 to see you all here today. And it's a really great pleasure to introduce Henning to Meyer. Henning, I'll do the official bits first, is the Professor of Social and Behavioral Science and the Felberg Chair of Maternal and Child Health at the Harvard School of Public Health. By training Henning's a physician and a social epidemiologist, had his degrees from the University of Bohnen Germany and his PhD from Erasmus in University in Rotterdam. And Henning's really been pivotal, absolutely pivotal in the success of one of the most impressive birth cohorts worldwide, the Generation R cohort. It started with over 10,000 pregnant mums and has followed these kids from birth now into adolescence. And using these exceptionally rich data, Henning has published extensively on the origins of child developmental problems and unraveling the interplay between genomics and the environment. And Henning's also played a major role in the equally impressive Rotterdam study which is focused on the other end of life. This study has produced literally thousands of publications on the origins of later life cancer and heart disease, many under Henning's leadership. Henning recently moved to Harvard where he's planning really innovative new directions for his group, focusing on some of society's most vulnerable and marginalized children. I'll just add by noting that I met Henning in 2005, 14 years ago, at the NAMIH whenever we were both junior postdocs in the child psychiatry branch under Judy Rappaport. And there are three things to note here, three things to note. First of all, working together as postdocs was the beginning of a very long and I think a very fun collaboration between my group and Henning's. Second thing to note is science is always much better than if you do it with people you like doing it with. And finally, and most importantly, I think you'll all agree that neither Henning nor I have aged a day. So please join me in welcoming Henning. Wow, what a sweet, what a special. I'll come every week to hear that. Philip, yes, thank you for the very kind invitation and the introduction and good to be here. Hello, yes indeed, 2005 I came and I introduced myself to Julia saying, I want to learn imaging to do a birth cohort. And she said, we don't need any birth cohorts anymore. And I try to convince her that if we do imaging it's worthwhile. And so 15 years later, she's not around, but I want to convince you that it was worthwhile. So because at that time I knew nothing about imaging. So here we go. So this is the study I had. We started that actually. In 2002 we started and then it was just data collection. So I took time off to come to Bethesda. It's talking this whole talk will be about generation R and a few sort of recent researchers, something I'm up for discussion on to show what we're struggling with how we address population neuroscience. So generation R. The R stands for Rotterdam. It's a prospective birth cohort. So it started in fetal life or early fetal life. And we promised our funders to get 10,000 mothers and we nearly got there. I think we should have continued one more months to get it, but we sort of had an end date too. And the other thing which is not so relevant for today's talk, but is that Rotterdam is more than other Dutch or European cities very multi-ethnic, so half of the population has parents born outside the country. That's a lot for European city. That's in the League of Toronto. So let me just briefly, very briefly because it's always the boring part, introduce the study. So we, most of the women started in when they were pregnant. And why we got them into the study is we promised them very detailed ultrasound measures and that's what they like. That's why they came. Some stopped immediately after they had their fourth assessment but many continue 15 years later. So we did ultrasound. We did a few other measures. Blood, it's important we got them from mothers, fathers and child for much of our work. And perhaps I don't want to go into that much. Questionnaires, a lot of questionnaires on child behavior cognition, that was the part. I was responsible and the MRI, which I had sort of this, that's why I came to Jay Geed's group to learn population MRI. And some years later when the children were six, we got the first pilot grant, pilot for a thousand MRIs. We completed that successful. Then we got money partly from the university, 4,000, not all of them successful. So most of the data will be three and a half thousand. And up to that time now overtaken by the ABCD, we were the biggest pre-adolescent imaging study in the world. And we have now completed three and a half thousand in the next wave, which would be 13, 14. I'll focus much of my results only at 10 wave of imaging because there we had a full, large part of the cohort. Obviously it's only half of all the people that ever participated, but quite a good response rate. And we had other measures which I'll mention on the way. So I'll start with, I squeezed in one more topic. I was gonna start only with depression, maternal depression. So we're looking at one of my research lines has always been prenatal exposures, rating the prenatal exposures to what happens to the child later, mostly neurodevelopmental being cognition or behavior or imaging. And this is the list of things we've studied. We're quite well published, for example, on things like thyroid, fatty acids. Cannabis use was, we had 200 women using cannabis during pregnancy, showing the effects, which are not that marked, so I won't tell you much about that. And so we have repeatedly measures, yeah, it's surprisingly little effect on depression is what I, and I squeezed in because we talked about two or three people I met today. We talked about family networks. So I have very unpublished recent work on family conflict which I squeezed into this presentation. So the depression work was coming from the idea that depression is much related, has been related many times prenatal or maternal depression at any time point to child development, not so much to brain development and what we have is quite unique data. We had it from different, here's obviously, the data, the questionnaires, I'll show you. So we have maternal depression data, which is in red, the red arrows at four different time points. So one is prenatal, then we have it actually, it should be at two months. So in the early postnatal phase, then we have it in childhood. I use the six year data and we have it essentially cross-sectional with the MRI data at age 10. And the first thing sort of, if you wanna think with me, what we're trying to do was the idea is there a time point when the depression is worth if you go by the MRI. So if you think when does maternal depression or when is it associated to any MRI changes? If it is, you could say prenatal, postnatal childhood. I'll give away, it's not at all time points. So it's only at some one time point. So you might think it's prenatal because that's when the brain grows fastest and this is a very important, the two months is a very important part for attachment. You could say this covers all the early childhood or this is the most recent, which is essentially the cross-sectional, could be the strongest association too. So that's what we looked at and we looked at it with just very global volumetric but also with DTI that is white matter tract measures. And that's just our mock scanner costs quite a bit of money actually but what am I showing it for? These are the sequences here I'm talking about the classical T1, this is a three Tesla but this is a T1 sequence for structural and the DTI. It's quite known to people just the classical. We also have a resting state which I'm gonna otherwise put it away. And here are the results. So if you look at it quickly, what you'll see is another probably more powerful pointer. What you'll see is we've got the different, four different time points, three months. It's actually three and six, it doesn't matter, I showed three here, nine, two months and postnatal and then you see it's, the first step is always for us to the sort of hierarchical look at the total and then zoom in so we look at total white matter and total gray and there's nothing much at all because white matter volumes and then there's something only here in the postnatal two phase. So indicating that if there's an association it would be just a postnatal two months. Now we look at that further, now comes the white matter tracts. So these are GlobalFA, very similar picture with MD. So GlobalFA, now you can look at it differently. I'll show you an easier picture that take away one of the models. You can see again it's very consistent the white matter tract is only at age two months. There's no effect. The other thing perhaps to note is that there is a lot of confounding so if you adjust for conflict and BMI and smoking during pregnancy you get rid of quite a few effects. That's what I think some studies do poorly but we found a very consistent effect on both the volumetrics and the DTI only with the early postnatal depression and that was part of the story which was a bit surprising. I would have, I don't know, I got my money with prenatal research so I always think it's prenatal. Often it is not, I can tell you that. And here again it is not, it is the early postnatal. Actually some child developmental people predicted I've given this talk before and the more you're in child development the better they can predict that. But then we thought the following went on the story a bit that really is this bit of a chance findings is really a pattern. Note that this was not a repeated measures analysis this was a simple one on one so one measure not adjusting so you could do that a bit more sophisticated. You could try trajectories, at least look at trajectories so we did trajectories to see what is the pattern you know are these the same women at different time points and that actually is a bit the story. So if you look at the trajectories one of them I'll warn you this purple one or whatever color that is is a bit unstable it's quite a small group but then there's a big group that has no symptoms as always there's some but then there's one group and that's the remark of a group that has clinical symptoms throughout so this is the cutoff of this specific measure of depressive symptoms the clinical cutoff is somewhere here. So these are women that are at every time point above the clinical cutoff so there's a small group only in Generation R but that's actually not so different than in other ages of women that have very high clearly always above the clinical cutoff. So three, two, three percent in the total population not those that are imaged but it's three, 3.5 percent that have high and these women who always have these high symptoms they just peak after pregnancy. So essentially it's that is the typical pattern and if we look at relating our outcomes for example these white matter tracks to these patterns essentially you have much less power because you just have four groups but you still find that that one fairly small group of being high and then going down a bit that explains what we find. So I would warn anybody to say it's just one time point what it really is it's probably the children of one group of women that have these slightly well altered or changed or perhaps differences in white matter and these mothers always have the symptoms they just peak after pregnancy. So that's a bit of a different interpretation which I think is a bit more solid and that in other words I think well this is where if you look at individual tracks which you could say it doesn't really I want to go into that I think this critical period approach which many people take nowadays and it's very fancy sensitive periods it does not work with depression. It does not work with depression because of the high correlation of the exposure at periods if you really are not adopted away let's say then likely that there is always something ongoing or that your mother has very few depressive symptoms. So there is a high if you wish carry over effect of the exposure among these periods. So I think if you want to study these sensitive periods you might do it with thyroid hormones and other things but not so much with depressive symptoms and then there's another interpretation which is a bit at odds with some of it but it also is fascinating to think there is some evidence that postnatal depression or postnatal depressive symptoms could be very different actually some say more genetic than childhood depression when you're very stressed by your job it's this very biological rollercoaster of your hormones that pushes some people who are vulnerable into depression. Hello, welcome. Welcome DD. No that was the first so I think that is that. Now I want to talk I didn't even have an introduction slide so it's a bit of a jump. I want to talk about the family conflict and that's just because I think it's a different picture to contrast that a different exposure. So we're talking now about a study of two and a half thousand very similar numbers. So two and a half thousand children again with imaging at age 10 and here we measured family conflict at three time points prenatally at age six that's why this time it is age six and age 10 and family conflict are things that's quite a long list of 12, 13 questions we can't trust each other or in family planning it's difficult because we always misunderstand each other that is not full blown conflict it's not like I beat my wife but actually we've shown that that is a very sensitive measure a very good predictor of divorce oh my God if you have that then every standard deviation will just give you quite a bit of divorce risk. Sometimes in hindsight I've just published that this year how much that predicts it's very, very remarkable that prenatal family conflict and it's actually very interesting we have it in mothers and fathers and I can tell you another thing the maternal measure is so much more predictive so if a mother or a woman in pregnancy says I can't trust my husband that's predictive of poor family and marriage and child outcomes so I was very interested in this measure in relation to child brain development again arguing that it captures it's not an etiological study but it captures many, many things but it's a very sensitive marker of problems in that and then I'm also interested I come to that later in a minute in the child behavior measured at the same time so this time in depression we didn't do that but in this paper I want to factor in the child behavior so here's again where we measured it so it's a bit of different time points we measured 10 to a minute other things but this is what I'm showing you it's a very, I don't want to spend too much time on the analysis perhaps important we do quite stringent correct for multiple testing quite a lot of confounders I want to point out that we control from maternal psychopathology some people think that's an intermediate of family problems that you get depressed because your husband is so ugly but actually you can also argue that you have many fights you know you're very neurotic and that's why you have family fights so in case of doubt I always adjust and you can take and pick it that's what I really want to show and then we look at gender effects both in parents and I'll quickly show you the results of to show you something so there's nothing much certainly if we correct for multiple testing we don't see these overall white matter and effects as we saw for depression it's perhaps it's well yeah but we do see no that was not there for cerebral white matter but we also had an idea that we looked at amygdala and hippocampus because this if it's early in life it could be affected and that's what you see so we rarely see a subcortical effects on the child brains at age 10 when we use measures at age five and at age nine which I think makes sense outside really severe child abuse there's very little idea that your hippocampus once you're age six seven will sort of shrink that much because of or grow less because of family conflict but this was quite a you know it's corrected for multiple testing it's quite a strong they should be 004 actually type of there so that's an and it shows this is the test of is are these different ages different the effects yes very much so just to show you a very different pattern how that contrast and it doesn't change whether you're correct for depression or not and then one thing what I fascinates me just to mention we have shown in a paper which I'm not presenting today that if anything it's more child behavior that predicts brain development if you measure it for example was white measure so it's that's our American Journal of Psychiatry paper where we show so again child behavior predicts child behavioral problems internalizing behavior with drawn behavior predicts how your white matter tracks develop rather than that your white matter tracks predict the change in behavior so since that paper I'm always putting be behavioral problems as a mediator of course that's a bit tricky because we don't know the hippocampal volume at all these different ages but again it shows that these probably either go parallel or perhaps that the child we don't know drives this but we can at least see that this conflict really to very strongly does is associated with behavioral problems very early and these correspond to at age 10 hippocampal it is not a sort of causal mediation analysis but a very interesting that it is corresponding not just to brain changes it does correspond to behavioral changes in these children too they go together that is family conflict I like the story actually even better than the depression so that's yeah that's just that's not important and perhaps one because we're on a Harvard student I want to show one result of a Harvard student now that I moved the rest is all Rotterdam student still that he analyzed the white matter tracks and again you see a very similar picture that the prenatal functioning rather than the mid-childhood family functioning is associated to global and of course once we see global we can then zoom in on the individual tracks but I'm always first interested because otherwise we do hundreds of tests is there an overall picture and many of these early life I don't expect them to be very specific I don't see why should something prenatal have an effect so I look at global effects first and you see it okay so I'll do I have enough time to do another jump do interrupt a few of urgent questions please do but otherwise I'll move on with another topic do imaging and then see if there's room for genetics I hope not I made a genetic talk so I've got a few handful of genetic slides at the end seriously, brain imaging and child safety so understanding externalizing problems so one of the things that I got to grant a few years back to relate externalizing problems to imaging essentially nearly cross-sectional it was a nice big grant so what you sometimes have the luxury and it works quite well if you put two PhD students next to each other on a similar projects because if they don't need to become friends but if they work together well it really does it's the output just explodes honestly that's my impression and so I had to but then the problem was that two things happened one is you have to have good ideas for two students you know just one idea won't do so you have to come up with more ideas and secondly we did a sort of pilot study and we just took the imaging parameters that's the sort of preamble of this talk and related it to externalizing symptoms as measured with the CBCL there was no association so essentially I had two students one idea and a negative finding and so we sat together and said that's not gonna work we have all room to do other things and just pretend it's externalizing problems or they said we have to dig deeper and it can't be we thought that there is nothing there so we had so one said we will split further the externalizing and the other said we will lump externalizing and I'll show you the two approaches also note that once we got going we contacted other people the imaging also who else? Jim Hujek I don't know what data he was working on and he also had found that nothing much was externalizing very simple in the general population we're not talking about conductors order in severe clinical cases what we're talking out is you know aggressive behavior in the general population where we see you know it doesn't correspond to big changes of the brain okay so then we had two approaches first we come to the lumping so that's Alex actually it's true story so Alex is the lumper and the background is the NIH sample imaging they found nothing much so we thought we'll start reconsider the phenotype we'll lump and what we did is we know and that's essentially the story of that is we started with this at that time quite a hype or still a hype the P factor hype and so everything is correlated with everything in the story so you know that you know that not only depression is correlated with anxiety or depressive symptoms as anxious symptoms but we also know that you know aggressive symptoms in the general population but also in clinical populations are associated with anxious symptoms so the idea essentially behind this is we don't find anything much because if we have a child who's externalizing he's probably also got some degree of anxiety so this is one of the classical figures but I'll show you the generation R model based on the following we have it based on the CBCL data first at age five but later we also combined two waves so essentially it's a cross sectional study but with behavioral data from two waves at age six and at age 10 and then if you have this typical behavioral phenotype what you get is externalizing behavior and internalizing behavior and that sort of loads on classical at that time that age attention later it is more rule breaking and aggressive behavior and then you've got these typical anxious or emotional reactive or withdrawn symptoms and then there's some others so and then what if you know the P-factor model what Ben Lehi introduced it he called it of course differently he called it the general general psychopathology factor which is a complicated name so Caspi coined it P-factor they said that we know that there is a relation between externalizing and internalizing and they said if you have another factor which loads essentially on all the different scales this is what happens so it's a general psychopathology factor loading on everything that is essentially one trait that describes your degree of vulnerability of symptoms of any psychiatric problem it's been studied by a few people and it sort of will show later but we also find that it's related to intelligence you can relate it's also been shown it's related to severe child abuse it's related to some very strongly for example to neuroticism and generally essentially severe the degree of psychiatric problems if you have that in the model note that it does not load on internalizing, externalizing however what happens and that's psychometrically not so easy to understand the meaning of internalizing and externalizing changes because essentially what you had now in the earlier model you had aggressive behavior and attention together is x into externalizing but now much of these problems is described by the general factor so this is the externalizing now is what is left after you've taken that out meaning it is what is left is the specific externalizing or it's something that is left and relates to aggression aggression once you've residualize or whatever you want to see it say it, taken out the total problems and let's take that forward so this is another factor from there's different variants from Ben Lehi this is one of the first where he did it in 2012 he had this general factor and externalizing and then he got two different internalizing but it's very much the same okay now let me just go through that quickly so we did that in a big group we also very carefully have multi-informants so we make it really theoretical now we use multi-formant to get a very specific and very stable these measures and this is a very complex model I don't want to go into it just to show you very quickly but it's really complex so it's nothing that sort of is suitable for individual diagnosis it's a very group latent variable describing the structure of psychopathology in the population and then you see what I find quite interesting how these factors this is the specific internalizing it's not the internalizing it's the specific so once the general factor is taken out say that again for example if you have negative affect it is still related to internalizing not at all to externalizing anymore that's what you really want to see that you have an externalizing factor that's not related to affectivity but it is related to strongly to something like surgency so I have argued and I don't know whom I told that today that it becomes like a bit of a oh yeah Dee Dee we spoke about that it becomes a temperamental trait to some extent I think that's sort of one of the best ways of in the CBCL this measure of child behavioral problems there is of course temperament in it and if you take out the problems you could argue that what you're left with here has something more it's not temperament merely but it is something more of temperament than psychiatric problem and what I was it's quite heritable we showed so this is a snip heritability which is comparatively high for child behavioral trait you can't compare it with this sort of twin study heritability so that's quite high it's quite remarkable we published that quite surprised how heritable it was and even a sort of smaller study and then we looked at the neurobiology the imaging and what do we find what would you predict we looked at the white matter integrity again so that's the bit of the theme of today with this externalizing but also with the general factor and it's 3,000 children so that's quite a sizable study where we could define the P factor of 3,000 children with MRI at age 10 and then also the and I told you when we just looked at internalizing or externalizing we found nothing with DTI just to repeat that and so this is the essentially model you've got on the one hand the white matter tracks and then you've got we actually at that age had to take out attention because it didn't load it was a separate factor so we had the general factor internalizing externalizing and also attention that's important to realize at older ages attention goes out of externalizing and this is the imaging protocol and then we find what we expected to the general factor is very significantly if you even if it just from any confounders related to that's what I'm trying to say less white matter integrity that's actually not so sensational not at all actually you know you've got this like sorry if you lump all the problems then you do find something okay but what I found I'm not saying it's sensational but it's quite remarkable is that externalizing we also find something but it's a different direction so all of a sudden you do find something for externalizing and that is mean to me it means that if we take artificial I admit artificial construct of pure externalizing without any of the problems which essentially doesn't exist in any one person but then we could see that there are two different things going on and that's probably why we find so little there are this sort of impulsive if you wish this sort of outgoing outwisness which goes together with more micro-archo structure and at the same time in these children if they are in real life they do have problems then you also see that they have something of the general factor and that goes the other way so essentially you could argue that's what we try to do it cancels out there's nothing much for it actually attention is the only different one if we take not specific attention but attention without that correction we do see less white matter that is the whole picture so it's the only one internalizing however we do it we find nothing okay so that summary of one approach that if you slump I would call that the lumping yeah slumping then you might understand better and then the other student is splitting so he had the generation died at the same ages and but he also went to he said I'll refine the externalizing phenotype and I know my mentor Frank Verhoel said when he heard that he said oh my god you're number 120 in my career that's trying to refine the externalizing phenotype I can't even stand it anymore and he I don't know he didn't want to he said it was just stupid but we pushed through anyway taking several this is quite a complicated study statistically so we took the data of different population registries who had the CBCL but also a clinical sample to show whether our sub phenotyping would hold and it's we showed it's called sort of a mixed factor it's a very complicated approach combining both late in classes but also different factor solutions but it's actually only the factors that do it well there's no classes much we get very much the same as the 120 before us got we got the well-known factors if you wish the physical aggression the rule breaking which is more the conduct this order or this yeah and the disobedience which is more the oppositional the only perhaps a bit interesting thing would be that we found the irritability as a very standalone dimension not as a jurist a jurist would say it's part of the oppositional we see it really as a standalone aggressive trait so that would be this you know emotionally reactive aggressive type so we have these four and now comes another question like was the depression if you have these four dimensions which or which one or more of these are related to white matter changes would be the next question so that was the question of Kuhn the second student so once he published this he immediately went forward and said which of them is related to white matter changes and let me tell you in this study of 2600 there's just one trait of the four that's related with a modest effect and that's the smallest group if you wish so less these prevalent trait is the rule-breaking is related so conduct is order not such a surprise the only frustration is that there's not so surprising finding which is you know nothing here and a bit here published most easily whereas the lumping is really people don't like it at all so you know I don't think he won the scientific debate but he certainly won the publication race so if you split you can split people have split before if you look at children with excellent traits and summary what does really change even if you don't take out the psychopathology what is related to changes in the brain is the very small group of children that have rule-breaking and then you must think of children that already at the age of 11 sort of skip class without the mothers knowing in generation that's a very small group that is truant at that age so that's that summary of rule-breaking is a rather severe uniquely associated as rule-breaking and that's what I haven't shown you it's nice that it's sort of consistent front of temporal tract perhaps what people would yeah not so surprising so this disruptive these different dimensions may help because in the general population I can tell you we and others have not found much with just taking the simple externalizing approach I'll show you one or two more results quickly rush through them because I have to do some genetics but you can have a question this is a good moment so that's a brilliant question and that'll fill the rest of the 20 minutes and I can avoid the genetics seriously it's in there's a few aspects of this one is power and again we talked about power I think population people completely get the power wrong in a way if they talk to people with case control studies because if you have two extremes of the you know it's not a normal distribution but it's sort of the skewed distribution but you take the very normal and the very extreme if you have 200 of that you probably have more power than our 2000 because you're contrasting to its extreme so I think for power we are overblowing the population base that's part of the answer so we're saying oh you have to do these big studies but really of course most people have a bit of symptoms or no symptoms and you have so many essentially controls essentially it's like a 10 to 1 design and even in GWAS they get it wrong I think it's very hard to do a good power calculation with a general population because the trait is so skewed and that makes it up now comes I think the beauty of these population approaches is in something else we can control for confounding much better and I do think that thus we can get rid of some of the background effects and perhaps I should start differently if you review the literature which we did for always do for these traits you know there is no one finding there's hundreds of findings or completely all over the place so if I sometimes do the joke and say can you find you know that autism is related to this part of the brain and you name it it's parietal it's oxidable they'll find a study so I think that is for several reasons it's multiple testing problems it's poor control for confounding and I think and that's what I would argue they don't have the possibility to it's not so much confounder control also it's to use other traits and control for anxiety because let's say if you have a control set that has no anxiety no aggression and you contrast it with those that have anxiety aggression attention problems you can't control for subtypes anymore because the one of the groups they empty cells and if you take Jay you know many of the designs we can show that that's not working other things are control for IQ or the whole debate that you know the contrasting groups have an IQ difference of 15 that's much more where I think many of the poorly designed case control studies struggle and that is why I think however the downside of population research is as I said it's power in a way that it's overblown and secondly of course we lose many of the severe kids to non-participation and if you have poor that then what are you talking about you know a tiny bit of aggression so it's not one or the other I think well-conducted needs goes for both otherwise it doesn't work so look at it carefully but I do think we need this population approach for something like a p-factor or whatever it is or controlling for anxiety or so yeah I don't know if that convinces you okay good then I'll have to go on so here we go so I'm talking about something else which is sort of a hobby from mine and I'll tell you a tiny bit the story was that in generation one of day we discovered that our ethical board said we have to do ethical consent after birth again because any consent for a non-living child does not go when you have a child and then you have to consent again and if you have to consent you have to do consent in face-to-face and we had no money to bring the people in so we were quite lost so if we had to send out a question there we would have to consent them so I was charged with solving that and they gave me a bit of money and so I said this is not enough money to go to anybody I can't do it and so I was lost and then I got money or we got money but I was in charge of it to get money for unemployed I remember that exactly for to hire people that have been unemployed for a while to go to the people's home so I was part of a training team to train I don't know honestly it was a team of six women elderly women mostly had been unemployed for years and then I said well if we do this and we had a good atmosphere I said I'll train you to do more I'll train you to do not only do the consent I'll train you to do home observations and I'll train you to do neurological examinations and I had a series of I remember very well I call it the pizza evenings because they always wanted to eat pizza which I don't like so we ate pizza together and I tried to train them with a friend of mine who was a physiotherapist in neurological examinations which was a lot of fun I can tell you so I had this team and actually they became quite reliable and they became wonderfully reliable I took them you know it was very very very nice and I learned to understand a slang of Rotterdam which I really you know never had the chance to hear so they did they went to four and a half thousand homes to do neurological examinations which each of them took 15 minutes and that was a beauty and they did home observations so they recorded and I sent them to the toilets to wash their hands and they recorded whether these toilets were filthy or not and that was quite a remarkable investment and I just had such fun with these teams so this is now years later revisiting that data in the following so then this is what they had to do you know they didn't learn these words so I had to explain that to them but they did all these examinations of tonus and you know move and we did it with dolls and we you know one of them I had some of the students even find small babies to train these things because they were elderly women and they had to train on somebody so you know they had to know to do this and they did it quite well so yeah we've got a good reliability so I we moved forwards and then now that I'm jumping to so when we took the genetic data note that in Rotterdam our genetic samples are always much smaller because we always have to exclude half the population which is non-dutch and then for some we don't have genetic data so if we do 4,000 very quickly with lost to follow-up it's not that big but mostly non-dutch which is exactly well not exactly half we're ending up with 1,500 where we still have genetic data and this analysis we did so I'm just showing a few things because I want to get to a different point really this is a nice publication we show that in a such a small sample we could show that the Schizophrenia PRS and Philip was saying everybody does PRS so we also did the PRS we did the PRS for Schizophrenia and we found an association and the interesting thing is was mania the PRS for mania was the other direction which is showing that Schizophrenia they were essentially neuromotor delayed with mania riskors they were neuromotor advanced which is actually not what I had expected but it was very clear so that was a nice story here is just the story I should have perhaps ordered it differently showing that the motor score if you have low motor scores gives you high symptoms if you have a good motor development you have low symptoms and that's very you know that stays with you so we found that the motor development at age six weeks reliably predicts internalizing not so much externalizing but internalizing problems is related to the genetics of Schizophrenia and here also to the autistic traits so we measured at age six this is we measured autistic traits and again we found that if you have a low muscle tone for example hypertonus low muscle tone your association it predicts if you wish autistic traits later in life I found that quite fascinating now comes something it becomes more and more complicated and then I'll show you something about yeah I'll have time about pleiotropy but here so we're still talking about motor development what we really measured is what did it quite well is the tonus then responses where you must think of reflexes and senses that sort of startle reflex that's also senses is eye how you know does it make eye contact we measured those things does it seem to hear and they get points for all of that and what you see is that a very different pattern you can't see anything here but that's because it's so light what you see if you do the ADHD risk score it predicts only those senses if you take the autism polygenic risk score it predicts the tonus very strongly and to something well mostly the tonus but what is then also interesting you can see that this polygenic risk score of autism predicts the motor development and later the autistic traits but also the ADHD Philip says it's related to everything indeed it's related to autistic traits too but we can even run a mediation model showing and that's a tricky what what does a mediation model mean here is that the risk of autism genes predicts the tonus and predicts the autistic if you put it into one of these classical mediation models you will find an indirect effect from autism via the motor development on that which would suggest and there are trials that make that plausible that if you have sort of poor motor development that either is an early phenotype of autism or it could mean that it is an risk factor that means it constitutes the development of withdrawn behavior which is a bit tricky but it's probably an early phenotype or it's part of a pliotropic effect that it does something with your neurodevelopment be it motor or be it so essentially if you take these models it could be the true sort of pliotropy it does motor and autism it could be that it does if this one or you know that one doesn't really matter if you've got the genetics it goes through the motor development and then later to autism and you can think of 20 other variations if you play around how it relates I don't really care but it you know essentially Philip would say publish as well I think what you learn from it is the serious thing is that of all the developmental measures I think it's quite remarkable the relation with autism and motor was the most striking let me summarize it that way and I do feel I don't think it's much I don't really know what it is whether it's this causal mediation but I do think it is and sort of an early symptom of this vulnerability later and that was quite remarkable I'll show you something else about the pliotropy and that's what I added because I'll tell you we were in a record store I remember having a coffee and we were talking about G by E and I promised you one day to show you the slides on my pliotropy which is sort of stalled but here is I discovered the slides I made and I never gave the talk before so it's a quite a risky thing and I have 10 minutes so here we go on pliotropy at that time what I was sort of hobbling around and see and hear what you think of genetics not show you more PRS I'll show you something very different so a fascination of mine has been the idea what you know as you notice from this thinking about that is how much of that is there and we've all got this cross disorder trait literature and all that so when you think of pliotropy you have genes that are associated to many outcomes and I'll show you what if you look at the way many people it's changing now I'm probably this is too late but what's been for a while what they have done is that they have done two G buses and both of them have a hit and then they say oh this is pliotropic you know that's interesting and that's what you see and then they sort of take for example the first papers from Smoller and Lancet he says every I'll take all those above the cutoff and I'll take all above the cutoff and then he sort of does a circle and says those overlapping ones are the pliotropy that's essentially descriptive pliotropy because what I would be much more interested is in something it's not these two it's something like that that there is a subthreshold hit which is nowhere near significance but it's subthreshold in two traits meaning that it's this is easier to discover but the question is how many of those are there so where there's a quite a clear signal far from significant and quite a clear signal and that is let me tell you if you look at the literature if you look at for example what they do very often is the Fisher methanolises of P values to combine two G buses to find pliotropic hits what they do is they what you find is such a heat map meaning that if if pliotropy is essentially pliotropic genetics is red and this is the P value of one trait this is the P value of other trait it's really driven by a significance you know on one of the two it's essentially driven I don't know what one and zero means I say I was take it as got lost here but what I mean is really let's say that would be P value is zero that would be very significant so essentially you want let me leave that it confuses me now what you want is not a pliotropic hit being defined by that the association is you know why it's significant and overwhelming in the other in one trait and absent in the other you will see if you calculate pliotropy what you want is a certain signal in both not just an overwhelming signal in one and very little in the other and that's what I try to depict here and I think I could get the story if I had three tries but I won't bore you as long as you follow me so what I'm not so interested in that you say let's say this is a genome-wide significant hit for schizophrenia which is wackingly significant some of them you'd easily find a significance of 10 to the power of minus 12 and then you say there's a bit in depression nothing much but together it becomes significant what you really want is a certain substantive signal in both of the traits and essentially you don't want and that's what I try to depict you want this area not just the extremes so if one player is close to zero to report significant part of the value from the second can take in value even one that's what I was trying to say it doesn't matter so I'll skip this it's just that not I'll skip this this is what it's about what what I thought of that we can simply do something very old you know when I did statistics things on a sort of not on a computer but on a calculator at that time I always did for some things some rank testing because very easy to do by hand so remember that the some ranking so essentially I said if we just simply some rank the test we could think of the ranks in two G buses and we could combine that and that's where my sort of knowledge ended that's what I want to do so I found a statistician in Rotterdam who did the math for me and said actually that would work as a pliotropy model that you do some ranking very simple but it works the first problem we had to overcome so what is the chance to have a snip with rank one and two if there are 10 million variants so if you have what is the probability of rank one and rank two and you know it's it's got to be the same so it doesn't matter which one so that's one thing you want if it's rank one and two or two and one should be the same probability so essentially you want the rank and you can do that and the first thing we came across of course that which makes it very tricky and much of the pliotropy anyway not only our model is the linkages equilibrium that you have quite of blocks of snips that go together so I'll tell you we did account for that we essentially took the rank we did something which is much more the rank of a block rather than a single snip otherwise you get quite confused so you saw that's based on the summary statistics so that would be the rank of that or you jump there so note that it does yeah you adapt that it doesn't the whole block gets the fifth rank not one two three but the whole block gets that rank of the middle so you do the blocks only you rank it that way and so it's I can I can skip the math first or because I don't understand it but you want to it is about two things this is the I would be the different snips and M would be the different different phenotypes you combine and we can not only combine two you could combine three so you could do depression anxiety schizophrenia you combine that and do the same theory but the interesting thing is you can boil that down to quite a simple well I I didn't I can understand that if somebody really takes me by the hand and through it but it is it boils down to a neat mathematical formula where you can take the ranking of these p values to the power of the phenotypes you're studying and then what we did is this is just the theory I'll show you examples in a minute and then finish off we get a very different heat distribution that was I was that's easier to explain where you see this is what you want you get a very neat very different distribution so that was simulated data where you see that it goes differently meaning that here you become trait so it becomes significant if there's a certain amount whatever that is of signal in each of the traits so you get much more hits of the the middle area so we worked was that is different and we did then simulated data so that we found simulated it with no hits so it didn't produce any artificial hits that's quite confusing comforting that's quite nice so one of the things always is you've got this wonderful method then you put non-informative data into it and then you do find many hits well we don't so that's good and then so we did that but then we did it with real data and there's an interesting case because Barbara Franke who's a friend of mine in Nijmiech another Dutch city she published that there was no overlap in subcortical structures and schizophrenia and there was no pliotropy which was a bit disappointing because everybody had hoped that that would be an endophenotype Philip and it seemed to be not or it was underpowered and we think we could show that it's a model problem because when we ran it we got quite a few hits you think why on earth would they be credible well there's many things you can argue with the biology I can just show you this is a fun exercise to think of pliotropy but the best thing is that this was quite remarkable that you know there was it was more intergenic than we would by far have expected by chance and the other thing of course which I love really I would point out that there's always not so much there's much more in schizophrenia and education that's also published much more pliotropy in different directions which is also fascinating to understand the brain if you wish I think that's the whole so the call here is and we can do that for different the Barbara Franka's cortical thickness schizophrenia we also did it for depression bipolar schizophrenia there we did not find that much more than has been described although we did find some other things and now I lost the slide which I had added somewhere this was the slide I wanted to add of that's the between the five different psychiatric phenotypes you find some that have been well described as being pliotropic very very well described we find with and if you go through them we have quite some reason to believe that they are not chance findings it's a very neat way a very simple way essentially of doing pliotropy and note there's very little sort of pliotropy GWAS which is quantifiable saying this is a you know this are the top hits so I want to go back that's it so as to Steven's question you know it's what is population population science what do you get from it well I think there's poor reproducibility of findings across studies not so sure we can address that fully but we can help poor generalizability to larger populations is one aspect which is not the same there is no developmental trajectories which we will increasingly now with repeated measures do and I think one of the big problems has been and that's what we what I didn't address my answer to you Steven is that I think we have overlooked seriously in non-population based approaches the effect of behavior on the brain and only with longitudinal data will we be able to show that and we show that that comes in my view from dementia studies where we always say well we've got dementia we've got the brain and then predicts dementia which is very true was very good approach but I think with five-year-old children that is different and there's a good evidence good reason good theory and now data to assume that you know and you see that from one of the studies I showed that if you are very very anxious your brain will subtly but it's all subtle will develop differently I don't know that much how that is for ADHD I doubt but for traits like aggression anxiety I would put my money it's reasonable and it's essentially just epidemiology to be honest and these are the two students so it's a true story he's published very well he's published okay but it's very sad he looks like that but that's it thank you very much I'm sorry about this yeah no you were perfectly on time perfectly on time I think we're trying to go on to Anna otherwise people... or mail or whatever oh no questions oh god come on then thank you very much all right thank you all very much