I’m really glad that you all could come tonight. We’re delighted to have a wonderful speaker tonight. And, this event, despite having empty chairs, this event sold out more quickly than any other event that I can remember or that, I think, Nicolas can remember either. In the first week, it was totally sold out. I guess a lot of people, maybe were deterred by the (inaudible), or something came up at the last minute. There was a lot of interest in this speaker and this topic. So, tonight’s event is sponsored by the UCSF Department Psychiatry, the UCSF of School of Nursing, the UCSF Depression Center, and the National Network of Depression Centers, which I’ll just explain a little bit. But, first of all, tonight’s speaker: The topic is “The Genetics of Personality of Depression: How Mental Health Researchers Try to Use Genetics to Understand Who We Are” And tonight’s speaker is Jonathan Flint. He received his medical degree and his postgraduate training at Oxford University in England. He’s an expert in the genetic and cellular molecular mechanisms in depression and how genetics interact with the environment to yield who we are. He completed, already, an amazing study that’s gotten a lot of press–that has Helped a lot of researchers. Sequencing DNA in more than 5,000 Chinese depressed individuals and 5,000 healthy individuals. It is the first study to find a link to a specific gene in depression. He’s also the Co-Director of a marvelous program at UCLA, called the UCLA Depression Grand Challenge. And, that aims to cut the economic and Health Impacts of Depression in half by 2050. He also is about to lead a 100,000 person investigation, “The Grand Challenge”, which will be the largest ever genetic study of a single disorder. So, that’s Dr. Flint. Just a bit about the NDC: National Network of Depression Centers has been in business for about five-or so-years. Ten years. Time flies when you’re having a good time. And, their goal is to foster connections and collaborations between the individual sites. To foster through the power of a network: greater research, education, and treatment of depression. Right now, they are 23 Centers of Excellence within the NDC. UCSF is proud to be one of them, among other top universities, such as Harvard, University of Pennsylvania, Stanford, UC San Diego, etc. So, but that is a background. I’m very glad and we very honored to introduce Dr. Jonathan Flint. Thank you for having me. So, I started my presentation by showing you this slide just to indicate this subject is of still great interest to the media and therefore used to the general public to show what is the genetic effects that make us what we are–personality, our mood. And, my task this evening is to trying to inculcate in you to some criticism towards some of these findings and make you a little skeptical about some of this journal and others might publish. And to do so, I’m going to have to teach you a little bit of Psychology, even less genetics, unfortunately a little bit of statistics. I’ll take you very easily with those subjects. But, most of what I’m going to tell you really, would be about personality. And, in order to do so, I thought this might be useful. So, this is a publication in a Railway Station, on sale to the general public. And, a couple of lessons itself self-scoring Personality Test. And if you notice the author down there, Victor, it tells you he’s the “Honorary International President of MENSA.” And, you might wonder why he makes it so obvious that he has that position. I think it’s because lots of us, when faced with something like a self-scoring personality test are a little bit dubious about what we might tell you. We all know that judging personalities is not so easy. Most of us have the experience of meeting somebody we think we like; taking them out to a dinner, and then realizing that we don’t like them. And even worse, you might be married to somebody and then realize they’re a psychopath. So, we don’t think a 5-minute self-scoring personality test is really not going to be much use. And, it won’t tell us anything right as much as our partner will tell us when we fail to do the washing-up, cleaning-up of the bedroom, and those sorts of arguments that go on with people open-up their hearts to each other. And, if you open up this magazine then, you’ll find that the questions are slightly strange. So what’s down here .. “Do you like to pick-up furry animals?” What is that going to tell you about their personality? Most of us would think about the problem in the following way. So, this one question: “Do you disregard other people’s feelings?” So, you have to answer “Yes” or “No” to that. Most of us will look at that and think, “Well you know, that depends?” If I want to impress you, you know, and I’m taking you out for a nice evening, then yeah–I’m going to my best to be nice to you. But, as a scientist if you’re standing in the way of a major publication and a scientific advance, I don’t really care what you think about me. I would do my best to disregard your feelings. There’s a situational element to the way we would answer these questions. And, I think that’s fundamentally why we are distrustful about this sort of approach. What I have to convince you is that this is not the way psychologist treat this data. For them, it doesn’t matter whether you tell the truth or not. Their interested in this data set because it’s a set of responses, which might be for example, predictive of whether you are suffering from a psychiatric illness or not. So, if I have a group of people who have schizophrenia and a group of people who don’t; and I give them this questionnaire. And, I find that the people who have schizophrenia are more likely to say “Yes”, they like to pick-up furry animals. I don’t really care whether that’s true or not. I just use it as an indicator. So, psychologists are using this because it tells them something useful. And, the other thing you should realize is that they don’t work necessarily with the individual items. So, the best way to show you this is to take two questions: Are your Feelings hurt? Do ideas run through your head, so you cannot sleep? Now, there is no logical reason why you say “Yes” to one question. We will say, “Yes”, to the second. But it so happens, I give this 7-questions to this audience or any other group of people, we will find that people are, in general, more likely to say “Yes” to this one and to that one. They come, as if in pairs. And, in fact, if you look across all of the questions, you’ll find this correlational structure. The correlational structure is what psychologists would call factor or personality factor. And, when that’s run through Victor’s string of set of questions, he pulls up something with his personality factor 2: Emotional Stability and gives you a definition of what that might be. I’m gonna refer that, not as emotional stability, but a rather older word, neuroticism (N). And, as I’ve said, we’d like to use this for a certain reason. And, we use it as a predictor of the onset of depression. So, people, when they’re given this questionnaire, they score on the (N) scale are more likely to get depressed than others. And, I also know that it’s subject to genetic effects that are shared with depression. That’s to say that whatever genetically is making people more likely to be depressed; it’s also increasing their risk about their score on the Depression Index. I’ll just continue for a few minutes more to talk about the ways that we assess personality and what it means. The Eysenck Personality Questionnaire (EPQ). This is what we use. We don’t use Victor’s one. If you read the instructions here, you’ll see, it picks up on the point that I made to you a moment ago. “Please answer each question by putting a circle around the ‘YES’ or the ‘NO.’ There are no right or wrong answers and no trick questions.” In fact, the no trick questions is not quite true, as I’ll show you in a moment. This business about there being no ‘Right’ or ‘Wrong’ is important. You’re not really concerned about the truth of it. “Work quickly. Do not think too long about the exact meaning.” So that’s essentially what we use. Here’s a questionnaire that we gave out; it’s been filled in by one of the subjects. It’s one of our studies. There’s 90-or-so items in the full questionnaire. And there’s one question down here, number 61. “Is or was your mother a good woman?” And, the person who filled out this questionnaire wrote this on the back Q. 61: “I found this very hard to answer. My mother was good, in that she provided us with clean clothes, food, a tidy and clean house, but she was and is emotionally cold and lacks empathy. I can understand why this is, but since warmth and empathy is so essential to the well being of children. I find it hard to deal with her now and over time, her cruelty seems to have increased. She had a hard childhood, years of abuse from my father. It’s not surprising. But hard to say whether she is good or not.” This is not a “YES” or “NO” answer. This is not the way you’re supposed to answer these questions. So, just to make it clear, you know, this is interesting but that’s not what we’re looking for. There are other questions in here which you might think of slightly strange. This is, again from EPQ. And, one of them is, questions like: “Are all your habits good or desirable?” “Have you ever taken anything, even a pin or button that belonged to someone else?” So, anyone who answers yes to either of those two questions Will score “1” on the Lies Scale. So, these are the trick questions, which I think tells you aren’t actually there. When he devised this, he thought it might be necessary to pick-up people who really were trying to cheat the system. In fact that turns not to be true. It picks-up something else, which is more like social conservatism. And, I’ll illustrate this to you. I’ll just show you data that I collected. This is from a small group of people. People were then working for me at the time. People were interested in this work. And asked for the questionnaires. I handed down a set of questionnaires. And in a group meeting, I presented the results to them, as I’m showing you here. I didn’t obviously say who had given what result. And I was just explaining the nature of personality and using the data to make these points. So, if there’s an (N) scale, that’s the neuroticism scale, and we’ll see (F) for Women here and the Male there. If you just cast your eye over these numbers. You don’t need to carry out any statistics. But you’ll see, that there are more scores over 10 in this group than there are down here. So, men are more neurotic than women. And then, I explained that there was the (E) scale, which is extraversion. People that like going out to parties and having fun. And, then there’s the (L) scale, the one that measures lies or or social conservatism. And, here’s the (P) scale. The scale that I set out to develop for measuring psychoticism. Psychoticism, as it called, another poorly defined feature, which usually the psychologists don’t use very much. If you look, most people score very low on this scale. When I was explaining this to the group of people who filled this out. But there’s one down there of 10, which is quite high. And, I was doing this, my colleague at the back of the room lifted up his hand, “Cool! I got 10 on the (P) scale!” I thought about that answer, and I thought, “If I had 10 on the (P) scale, I’m not sure I would want to broadcast it to the entire group. And that definitely made me think that maybe this is picking-up something of importance. But what that actually is, isn’t clear. So, that’s a brief Introduction to the Psychology of Personality. And, the important points to bear in mind is the questionnaires; in terms of the data they provide us are telling us about the predictive factor. And, what we’re trying to predict here, is what we learned from the (N) scale about the risk for depression. So, the second thing I need to tell you a little bit about, for those of you who don’t recognize what’s written here, it’s a DNA sequence. It’s just a little bit of genetics. So, you all have two copies of your genome. One comes from your father and one comes from your mother. And, the genome consists of a string of letters. Molecular Letters A,C,G, and T. And, you have a six billion. 3 billion from your father and 3 billion from your mother. And, I’ve shown you a very small number of them. In most cases, there are identical. The copy you get from your mother is identical. But there are places, which you should be able to see in the red just about here, where they differ. And, I’ve just shown you one difference. And, it’s random. Your father might have C, and your mother might have G. Or, that might not be the case. And, the consequence of it being random is that you can have either two G’s, two C’s, or you could have, as shown here, one of each. So, in a population, we have three possibilities for that particular variant. And, we call these genetic terms to be a genotype. So, there are three genotypes. And, each of those positions, which are variable in the genome. That’s really all the genetics you need to know and what we, as geneticists, want to know is whether any of those genetic changes predict the phenotypes we’re interested in. Personality. Depression. And, the way we do that is to simply ask: Do we find more people who are in, this case, severely affected with depression? Who have that genotype? And, unaffected people have this genotype? Is there some pattern that you can see? All you need to do, once you have that data is, run a statistical test. And, it’s straightforward to do and ask whether it’s significant or not. And, that’s something that even psychiatrists can do and have been doing for a large number of years. And, just to show you that: So, recently I took all of the Association Studies–that’s a technical term, to be asked just about the association between the genotype and the disorder. All the Association Studies in Psychiatric Genetics. And this axis is the number of publications per year. And, these are the years. So, people started doing this in, about, 1990. By about 2004, a bit earlier than that, about 2000 or so. And, it’s coming out about 1 per day of 365. And we’re now up to about 2, 500. So, that’s a huge amount of information that’s been generated using this research. Now, I’ve shown you, when I started, my first slide was to tell you this is a subject of interest. So, when one of these papers get published, the press picks up on them. The question is: “Surely, Dr. Flint, we’ve must have learned a lot if we have all of this information is pulled out.” I’m now going to tell you some very interesting things from this literature. So, I’m gonna give you a couple of stories. And, my stories are around something I’ve mentioned to you before, which is the personality trait neuroticism. and the other thing you need to know a little bit about is something called the serotonin transporter. And, the Serotonin Transporter is the target of a commonly used antidepressant, prozac– part of them molecule that it binds to. And, for that reason, is attracted to a lot of interests. And, people have wondered whether it might be involved in many psychiatric conditions that the drug itself affects. So, what is, what do we mean by the serotonin transporter and its genetics? So, this Is a little diagram, you know, fully into the details of how geneticists, molecular geneticists think about the Serotonin Transporter. So, the Serotonin Transporter is a protein expressed in the brain. And, every protein in the body is encoded by a gene. And gene’s have components to them. And, components, which are called exons; and there’s the first exon and more the gene is spread out along the chromosome. But, we’re not interested in that because the bit that the molecular biologists are interested is something is; is something in front of the gene. And, the front of of the gene is like the regulatory region–the thing that determines how much of that protein is being expressed. And what was noticed, some time ago, is that there was a small region here within this regulatory region, which varied. In other words, there was a little deletion or insertion in this regulatory region. And, that affected the amount of the protein that was being, at least the amount of, the message of the protein that was being produced. And this is a genetic change in the DNA. It’s just like what I showed you before–the G’s and the C’s. Except it’s, rather than affecting one, basically; it affects a number of these letters. There’s 44 of them. And, we can therefore express this as, in the following way. So, you could either have a short one or a long one, an insertion, or you can have a deletion. So, that works just like the G and C, that I showed you before. You have a probability of inheriting one or the other. So, we can therefore produce the three genotypes. So, that’s “l” for Long and “s” for Short. That’s somebody who’s got two short alleles, two long, or one of each. And, then what you can do, and this work was carried out some years ago; is see whether those three are associated significantly with our measure of neuroticism. So, you collect some data, as the number of people that fall into those categories. We score them with their neuroticism score. And then, we ask if there’s a difference between noted categories. And the answer is significant, by sine standards P-value. And, as a consequence, this was work done in the 1990’s. You will publish a paper saying there is an association of anxiety related traits. So, remember, neuroticism is a personality trait that predicts psychiatric conditions–anxiety and depression. Polymorphism, a sequence variant in the serotonin transporter gene regulatory region. And, this was published 20 years ago. 3,254 papers since then, have cited this paper. That’s to say, this has been a very influential piece of work. And, therefore, by some criterion; you think was a really important observation. Now, what I’m gonna do next is just try and think a little critically about how we evaluate that finding. And, to try and show you that things are not quite as simple as they appear. So, this came out in in 1996. And, I told you that doing this study is quite easy. I can give you the questionnaire. You can fill it out in about 2 or 3 minutes. You can all before, you leave the room, spit in a small pot and I can extract the DNA. I can genotype, get the information for this “s” and” l” polymorphism. It would take me about a day. What were the numbers here, I think it was …300-400. Well, we’re not quite up to that number in the audience, but I could probably get some passersby. And, we could certainly collect a sample of equivalent size within a few hours. And, if I did that, then I might publish a paper. Here is a series of papers published soon after. Three in 2001, and one in the year 1998. And, I’m showing you the number of people they recruited. And, this is the significant result that they get. And a P-value that is less than 0.05 is generally regarded as being significant. And, you can see this person found a significant result. This one wasn’t. This one wasn’t. This one wasn’t. So, I’ve headed this “Inconsistent Association Studies.” So, they’re not consistent with the the first papers some people are finding it. Some people aren’t finding it. It’s not so straightforward. So, our question then becomes: How do we know who’s right? Is there really an effect there? Is this song just noise? What’s going on? And, there are two ways we can answer this question. One is that we can take all of the published paper and we can ask: Is there some common pattern going on? And, the technical term for this, is to carry out a meta analysis. I’m showing you this picture–this is my friend Marcos Minako. He and i sat down looking at these data about, almost, 15 years ago now, to try and answer this question. We didn’t know how to carry out a meta-analysis. We knew there were such things but didn’t know the technicalities. So, we asked a few friends, worked with them, and we published a paper. And, in this meta analysis, we found that the evidence for the genetic association between N-Neuroticism and the Serotonin Transporter, that’s the gene I explained to you, and the evidence was that (nothing). So, that’s one way we can address this question. The other one is to say, “Let’s carry out a really big study and collect lots and lots of people and see whether we can see it.” So we embarked on these two large studies using EPQ. I, was at that stage, living sort of, here, in Oxford. And, we sent out questionnaires to 88,000 people. It’s very cheap to do; it’s just the post issue. You send out the questionnaires and they send them back. So, we got responses from 88,000 people. And, that’s shown here. Just to show you what the distribution looks like. This is the mean. And, I’ve centered this around zero to make it easy to see. And, those are the two extremes. Now, it turns out that you can get almost all of the genetic information by just getting the genotypes from the two extremes. You don’t need to get data from all 88,000 people. You do, just as well, just to go down to this lot and down this lot. So, we designed the study, whereby you wrote to the people here and here. We said, “Could you send us a cheek swab, so that we can genotype and carry out this very important study.” Let’s think about this …I didn’t … We are going to the most neurotic people in Western England–the most neurotic. And, we’re saying, “Please can you send me a cheek swab ?” Now, some of them did write to us … ‘I just received your request to a mouth swab. This is the straw. My holiday has just been cancelled, my son has influenza and my husband has left me. Now this from you! This is the END!” So, that tells you that we are getting the right people. They are very, very neurotic. So, we’re quite happy in terms of how this is working. Unfortunately, because they’re the very neurotic people, they’re less likely to take part. So, I’ve shown you what we’d get back. And, this is the more neurotic group and we’re suffering a bit from ascertainment. Still, we work it out and decide that it’s still a great part of the study. So, we analyzed the data and this is what we find. So, I’m showing you here: I’ve already pointed out to you that the scores in women are higher than men. So, I’m showing you separately men and women. The gray are female. And the darker ones, male. And, we’re looking for a difference in these three bars: ll/sl/ss. You can see that there is no difference. We had some colleagues who are doing this same thing. They had collected data from 33,000 men and women. Only, now this is the East of England. And, they again, sent out this Eysenck Personality Questionnaire. They got 20, 000 men and women who answered. And, we did the same thing and selected 5, 000 at the end. And, this is their result. So, we thought, that was pretty conclusive that it doesn’t have anything to do with anxiety and depression. But while we’re doing that work, it takes some time; and something else has happened. Which is that in 2003, this paper was published. Again, this is in a journal that all scientists want to have their name associated with. It’s called, “Science.” If you can can’t get published in “Science” than you want to have it published in its sister journal “Nature.” And, then you’ll be a famous scientist. This paper has been published in “Science” And, so, “Influence of Life Stress and Depression: Moderation by a Polymorphism in the 5-HTT Gene. Now, let me just explain what that means. So, this is what we’ve seen before; this is the Serotonin Transporter; so it’s the same gene, and depression is clear enough. But, this business about “Moderation” isn’t quite so obviously. I think the best way to explain that would show you the critical result from that paper. So, let’s talk through that survey. So, let’s start at this side: No ‘s’ and ‘l’ stand for the genotype. So, this is the s/s (short/short), s/l (short/long), and l/l (long/long). And, what we’ve said before, is that that the initial claim, is that differences between these genotypes was associated with increased risk for depression through neuroticism–or increase neuroticism scores. And, our data suggested that wasn’t the case. And you’re seeing here this vertical axis is written, “The probability of Major Depression or Major Depression Episode”–That really Is the neuroticism score. The higher your (N) score, the more likely you are to become depressed. Just think of this as the (N) score; it’s the same thing. And, the claim we saw earlier was that if you have higher, or more of this genotype, s/s; you’re more likely to get depressed. So, this is what study what we were looking at earlier? Now, the difference is the siler (inaudible) stuff on this on this graph. That’s to do with whether you have bad things happening to you, “Severe Maltreatment.” Whether “Probably Bad Things Happen to You.” And, on the left hand side, as I hope is true of all of you, “Nothing is Happened to You” And, what you see is that the difference between the genotypes depends on whether a “Bad Thing Happened to You.” In other words, if I carry out my study of the relationship between this Serotonin Transporter gene and depression or neuroticism are the group of people for whom nothing has happened–no bad things at all. And, there’s no difference. These three genotypes are exactly the same. But, if I do this on a group of people who have had “Bad Things Happen to Them”, then I do see a difference. So, what they’re saying is the genetic effect depends on the context–depends on whether bad thing happen to you. So, that was the observation and this looked extremely interesting. Now, you might begin to suspect where we’re going next. The first observation is that not everyone found this. Other people tried to do it, and there are problems here because it’s not quite the same as the study I’ve described before. So, let me Just pick up a couple of issues. So, this is a picture I showed you the first time, And, I’ve tried to simplify it. So I’ve made it very simple, I’ve said: “Depression” is the risk here and then we’ve this axis of “Bad Things Happening to You.” This is not anything Bad and this is Bad Stuff Happening. We said there are different genotypes. I simplified: They have three. I made it very simple and said, let’s just make it two. Here, you can see there’s a difference between these genotypes and that difference is constant. So, regardless of whether a “Bad Thing Happened to You” or not, we’re always see a genetic difference. Their claim is that’s not true–we don’t see this. So, here we see a genetic effect. Here, we see it dependent upon the environment. Just hold that picture in your head for for a moment. Now, the next picture … It’s splayed out and I’m just repeating, in a way, what the first figure showed you but just making it simple. And, seeing a bigger difference where there have allowed that to be a genetic effect here. So, that’s sort of the same as what they claim. But, what about if you saw this in your document; would you think that’s the same? It sort of levels-off here. What about if you saw something like this: They start the same and then they, sort of, diverge much more. Is that the same? o Or, even more strangely, what if it’s like this–it crosses over? All of those are possible and they all have different biological interpretations. So, if somebody carries out this study and they say, “I have a significant result, I find there’s a change of the genotype effect dependent on the life events. We need to ask whether it’s really the same thing because it might not be. If we plot out; it might not be. Let me just give you one example: So here is data from one of the papers that claimed they had found the same thing. And, “Oh! It doesn’t quite look right. It’s not what we were seeing before.” The genotypes are now written at the bottom here– the graph is different. And, this is where the likelihood of becoming depressed is. And, they now put the two options–low environmental risk and high environmental risk–they plot it out like that. It’s very hard to see whether that’s the same or not? So, what I’ll do is that I’ll plot that in the same way as the original study. So, I’m going to put on this Axis, the environmental risk. And on this Axis, the depression symptoms. And, this is what we plot out. And, you can see this is not the same as the initial graph–it crosses over–it’s not the same as that. But, the authors claim it is the same. They say, “We’ve replicated. We found the same thing.” Hmm? Hold on, I don’t think you have–this isn’t the same. So, how do we resolve all of this? Well, guess what? There are two ways we can do this. We can do the meta analysis or we could look at a large sample. So, I’ve shown you a picture of another scientist, not my colleague Marcus Munfao this time, but Kathleen Merikangas. And, I’m showing you this because she, and a colleague of hers, is in fact at UCSF, Neil Risch. And, also undertaken the meta-analysis and myself, my colleague, Marcus, and Kathleen, and her colleague Neil. We both ended up publishing papers. And, both of us, we found the same thing. You couldn’t see this effect. By the time you’ve taken all of these various complications into account, you couldn’t replicate it. And, as a consequence of that, a paper was published, summarizing the results. This was in “Science” again. It’s entitled, “Back to the Drawing Board for Psychiatric Genetics”, and it said there had been all of these papers published claiming an effect. And then, these spoilsports, Dr. Risch, and Dr. Flint, and Dr. Munfao, they say it’s not true and they’ve really made a mess and upset the psychiatrists–it was so terrible. What was terrible is that I got hate mail. I tell you, we had emails coming in from all over the place. You know saying, you really don’t know what you’re talking about. This can’t be right. This is just not a lot of truth there. We’ll continue this story in a moment. But, I now want to take my third component and tell you another way that people have thought about with these problems. So, this is another title of a paper with a slightly hard to understand subject matter: “Intermediate Phenotypes and Genetic Mechanisms of Psychiatric Disorders.” So, this is really dealing with subject of today’s presentation–the genetic underpinnings of what makes us healthy or not healthy psychiatrically. What does this mean: intermediate phenotypes and how is it related genetic mechanism? I’m going to explain this to you with a figure from that paper. It’s a very beautiful figure, I think; it has many colors in it. So, I’m a psychiatrist. And therefore, I work with conditions called schizophrenia and depression. And, can you see those are in gray? Smearing, poorly defined. You poor psychiatrists, you really don’t know what you’re talking about. You just chat to people. That’s how you get your information. This is not what you and I would call science. Now, there are others called psychologists. And, they work with things like episodic memory–complicated stuff. Emotional regulation And, note that these are better defined and brighter colors; that means they’re more important, than what the psychiatrists do. But then, the brain. The brain has bits labeled in Latin. This is getting much more scientific. Anyone who studies the brain itself clearly they’re getting to be a real scientist. And, the things that are measuring might be the imaging of the brain–the brain function itself. But they’re still not really the hardcore scientists. Oh no. The really hardcore scientists are down here … they’re the molecular people. Dealing with the molecules of life Itself, of DNA. Serotonin Transporter Gene–that’s a real molecule. And, there’s other genes here on chromosomes. So this figure has a message. And that message is that there’s a sort of pattern going through left to right. And, you can see that indicated by the arrows. That you start off with really hardcore science and it gets sort of blurred as you go through the brain. And then finally, you end up with the stuff psychologists use, which is probably okay. But, the last thing you want to deal with is the depression and schizophrenia because that’s so hard to work with. It’s much easier if you can work with this stuff; and even better, of course, if you can work on this side of the graph. So, what that means is if you are a scientist working this area, you really ought to be measuring things going on in the brain. And, that can help you with your genetics because that would be much more rooted. So, they describe this in the following way: So, the assumption of this intermediate phenotype strategy–that’s a big, pretty picture graph. Is that the genetic effects at the level of the brain are a more direct effect of genetic variation than is complex behavior. So, it’s much easier to work with those things and that’s the studies we should be carrying out. So, is this true? This is another paper in “Science”, my favorite journal. “Serotonin Transporter”, so you’ll know what that is now. “And the Response of the Human Amygdala”, so that’s a region of the brain that’s involved in emotional regulation. And, you can do a study where you can present to people a frightening stimulus– they use faces. And, then they can look at the activation of the amygdala. And, you will get out of an fMRI scan, and here is the data you get. So, we’re looking at little red dots here and here–this is in the amygdala. And, this is the response to the fearful stimulus. And what this group have done is they’ve genotyped two cohorts. And, if you look up here, you’ll recognize “s” for short, “l” for long. So, I hope you’re familiar now and you’ll know that’s the short and long alleles of the serotonin transporter. Trying to see if that gene–those genotypes–predict differences in the responses. Let’s see what they find. So here are our results: These are the genotypes for the short groups, and these are the scores. There’s a big score point tabulation and a smaller score there, for this image. And, then they work-up the P-value, which is less than .05, and they can publish. We’re trying to address the assumption behind this: Is whether the effect on this signal is bigger than the effect that it might be on personality? So, this is closer to the genes, it’s like really important to be be close, and therefore it should be a big effect. The people who wrote this paper didn’t report this particular fact. So, I’ll give it to you. So, what they say is that genetic effect is 28%–that’s to say–that 20% of the variation of their genes is due to this single gene. So, in theory, of course, that supports their hypothesis. And, it would suggest that we’re going to carry out studies like this, we really would be much better working with these sorts of phenotypes. I hope you’re beginning to realize what’s going to come next? So, let’s query a little bit this figure: Is it likely to be true? Now, we know the answer to this question because in the last 10 years or so, it’s been possible to, rather than interrogate a single gene; interrogate all of the genes at one go. And, therefore give us a very broad, very comprehensive picture of how genetic effects work. And, the way these results get presented are in diagram for the flightless. What we’re seeing here is the name of the phenotype. Its diseases that were looked at. And, these numbers here represent chromosomes. And, the vertical axis represents the likelihood that they m actually find something. And, where it turns green is an indication that the scientists are pretty convinced they’ve found something. This is a robust technology. And, let me just show you the results for the phenotype that we’ve been talking about a lot, Neuroticism. Here’s a scan for neuroticism, and it’s finding something here on chromosome-8, and chromosome-9? There’s other evidence, which I won’t go into to support this, but basically they find genes. So, now we know you can find things using this technology. But, the question that we’re interested in here; is not whether we can find things or not but how big is this effect? Is this one here explaining 25 % of the variation? Less? And, the one we’re really interested in is the things that affect the brain. So remember the claim is, that we look at things that affect the brain, we’ll find very big effects. So, that is now being mapped. This is another Genome-Wide association Study, Here’s another significant threshold here, and here’s a little peak and there’s some effects there. So, how big are those effects? So, in order to answer that question, I decided to get all of the data that’s been collected in the last 20 years. I took all of the studies, thousands of them, and I got the effect size and I plotted them out. Here is the result. So, this is the effect size here and this is the number of studies–thousands of them. So, what we’re looking for is something like 28%. That would be over here–just to make that clear. Nobody has found anything like that–not a sign of it. Absolutely nothing. So, I’m going to bring my talk to an end by asking, what I think is the most important question: I’ve showed you now, a number of stories, where people claim things, which would be very hard to substantiate. And, where a lot of the evidence suggests that it probably wasn’t right. So, I think we have to ask: Why? Why should publish their studies and make these claims? And, why should there be so much heat, so much disagreement? What is going on here? So, this is something that anyone can talk about. You don’t need to be a scientist. You just need to come up with some ideas as to what would make people come to different conclusions. And, I had this conversation with my colleague Marcus And, we thought that maybe what was important here was where you lived. Why do we think that? Well, because we know there are different scientific cultures, different countries, different Ways of thinking depending on where you’re brought-up. So, we wanted to test this hypothesis. So, in order to test this hypothesis, we took all of the literature that we could find. And, we asked: How likely is it that you have a positive result dependent on where you live? So, we take a measure of the effect: A bigger effect; more likely to be positive. So, we divide up the studies, in terms of the effect size. And, then we look where they were published. So, our “Measure of Effect Size” is called an Odds Ratio (OR). So, if you are in America, the Median was about 1.1. In Europe, it’s 0.96, and elsewhere 0.95. So, that means, that more likely, regardless of the study you do; if you’re in this country, you’re more likely to publish a positive result. Regardless of whether it’s true or not. So, where you live depends on whether you find something important. And then we asked, “Well, what might be explaining that?” So, we decided to workout a slightly more refined measure like this. What we came up with is what we called a Bias Index. And, I’ve defined it here. Basically, what we’ve done is we’ve taken the effect size, Odds Ratio (OR) and we’ve divided that OR by what everyone else found. So, if you 100 people did the study, something found 2, some people found 0. You take the mean of that; let’s say it’s 1. And, we take the mean of the individual countries and then we divide it by one by the other. So, if you overestimate, it will be over 1 and to underestimate less than 1. That’s nothing to say about the truth of the finding. It’s just taking the published studies. So here, we plotted this measure of bias against the amount of money that was going into science. And, there’s a really nice relationship. So, if governments pump money into science, they expect some return. So, if you really need to publish a positive finding; that’s what that’s telling us. So, that seemed to be one reason why people are publishing things which may not always be true. The second thing that’s very important is what’s called an “Impact Factor.” An Impact Factor is what a journal has–a journal publishes a finding. Other people, if they think it’s interesting, will cite that paper. And, more people will read it. And therefore, you can reach journal come up with a measure of the journal’s impact by looking at how many people cited the work. So, we put this into the context, if you publish something in “Science” or “Nature” (I told you those were important journals); the impact factor is about 30. But, if you publish in a journal of not very interesting science, the factor will be like 1.2 or something. So, we then publish our findings in important journals and look to see whether those relate to this bias index. So, this is our measure of bias on this index here. So, anything above 1 is biased in favor and below 1 is biased against. And, here is the impact factor. And, what you can see is that journals with bigger impact factors are publishing, in general, biased results. So, they’re showing papers with Odds Ratios, which are above what we would expect. This number, is like, 25, so if i told you there are not many journals, which are that high. So, this in fact, is “Nature” and “Science”–these are the really important journals. So, the really important journals are publishing biased results, which is slightly an unexpected conclusion. Now, the other thing that we did when we published this, is that we showed the results as people with the circles. And, the people with the circles, defines the size of the study. So, if you did a very big study, you’re going to get a very big circle. Then, if you did a very small study, then you get a little tiny circle. And, if you remember, I said there were two ways to determine the likelihood finding something that was true or not: To do a meta-analysis or a very big study. The bigger the circle, the better result–the more likely is is you get the right result. So, the other implication of this is: If you look at the big circles; they’re down here. And, the smaller circles are up here. So, that means that the really high profile journals are publishing biased results in small samples. The Least Likely to be the correct answer. So, why should this be? There’s a lot of discussion about this. And one of the reasons is that scientists, to make their career, really have to publish here. It’s really important for them. You won’t survive very long. If you’re publishing down here in these lovely flat journals, you’re not going to get a job. And, in order to get here, you have to have a really interesting finding. Something novel and interesting. That’s what the journals look at. So, you send a paper in and they’ll make a decision as to whether they think it’s interesting or not. And, you can see, the things most likely to get in are those which show a positive result and that’s interesting. Even if it is in a small sample and therefore less likely to be picked. So, there’s a issue in science, which is causing us some problems. And, there’s been a lot of discussion about this and we still don’t know how to overcome this bias. Something inherent in the way that science is carried out. And, I’ll finish with one last slide. And if you want to look at this in detail, I strongly recommend this. It’s freely available to download yourself. And, this is one title that doesn’t need explaining at all. It’s completely obvious as to what this is about: “Why Most Published Research Findings Are False.” And, I’ve given you a couple of hints today as to why that’s the case. And, I hope that by thinking through these issues, you won’t fall into the same pitfalls that many people do when they open up a newspaper and say, “Shock! Horror! New Genetic finding. This is the cause of Infidelity. This gene to blame.” You’ll remember my talk and and say, to yourself that, “Dr. Flint told me this was rubbish.” Thank you very much for your time. I have a few minutes for questions. Q: What about looking at people with diagnosed with clinical depression? A: So, that is something that my group has done. And, so the requirements for carrying out such a study are that you would need a very large sample size. Which is, in fact, what we did. So, I showed results for the personality trait Neuroticism. You need about 170,000 people to get that result. We didn’t need so many for depression–it’s a clinical disorder. So, in a sense, it’s a bit like doing the extremes. Remember I showed you, if you collect a large distribution, you get the information from the extremes. But, yes we have done that. And we do find genetic effects that we can replicate. Q: With Looking Like? A: It’s looking like we’re just getting the first clues. And, we know that stuff’s there but it’s a little too early to interpret–this was published last year. Q: You mentioned that meta analysis being one way of evaluating the results of your research findings. But, I guess, at least for studies on drugs, probably one of the most common meta analysis is the pooled study or pooled analysis. And, I got into an argument with a friend of mine who has a background in Math. And, he was saying that pooled analyses are statistically not very good. And, he pointed to me some journal articles that said the same thing. So, yeah, it’s not ideal. It has a big advantage in that it’s very cheap. You just need to pull the data off the internet and run the analysis yourself. Anyone with access to a laptop can do this. Because the studies are carried out in such different environments and there are so many biases that might be hidden; it can be hard to know whether the results we’ve got our robust or not. So, one of the things I didn’t go into is that you can carry out analysis to indicate whether the individual studies are heterogeneous or not. So, if you carry out a meta-analysis without taking into account the fact that maybe three courses of the studies you’ve looked at are carried out in women and the other quarter are carried out in men. And, I’ve shown you that there are differences between men and women in these scores. Then, you could be led to draw incorrect conclusions. So, you need to carry out tests to detect those differences. But often, the researchers themselves may not tell you. Or, they may not even know what those bias might be. So, it’s you know, if you got nothing else, it’s a good way to go, but ideally you should do a properly designed study. And, in this case, it would need a very large sample size. Q: So, maybe this is a very basic question for some, but what will be the next step after we come up with a study that is positive and we are confident it’s positive, when we find that these phenotype actually corresponds to these genes and we know that these genes are responsible. Then, what would be the next step? Like gene therapy? A: No. I don’t think that’s not gonna work. So, what we know about these conditions, and I emphasize this: Is that things are really complicated. It’s really hard to get yourself these results out at all. And, once you’ve got these results, the real problem is trying to put them into some biological framework. So, I’ve just shown you little pictures with lines on them. That’s really all they are–it’s just telling us we got a P-value. So, you’ve got a huge problem turning that into the effect of a gene. But let’s suppose you do those experiments and you can now say, “Well, I’ve got a series of genes that I think are involved in this phenotype.” You’ve then got come up with some hypothesis about why those genes are having no effect. The nice thing about genetics is that it’s hypothesis free. You make no assumptions about what the genes do. All we’re saying is there’s a genetic effect. But the bad things about genetics is that it doesn’t generate a any hypothesis for you–it’s entirely up to you to do that. So, you look at those genes, many of them have something like a weird number ZNF567orf63–meaningless. Generally, because people have no idea what those genes do. Just a bit of random sequence in the genome. So, you’ve been faced with a lot of work trying to understand what that gene does. There’s, lots of people around the world working systematically across each gene Doing mutation studies in mice and doing expression studies to try to get a broad picture, so that hopefully, by the time you do get to the stage we’re just about to get to now, You can take that information and begin to say, “Oh! I know what that gene does, it’s involved in the processing of this molecule in this neuron.” So, maybe what’s happening is that pathway is a little different in people with depression than without. And, then you can go and try and test that hypothesis. Unless it was, that you’ve done all of that work, and you now know that the genes you find are involved in determining how neurons get connected in the brain. There might be some developmental effect. You, then, like to know: Well, can I alter that? Can I change the way, so that I can improve my patient’s health. But 10-15 years, maybe longer? It’s a long struggle at the moment. (Inaudible bit of chatter between audience member and speaker before next question.) Q: The Pinepeck Journals you had over on the right-hand side with the small sample size. Has that been historically the same all along? Or, is that evolved over time that they just picked and chose the most exciting articles? Well, that’s a very interesting question, I don’t know? I think one of the issues has been that the studies of the sort need large sample sizes have really only come to the point where they can be submitted and published in these high profile journals relatively recently. So, certainly in psychiatry, psychiatry was, you know, a graveyard for careers. It’s the last place you wanted to do research. Particularly, if you worked in hardcore molecular genetics. Everyone would tell you 20 years ago; don’t go anywhere near it. You’re never gonna find anything. And, the solution has been needed to do these very large studies, so it’s taken quite a long time to get to that stage. What, I would hope would be the case, is that what I’ve shown is this sort of blip on the ground. That, if we look five years down the line, we’ll see that; now they’re only publishing these really well powered studies. And, that there’s some evidence that’s true, certainly in genetics. People now know the sorts of problems that I’ve described to you. And, people are now much more critical about what they’ll accept. But, it’s not true of all disciplines. I mean, certain areas of Psychology has suffered this with lots of bad press coming out about unreplicated findings and so on. The area that might be worth looking at is, it might be, would be in physics. Actually, Physics has had this problem. When physics moved away from just saying, “We know what an electron, you know, the basic structure of what a nucleus is. The experiments they now do, those things need enormous machines buried under mountains in Switzerland to get an answer. And, what they do, is they generate vast amounts of data, and colliding particles together and they’re getting a lot of data. And, they’re just looking for really, really rare events. They’re looking for what collision that’s produced to predict a particle. And, that doesn’t happen very often. So, they have to go through lots, and lots, and lots of data to try and find the result that either proves or disproves their hypothesis. And, I met the man who published, one of the people leading the study–finding the Boson–this was published as the “God particle”when it came out in the press. And, I was asking him how they know they found it? Because they have the same problem–they’ve got lots of data–sometimes it’s right and sometimes it’s not. How do they know it’s right? They have a criteria there; it’s called the Five Sigma Rule. Which means that the thing has to be more than five of these units away from what they’re observing. I said, “How did you come up .. how did you work out this rule? How do you know that’s the right one to apply? And, because he’s a physicist, I thought he would then write out lots of equations and explain to me unambiguously, ‘This must be the right answer.’ But, in fact, he didn’t do that. He just said, ‘Because three sigma didn’t work.’ It was just arrived empirically. So, they’ve had the same problems that geneticists have. So, maybe going back to your question, we went back and looked in the physics journals, you’d see the same sort of problem. That there were small sample sizes giving unreplicated results. And, that’s now being resolved. They now have a better threshold and they get better statistics to correct. Interesting question … I’ll think about that. Well, thank you very much for coming up. Sorry we couldn’t get anymore questions, but I need to get Dr. Flint to the Airport to go back to Los Angeles. Thank you all very much for coming.