many papers that I bookmarked for “reading later”, this preprint was in the top of my list, and I managed to read it today. There are two reasons why I had a special interest in this preprint.

First, it addresses a question that I—and probably many others—were thinking about more and more recently: why some individuals carrying a large effect mutation(s) for a particular disease do not develop the disease; what factors make them resilient to the disease?

Second, I learnt about this preprint through Dr. Sébastien Jacquemont (a co-author in this paper,) a scientist whose work I’ve started admiring a lot recently. For e.g. I found two of his papers on the associations between copy number variants and intelligence highly insightful, which I have also summarized as twitter threads (thread1, thread2.)

In this paper, the authors have analyzed the exomes of around 10,000 ASD cases and 100,000 non-ASD individuals (who were mainly from the UK Biobank) with one particular aim: what makes non-ASD individuals who carry loss of function (LOF) mutations in known ASD genes resilient to ASD?

The authors have looked into a list of 156 genes what seem to be “robustly associated with ASD.” For each gene, the authors calculated the ASD penetrance (in terms of relative risk, which is “the risk of being diagnosed with ASD when carrying such variants,”) and compared the penetrance estimates with the LOF intolerance measures, and found them to be highly concordant. This reinforces the notion that genes linked to ASD are simply genes linked to brain in general. The most robust metric of a gene’s specificity to brain is its mutation constraint. And we see it for the ASD genes as well, also intellectual disability (ID) genes. Note that there is no clear distinction across ASD specific genes, ID specific genes and brain specific genes. And this has been the basis of a paper titled ‘Insufficient Evidence for “Autism-Specific” Genes,’ that came out last year, and stirred some tension among the ASD researchers.

Despite the high concordance between ASD penetrance and mutation constraints, there seem to be some interesting exceptions, e.g. PTEN, SHANK2, which are highly penetrant for ASD, yet less intolerant to LOF mutations. That is, despite mutations in these genes often leading  to ASD or ID, they seem to evade natural selection. How? Such outlier genes, perhaps, hold the key to opening the ASD mystery? The authors also report a list of 25 genes that are fully penetrant for ASD. That is, among the 100k non-ASD individuals, the authors found not even one person who carried a LOF mutation in any of these 25 genes and did not develop ASD. I find this list extremely intriguing, and I plan to dive deep into this list later.

The authors found that the cognitive functions of non-ASD LOF carriers in the UK biobank are significantly lower compared to non-ASD non-carriers.  More penetrant the genes are for ASD, higher the differences in cognitive functions between carriers and non-carriers. My immediate thought on reading this finding is if the authors repeated the same analysis in some brain-specific highly constrained genes (irrespective of their association with ASD), they would have seen the same. In fact, that’s what exactly was shown in Dr. Jacquemont’s previous paper. My point is all these analyses are reflecting brain specificity, rather than ASD specificity.

Importantly, the authors found that the liability threshold model holds true for ASD; non-ASD LOF carriers had lower burden of common ASD risk variants than ASD LOF carriers, which seems to indicate that the decreased common variant burden might have contributed to the resiliency of the non-ASD LOF carriers.

Finally, the authors searched for clues that might support the “female protective effect” hypothesis. The authors expected (or may be they didn’t) that female sex might be one of the contributing factors for ASD resiliency in non-ASD LOF carriers. Unfortunately, that wasn’t the case. The non-ASD LOF carriers were equally distributed among males and females.

Yesterday, I came across a fascinating paper, which reports that a protein truncating variant in ACTN3 in the homozygous state (i.e. complete knockout of the protein) is extremely common worldwide.  It seems there are around 1.2 billion human knockouts for ACTN3 worldwide. 1.2 billion! (That is the number of humans living in India). If you’re wondering why this null variant became so common, there’s an extremely fascinating reason for that.

The gene ACTN3 codes for alpha-actinin-3, a muscle protein that is present only in the fast-twitching skeletal muscle. There are two skeletal muscle types--type 1 or slow twitching and type 2 or fast twitching. While the fast twitching muscle is used for rapid contractions that lasts for short duration (used during sprinting), the slow twitching muscle is used for sustained contraction that lasts for longer duration (used during endurance activities).

In the current paper, the authors show that individuals with the null mutation (XX) withstand cold for longer duration than individuals without the null mutation (RR). The authors immersed 15 XX and 27 RR individuals in cold water for around 2 hrs (with multiple 10 mins breaks,) and tested how many were able to maintain their core body temperature (measured from rectal temperature) above 35.5 degree Celsius. While only 30% of RR individuals maintained their core body temperature, around 70% of XX individuals did the same. That is a large difference. The authors didn’t explicitly mention any effect size estimate. But I calculated a crude estimate of OR-4.75 by making a 2x2 table from their data. That’s a large effect size, and it makes sense given that they were able to show a significant association in such an extremely small sample size.

It seems that this beneficial effect (ability to tolerate cold for long) has led to the positive selection of this null mutation. The authors speculate that the mutation must have provided the ancestral humans who migrated out of Africa a survival advantage against harsh cold climates.

Interestingly, the same mutation leads to decreased athletic ability, specifically sprinting ability (sprinting requires fast twitching muscle, which the XX individuals cannot use). A previous paper shows that the XX genotype is underrepresented in elite sprint athletes. Again, I crudely calculated the effect size by making 2x2 table comparing only XX and RR genotype individuals (to make the effect sizes comparable between the two studies,) and found that the effect size is OR-0.18 (inverse OR-5.5), which is similar but opposite in direction to the XX’s effect size for cold tolerance (OR-4.75).

Once in a while, I trip over a bunch of interesting articles, and go into a frozen state, unable to decide which one to read or which one to tweet. Today it happened. My solution to escape from this state of indecisiveness is to just list down all the articles, perhaps describe a little bit, and save it for reading some day. It’s likely that “some day” will never come. Yet, at least, I’ll have the memory of stumbling upon these article titles. Such memories will surface back as feelings of deja vu at times when I trip over similar articles again in the future.

A paper in Nature Communications report in detail about the architecture of the trans-ethnic genetic correlations for ~30 complex traits. The authors have adapted the stratified LD score regression framework to partition the trans-ethnic genetic correlations based on various functional annotations. I haven’t had time to read the entire paper, but I managed to glance through the abstract, figures and discussion. One of the findings stood out: the authors show that trans-ethnic correlations (meta-analysed across ~30 complex traits) are significantly lower in genomic regions encompassing genes expressed specifically in skin and immune related tissues. This makes sense as many population specific associations were reported for skin traits and immune diseases, both of which are liable to environmental influences such as sunlight and pathogens. On the other hand, the trans-ethnic genetic correlations are significantly higher for genomic regions encompassing brain-specific genes. This is an important finding because this suggests that it is less likely that we will find any population-specific GWAS associations for brain related traits (e.g. intelligence) and diseases (e.g. schizophrenia). So, all the GWAS loci identified for brain related phenotypes are likely shared across all populations. In fact, the near perfect genetic correlation (rg=0.98) for schizophrenia between East Asians and Europeans aligns with this finding. It is, however, important to note that still the polygenic scores will not be portable across populations because the LD differences between the populations will screw up the polygenic predictions. In a hypothetical scenario, if one is able to experimentally identify the causal variants at all the loci, then perhaps the polygenic scores constructed using only the causal variants might become portable across populations.

Speaking of the devil, another interesting paper in America Journal of Human Genetics (AJHG) reports on the genetic architecture of inflammatory bowel disease (IBD) in 3,418 African Americans (1774 cases and 1644 controls) and how they differ compared to IBD genetics in Europeans. As expected, there seems to be a big difference: out of 241 lead variants identified in Europeans, only 41 replicate, and even among these replicated loci, the effect size correlation was only 0.68 (this poor correlation could be blamed on the LD-differences). Anyways, this paper made me go back to the Nature Communications paper I discussed earlier, and confirm that the tissues with low trans-ethnic genetic correlations include small intestine, colon and ileum. It is great to see that the results from both these papers fit perfectly together. I’ll definitely dive deep into this paper soon.

An another paper in AJHG’s “sister journal” Human Genetics and Genomics (HGG) advances reports a monogenic (sort of) case of ADHD caused due to highly penetrant translocation mutation in CTNND2. I work a lot with ADHD, yet my knowledge on the highly penetrant rare variants causing ADHD is poor. I remember a few large deletion-CNVs causing ADHD, but my mind is blank when I think about ADHD-specific mutations (may be, I always imagined that there aren’t any). So, the words “penetrant attention deficit disorder” in the title pulled my attention. I’d like to read this. Also, studies like this are great examples to demonstrate the genetic component in the etiology of ADHD, which still many people (including doctors!) deny.

A Nature paper reports the role of creatine kinase in the heat generation by brown fat. This paper reminded me of a guest lecture that I attended in 2011 while I was in my first year of my MD at Christian Medical College, Vellore, India. The guest speaker, who is an alumni of our department (I don’t remember the name), spoke about the accidental discovery of the existence of brown fat (or beige fat) in adults while PET scanning adult cancer patients. He presented his molecular research work on understanding the mechanisms underlying brown fat thermogenesis, and how his research findings might help to discover new anti-obesity drugs that can induce brown fat generation in adults. I am curious to know what’s the “big discovery” in the current Nature paper.

An another Nature paper titled “Million-year-old DNA sheds light on the genomic history of mammoths” has come out. Common! Who will not be interested to know what the DNA of mammoths have revealed? Though this is not within my area of expertise, I’m still extremely curious. At least, I’ll read the associated “news and views” article which will be understandable I hope.

An another Nature Communications paper by Peter Visscher and his team has come out. Any paper from Peter Visscher is worth reading, and should be read. No scientists will disagree with me in that, I guess. This paper reports how the heritability estimated by correlating the phenotypic similarity with genetic similarity changes from unrelated individuals on the one end of the spectrum to monozygotic twins on the other. It seems like an interesting paper.

Yesterday, I had the pleasure of reading an interesting GWAS of Lewy body dementia (LBD) published in Nature Genetics. LBD shares pathological and clinical features with Alzheimer’s disease (AD) and Parkinson’s dementia (PD). So, I was curious to know how the GWAS results of LBD compare to that of AD and PD. As expected, the APOE locus towered over all other loci.  The effect size of the APOE locus was OR-2.46. This is much smaller than the APOE’s effect size for AD. For e.g. this study reports OR-4.6 for AD in heterozygous APOE-e4 carriers and OR-25.4 in homozygous APOE-e4 carriers. So, although APO-E locus seems to account for the largest number of LBD cases (smallest P value and common allele frequency) in this study, its penetrance is markedly lower compared to AD.  

Whenever I read a GWAS, the first thing I look for is which locus has the largest effect size. This cannot be identified using a Manhattan plot, as P values do not reflect the effect sizes. The summary statistics table in the paper showed that a chromosome 1 locus had the largest effect size. At this locus, the index variant is a missense variant in gene GBA, which codes for a lysosomal enzyme glucocerebrosidase. (This is a known locus. A high penetrance of up to ~33% was reported for GBA mutations in LBD in Ashkenazi Jews.) Despite having the largest effect size, this locus was not the one with the smallest P value. Because, this variant is relatively rare and so its P value was not as smaller as APOE’s. This is something that I often emphasise in twitter. A highly penetrant rare variant can have less significant P value than a moderately penetrant common variant. This doesn’t mean that the rare variant is less important than the common variant. It’s in fact the opposite. Variants with larger effect sizes point us to the core disease pathways. In the current study, for e.g., the GBA locus points us straight to the core pathology of LBD: lysosomal dysfunction.

Another perfect example for the effect size vs P value debate is the ACAN locus identified in the 2010 GWAS of height published by the GIANT consortium. I’ve written a twitter thread on this. This locus had the largest effect size, but did not tower tall enough to grab the authors’ attention (a word search for “ACAN” will reveal zero results in this paper’s main text.) The authors wouldn’t had any idea, I believe, that 10 years later, two independent research teams (this and this) will show that the ACAN locus’s large effect association was driven by a variable number tandem repeat (VNTR) within ACAN. The effect size of this VNTR is astonishingly large, in fact, one of the largest ever reported for height. Also, the gene ACAN codes for a cartilage protein thereby falling in the core biological pathway of height. Fascinating, isn’t it ?

A preprint that I read recently had an example demonstrating that even if you statistically force the phenotypes to be orthogonal, the genetic correlations between them persist. In fact, I’ve seen this in my own work. I constructed orthogonal principal components from school grades in various subjects, and found significant genetic correlations across them. This reminds me of a question that I—and many others might—have been pondering for a while: do the genetic correlations between phenotypes represent true biological correlations or simply the phenotypic correlations that were unaccounted for in the GWASs from which the summary statistics originate or perhaps, a mixture of both? The last one seems most reasonable. Although the preprint here demonstrates that genetic correlations persist beyond phenotypic correlations, it cannot be taken as a proof that the genetic correlations being captured here represent true genetic overlaps, meaning, the underlying biological mechanisms are shared across the phenotypes. This is because the genetic correlations capture the phenotypic correlations beyond what are measurable in the sample at our hands. This is both good and bad.

Good because it enables researchers to deduce phenotypic correlations between phenotypes that cannot be easily measured in the same sample due to time or ethical constraints. Bad because it doesn’t enable researchers to differentiate simple phenotypic correlations from true genetic correlations. A nice example for this is the genetic correlations that reflect the comorbidity between two disorders. A person can be comorbid for two diseases either because the person is truly suffering from both the disorders or the person was misdiagnosed for one of the disorders.

Thanks to Mendelian Randomisation (MR), which has circumvented some the cons of the genetic correlation analysis (but remember, the MR has its own flaws.) The MR works really well when the biological mechanisms underlying the phenotypes are clear. A very good example is the non-causal association between HDL and myocardial infarction. However, when it comes to behavioural phenotypes, the MR is, I think, as much blindfolded as the genetic correlation analysis is.

A paper published in NPJ genomic medicine reports an association between human salivary amylase gene (AMY1) copy numbers and adiposity traits—BMI, total fat mass and total trunk fat mass--in the Middle Eastern population.

This brought back memories from the time when I was in the second year of my MD at Christian Medical College, Vellore, India (Unlike in USA, MD in India is a three years post graduate degree, which can be done only after completing the 5.5 years undergraduate medical degree MBBS.)  I was then just beginning to learn statistical genetics. Somehow, I came across a Nature Genetics commentary titled “Adaptive drool in the gene pool”, which immediately grabbed my attention. The commentary was written by John Novembre, Jonathan K. Pritchard and Graham Coop. (At that time, I had no idea who these scientists were. Only years later, I learnt that they are some of the world’s top population geneticists.) Fascinated by the commentary, I read the full original article by George Perry and colleagues, and was awestruck by the fact that the entire AMY1 gene was duplicated multiple times over the past thousands of years because humans shifted from a fat based diet to a starch based diet (reflecting the shift from hunting to farming). Inspired by this paper, I came up with an hypothesis that individuals with less number of AMY1 copies might be intolerant to present day starch rich diet and so might be susceptible to diseases like type 2 diabetes (T2D) and obesity. I went close to submitting an application for an institutional grant to conduct a pilot study, but then the head of the endocrinology, who initially agreed to help me with recruiting T2D patients, stopped answering my emails and the project has to be killed.

I searched through my old emails and even managed to dig out the IRB application. It is a great feel to read something you wrote nine years ago. It brought back many memories. I laughed at how naive I was. I thought I could do this study by recruiting just 100 patients and 100 controls. I didn’t even consider the allele frequency of this CNV in the Indian population or what are the chances that I’ll find different copy numbers in such a small sample.I am glad now that I didn’t do that study.

I’ve been reading Oliver Sacks’ autobiography “On the move” recently. A response letter that Dr. Sacks wrote to a medical student inspired me to the core.

This happens in 1976, three years after Dr. Sacks’ Awakenings was published. By then Dr. Sacks was fired from the Beth Abraham Hospital at New York, the very hospital, which he places in history through his book Awakenings.

Inspired by Awakenings and Migraine, Jonathan Cole, a medical student at Oxford, writes to Dr. Sacks expressing interest to do a two months elective at the Dr. Sacks’ department.

Dr. Sacks writes back a heart-felt sincere reply. He apologizes first for being “so long in replying,” and reasons out. “My delay in replying is because I don’t know what to reply”.  He then explains his situation in his own poetic way.

“I don’t have a Department. I am not in a Department. I am a gypsy, and survive—rather marginally and precariously—on odd jobs here and there.

When I worked full-time at Beth Abraham I often had students spend some time with me for their electives—and this was an experience we would always find very pleasant and rewarding.

But now I am, as it were, without any position or base or home, but peripatetic here and there. I can’t possibly offer any formal sort of teaching—or anything which could be formally accredited to you”

Imagine, how many professors will have the heart to admit such a thing after having written two famous books and gained a reputation in the community. Everyone in academia gets ups and downs, but people tend to project only the ups to maintain a social status. Dr. Sacks has clearly been a beautiful anomaly in the academic world.

Then the letter gets even more beautiful. Having explained his situation, Dr. Sacks offers his interest to train the student informally. The way he explains it, I think, will tempt any student to accept the opportunity despite that it offers no accreditation.

“Informally (I sometimes think) I see and learn and do a great deal, with the extremely varied patients I see in various clinics and Homes, and every seeing-and-learning-and-doing situation is, eo ispo, a teaching situation. I find every patient I see, everywhere, vividly alive, interesting and rewarding; I have never seen a patient who didn’t teach me something new, or stir in me new feelings and new trains of thought; and I think that those who are with me in these situations share in, and contribute to, this sense of adventure. (I regard all neurology, every-thing, as a story of adventure!)”

As Dr. Sacks concludes, he reminds the student again that he is delighted to offer the training, but the student will not earn any credentials of any kind out of this.

“Do write and let me know how things work out with you—once again, I would be delighted to see you in an informal, casual, peripatetic way, but I am in no sense “set up” for any formal teaching what-ever.

With best wishes—and thanks,

Oliver Sacks. “