 Thank you very much for coming on a queue on the historian of the National Genomics Institute and I also helped manage the NHGRI history genomics program and this is the NHGRI history of genomics and molecular biology lecture series and the purpose of this lecture series is to invite really, really, really top-notch philosophers and historians of science, particularly of biology, of contemporary modern biology, with the purpose of getting them first more interested if they aren't already interested in genomics, also basically trying to increase engagement, collaboration, discussions between historians and philosophers of biology and the scientists and investigators whose work they are discussing and also to turn the NHGRI history genomics program into a kind of research node, research center for the study of contemporary sciences, contemporary biology genomics in particular. Hello, it's a great pleasure that I introduce the third speaker in this series who is Dr. Philippe Coulombe and Philippe is a philosopher and historian of biology, is the director of research at the Institut d'histoire et de philosophie du science at CNRS, Paris Saint-Saure-Monde. His research interests include the history of the modern synthesis, the history of human variation research and the changing understanding of genes and mechanistic explanations of biology. He also works on issues relating to the concept of organism and its role in current biology. His editor of the book series, History, Philosophy and Theory in the Life Sciences and has been funded by, among others, the France-Canada Research Fund for the Project Challenges in Evolutionary Theory, Development and Apparition, is the author or co-author of more than 40 peer-reviewed academic articles and nearly 50 book chapters. His lecture, which he is giving today, is the multi-passive legacy of the Human Genome Project a program for evolutionary biology. Philippe, the floor is yours. Okay, thank you. Thank you very much for the invitation. Actually, my research and my main competence is rather on evolutionary biology and its philosophy than on genomics. I am starting to look at things in genomics that are relevant to the questions I am generally asking on evolutionary biology. Some of the things I will say will be very tentative, especially about the role of genomics in the Human Genome Project. The context of my talk is in evolutionary biology, there has been a wide contestation of the classical evolutionary series in two decades. This contestation comes from various disciplines, evolutionary development, biology, molecular biology or ecology. And as a symptom that has been last year in nature, an interesting paper that was confronting opposed viewpoints on this question, does evolutionary theory need a rethink? And on the one hand, some people who were either from evolutionary development or biology or studying epigenetic inheritance or studying multi-level selection were saying, yes, evolutionary theory needs to be revisited and reconceived. While on the other hand, some people like Hopi-Huckstra or Dor-Dre and many others were saying, even though we know new things since the times of Meyer and Dobrzanski, we don't need to rebuild the theory. And actually, since at least two decades there has been publications and questions about should we really change our theory? So those are some of the books and papers that have been dealing with that. And the two recent ones from Baladistonia Sultan, Evolution and Development, that's mostly about plasticity, and Philosopher Denis Walsh, Organism, Agency and Evolution are trying to put forth an alternative to the modern synthesis. And actually, last November there was the Royal Society meeting devoted to precisely the current state of the evolutionary theory and the prospects of an alternative theory. The contestation doesn't really concern the fact. There is a wide agreement on a bunch of facts that are some cases of epigenetic transgenerational inheritance, the complexity of variation in general, the importance of phenotypic plasticity, the fact that there are selection at multiple levels, the role of developmental timing in evolution, the role of developmental processes in evolution, the role of neutral evolution. And at the scale of macroevolution, the role of mass extinction, the process of niche construction by which organisms are changing their environments and therefore changing the natural selection that they are undergoing, or the massive amount of lateral gene transfer in archaea and in prokaryotes. And the list is not exhaustive, but it's to give an idea about all those facts on which people agree, but they disagree on the theory. And more precisely, they agree about whether we should change the architecture of the modern theory to account for those facts highlighted in Blue Day, the facts that concern mostly are partly genetics and genomics. Some of the challengers of the classical evolutionary theory, which is the modern synthesis, so the body of theory emerging from first the synthesis of Mendelian genetic and Darwinian theory of natural evolution by natural selection. The challengers of the classical theory often say they are aiming at an extended theory of evolution, extended because it saw an extended synthesis. So it's forged on the phrase the modern synthesis that gave its name to the classical theory of evolution, and modern synthesis was the title of Huxley's book in 42. And so Gert Müller and Massimo Piliucci edited this volume, Evolution and the Extended Synthesis in 2011, which includes chapters on many of the facts that I was indicated before. And so extended because they want evolutionary theory to extend towards new facts and use new concepts to account for those facts. And the talk here will develop the idea that what's at stake here is the architecture of the theory that is aimed at by the challengers of the classical evolutionary theory. In this book Massimo Piliucci and Gert Müller, so Gert Müller is an evodivo researcher, Massimo Piliucci had been a researcher in plasticity for a long time, and they are trying to come up with a philosophical rationale of why we should change the architecture of the theory of evolution. And so it's in the introduction. I mean all the chapters are not, so they are very different and dealing with very different topics, but what have those chapters in common Müller and Piliucci say? Well, the modern synthesis was a statistical approach of the inter-generation change of trade values and early frequencies extended from micro-evolution to macro-evolution. It was a statistical approach because people who did the modern synthesis, so population geneticist Fischer and Wright and Haldane and then Dobrzanski and Mer and Simpson who extended the core views to other disciplines, they did not know the underlying mechanisms of variation and inheritance. But now that we have an understanding of the mechanistic underpinnings of the phenomena of variation and inheritance and understanding that's given in developmental and molecular biology, we should rebuild the theory accordingly. So what they say is the shift of emphasis from statistical correlation to mechanistic causation arguably represents the most critical change in evolutionary theory today. So the change of the architecture of this theory, they interpret it as a call for going from a statistical modeling of evolution to a mechanistic causal understanding of evolutionary change. So in my talk I'll try to question this claim actually. So first I'll talk about the modern synthesis and some of its challenges. Then I'll look more in detail about the role of the gene in the synthesis and then I'll address some aspect of what's been called the post-genomic turn and its relationship to the evolutionary theory and I'll develop a few epistemic aspects that I think are central in this turn. So what's the core of the modern synthesis? Even though the people from the modern synthesis were disagreeing on many, many issues. I mean Wright and Fisher were disagreeing on the role of selection and the targets of selection and the gradualism and many other things. Even though there is a core of the theory which might be summarized in this letter Huxley sent to Merah when they were preparing a volume of Centennial of Darwin that was supposed to summarize their theoretical views and Huxley said what we agree on and the message we want to convey is natural selection acting on the heritable variation provided by mutation and recombination of a Mendelian genetic constitution is the main agency of biological evolution. And by natural selection there were meaning of course you have random variation of heritable traits and those heritable traits are causally related to the differential predictions of trade bearers and that leads to change in trade frequencies and trade values in the population. So that's the way natural selection goes but of course I'll be quite quick on that. And more random variation if I refer to what Huxley was saying is provided by mutation and recombination and the fact that variation is random meaning it's not directly correlated to environmental parameters has been established or corroborated by Luria and Delbrug experiments in bacteria in the 50s and heritability is given by the fact that populations have a Mendelian genetic constitution so heredity is the subject of heredity is genes. So that's how the claims are should be cashed out more precisely which means that population genetics is at the center of the biological evolutionary theory it's the science of evolution by natural selection it shows how in Mendelian population natural selection somehow works I mean even small differences in fitness leads to fixation of alleles and changing of trade values. So population genetics is mostly statistical modeling it uses statistical concepts and in considered fitness as a probabilistic concept and models the evolutionary change in this way and the population genetics in the terms of Fisher and Wright very like Rathley said it analyzes the conditions of which the forces that acts on populations so selection, migration, mutation and drift how those forces act on population and in which conditions selection can dominate the process because selection is the most interesting to the extent that it's the only force among those that produces adaptation or design that was something Fisher or Wright or Harlein were agreeing and so those are the four causes of evolution, selection, migration, mutation and drift sometimes they are called forces and actually philosophers have been since a decade arguing whether selection and drift should be considered as forces or whether the world forces like superfluous whether you just have statistical correlations and there is a nice paper by David Dupuis, a recent paper who is analyzing how people from the modern synthesis have been saying the causes of evolution, then the forces of evolution then the factors of evolution and that those three words they were fluctuating between those three words to characterize the status of selection, mutation, migration and drift and the challenges to the modern synthesis are not at all new and actually the history of the modern synthesis to some extent is also the history of the challenges that researchers have raised against it and if I want to schematize things I would say there are issues about what is genuinely explaining evolution and then diversity of the living forms and unity of the living forms and issues about where should the explanation take place and if I want to sketch that the space of controversies I would say that there is one controversy about optimization or selection and should we start by assuming selection and the traits are here because of selection and so it's what's called the adaptationist controversies and there are various brands of adaptationism but I'm not interested in that here and the other line is the divide between whether organisms or genes is the level where the action takes place the action of selection takes place and actually all the combinations between the positions were possible within the modern synthesis so some people like Dawkins recently is the one who is both situating the action of selection at the level of genes and saying that the main force of explaining traits is selection so he's a pure adaptationist and on the other end of the lower case of the sketch Stephen Jay Gould would say well it's selection targets organisms or at higher levels species for example but not genes and you don't have only natural selection that explains traits and that's his famous critique of adaptationism in his paper about adaptationism is that you would say most of the traits in the living domain are mostly explained by natural selection so they are adaptations and it might be understood as an empirical claim about the majority of the traits or it might be understood more weekly as a methodological claim and actually that's what behavioral ecology is doing you assume that for example the foraging time of the flies is optimized by natural selection you do a model and then you predict the foraging time and you look in the field whether your data match your prediction so you first start as human adaptation and see what happens so it's methodological adaptationism and some people would say well it's mostly about genes but people like David Lynch who wrote the book about the architecture of the genome he's someone who says well it's not on... he says he's sort of paraphrasing Dabrzynski and he says well nothing in evolution makes sense except in the light of population genetics so for him genes is where the action takes place but he's not at all adaptationist he would say most of the architecture of the genome in your career in metazoan is due to drift rather than selection just to give a sketch of possible positions oh see that's the modern synthesis all that that's positions within the modern synthesis yeah yeah right absolutely so the thing is that within the the history of the modern synthesis the first answer is that for example it integrated neutralism neutral evolution yeah yeah yeah right but then the other issue actually is that there is in the classical theory there is behavioral ecology which is mostly adaptationist and population genetics which actually in population genetics you can have various outcomes depending whether drift or selection or mutation is dominating but there is nothing within population genetics that would force you to say well what's it really important is selection so there is a sort of tension within the modern synthesis between behavioral ecology which really centers on selection and population genetics which gives you like all the possible outcomes of evolution in possible world according to possible genetic make up and actually Fischer's theorem of natural selection a way to understand it was Fischer was really interested by design but then he was doing population genetics and which as such may not be directly telling you that the living domain is designed and organisms are well designed so the theorem is supposed to give you a direct link between adaptation and design on the one hand and selection in population genetics on the other hand so that's an internal tension within the synthesis so and the way to understand the history of controversies and about what's explaining so it's mostly the question about what's mostly explaining traits evolution or diversity is to think that you have like three poles of the main explainance the main things that explain it might be selection might be chance or it might be mutation or variation and so and actually the if you think of selection versus the neutral versus selectionist controversies whether a chance or selection is mostly explaining what you want to explain and that also goes on at the level of what Steve Gold was saying about the history of life in general he was saying the history of life at the level of clades or at a very large scale is mostly dominated by contingent large scale event so you would have chance you would have like selection at the level of species or organisms but chance at the level of the overall history of life if the atheroids hadn't he dears the dinos would still be there we wouldn't be there and so on so that's that and the other line of the triangle is selection versus mutations or the processes that give variation and I'll expand later on this difference and so that's something that goes on from the beginning of the century the Mendelian was saying while it's not selection that drives evolution it's rather the macro mutations and then at the times of the modern synthesis it was a ants mare and the people from the synthesis that goes against a goldsmith who had this theory of the hopeful monsters so macro mutations in the species are the variants that are not fit at all but if you have a large environmental change it would be great so and in the controversies between evodival and the modern synthesis there is also an emphasis on the mechanisms of variation it's not mutation here it's all the mechanisms of variation that might be more relevant to evolution than the modern synthesis sync so for example if you have a trait and the constraint on the variation of the straight are very severe like only a very small cluster of variants is possible actually the selection cannot do much and if the selection cannot do much what explains the straight is rather the constraint so the variation the mechanisms of variation so the challenge of the people from the evodivo is maybe development is more than just the expression of genes maybe and maybe they can directly produce adaptive phenotypes so I'll pick up that later and if you look at this controversy between selection and variation actually the philosophical issue here is that it's related to the philosophical issue about what does selection really explain so is selection explaining the arising of the traits or is it explaining the spreading of the traits because perfectly sound view would consist in saying the trait arises in one organism so there is an individual developmental process that makes a trait arise and then it may be successful or not but that the spreading of the trait not the origin of the trait so and against that actually the modern synthesis authors were enfasizing the creativity of natural selection so and they used the word which is a bit bizarre but it was more than a metaphor it was because they wanted to say that selection is more than a sieve it's more than a filter just filtering the unfit selection is cumulative selection so it is explaining complex adaptation it's taking one allele and another one each of them being slightly fitter than the other ones and by this cumulative process it makes new traits more adapted emerge that will not emerge with only the effect of mutations and variations so for them that I think one of the important thing in the modern synthesis was precisely that this opposition like variation which is the arising of the trait and selection which is spreading is not really relevant because selection is the sense of the modern synthesis is about the creation of complex adaptation by accumulation of small variations and so small effects allele basically so what Mueller and Piliucci were saying what they were saying that people from the modern synthesis are not have just statistical models but now we can have a mechanical understanding they are first rejecting this emphasis on the creativity of natural selection basically they take for granted that population genetics models are spreading of traits but not the mechanisms of how they emerge and so they say now we do know some of the mechanisms so a new synthesis is ripe and though the rest of the talk is about trying to say a few things about what post genomics is doing here in this sense of yeah sure sure sure selection itself so for example isn't it possible to say that the uterus itself is a site of selection and it is selecting against conceptuses that are whose constitution is not compatible with interior life are you trying to force development from selections on that? Development from inheritance actually so the development is the process through which the zygote goes into the becomes a reproducible adult and the claim that the modern synthesis is supposed to hold is that whatever the password from the zygote to the adult what counts is actually the number of you know like the fitness so the the offspring number in the left time of the individual so if you think in those terms whatever password from zygote to adult doesn't really provided that the number of offspring is the same doesn't impact natural selection because what counts is the number of offspring i mean that's maybe of course that's not so simple but that's the claim that at least the challenges of the modern synthesis are lending to the modern synthesis actually. This differential reproduction could have been shaped during the melt. Yeah exactly. It's just a window for manipulation and it also opens the window for change. Changes that aren't in the genes for the problem. But and actually just what I wanted to say was exactly on that so today they the fact that roughly said gene as it was considered like in the times of classical genetics up to like mono and a Jacob is has this double nature which is it's a key factor in development and so it has been formulated in terms of genetic program since the 60s and it's also the substrate of inheritance so we have them and they are the determinants in Mendelssohn's and then have been Christians genes and with age and then chance and distinguish between genotype and phenotype and it's the substrate of inheritance that concerns evolution the development stops at after one generation and oh sorry for and on those two sides have two distinct approaches. The first one about evolution. Well they say about development I inverted the slightest. It's at the individual level you decompose causal to causal factors of the process of evolution of development. So genes is a causal factor and environment is a causal factor and then you can find great environment in many of course into many layers whereas at the level of evolution from the side of evolution what's crucially important is to decompose phenotypic variance so you isolate the part of the iteration which is due to inheritance the one which is due to an environment like regarding one trait in one environment and then you of course they are like gene environment interactions variance but anyway what's important here is that the variance due to genetics and to genes and mostly variance due to additive mostly additive genetic variance is the thing that define heritability and that's interesting. And economy driven by theory or by availability of tools. By theory I think yeah but I mean when for example when Fischer was like writing forwarding the concept of heritability and additive genetic variance and the composition of the variance I mean it was his way of conceiving of how inheritance could be modeled within a in statistical terms and explain why small fitness ultimately explain why small fitness advantages can lead to evolution so why small fitness advantages can be retained generation after generations that's the key point and this is the concept to use it whereas in terms of development it's mostly yeah you have contributions to this the developmental process by environments and by genes and which are not additive and not additively interacting but the idea it's... Fischer knew how to use a scalpel. He still would have done variance component. I don't know. I mean it's very well actually I would say that all those developments are constrained by the techniques that are and that is true now and that was absolutely true then but and so while the first you know the Mendel and Johansson and the key steps in the history of the gene Fischer used the first of course the first classical Mendelian gene concept whereas the metaphor of the program and the genetic program in development that arose from the discovery of the DNA and actually we have lots of literature about the relation between those two concepts of gene but I really don't go into that and neither into the notion of information but the point is that there is no abuse translation from one to another kind of role so to the role of gene in development as a one factor in development that has to be composed as a factor and the genes are involved in the partition of the phenotypic variance it's not mostly translatable and it doesn't mean that the gene constructs everything in development for instance and reciprocally so and the consequence of that which is very clear in the literature of the modern synthesis so the Transky and Mer is that inheritance and development are distinct which historically is interesting because if you read people from the 19th century it would be the same for them like organisms develop and by developing they develop their parents and they develop into their descendants and inheritance which is focusing on traits and development which is focusing on organisms on one organism were the same in the 19th century whereas now they are very distinct and Ernst Mer for example developed it into the distinction he makes between this is very useful for us between the biology of ultimate causes and approximate causes so ultimate causes has to do with inheritance approximate causes it's the way organism function and develop and it doesn't concern the ancestral states of the population species and there is a history to be told about how this difference between development and inheritance that tied to the difference between partition variance versus compounding a kind of it's your incoming call that's okay okay okay so yeah so there's so did the so and actually when like recently someone really wants to say something to someone so yes so interestingly the modern synthesis was a synthesis between genetics and Darwinian selection evolution by natural selection whereas when people like Gilbert and Raph in a paper in like 20 years ago were among the first to say we should re-synthesize evolution and evolution theory and lever-bentale biology from the viewpoint of the medieval they call for a synthesis between precisely evolution so inheritance and development so the two natures of the gene that I was talking about so I had a small development and the fact that those two different uses of the gene natures of genes like in development or in inheritance may be traced back to two different families of meaning of the term causation so so A causes B may mean A makes a difference to B or A causes B may mean A produces B and so in the first case so you have like families of views in the philosophy of science the one which say A causes B means A makes a difference to B so it might be in terms for example of if A hadn't been there B wouldn't be there so it's a counter-factual conception you can say this in terms of probabilities and the process views are they are says that the causes are always embedded into the basic processes that define the world so each real process is defined by conservation of a quantity and so there are real processes and pseudo processes and that's very two distinct analyses of causation and I think that the difference making view correlates to the view that of the gene as something that is supposed to partition the phenotypic variants in the population whereas the process view is about what role a gene as a causal factor plays in a general process but and so the modern synthesis is tied to this epistemic heterogeneity between the causation in the sense of the gene as a unit of development and the causation in the sense of the gene as a unit of evolution and so and if you have this epistemic heterogeneity between two sense of causation you cannot make development relevant to evolution because it's two different notions of causation whereas in the extended synthesis the idea is that all the roles of the genes can be integrated into a general understanding of gene causation as a mechanism and that would bridge the gap between development and inheritance. What's some of the few things that are going on in the post-genomic term regarding especially the genes in development. I mean it's just very tentative and actually you know this like 10 times better than me but that's the kind of things that so sort I might look at it more in detail but the first move is that from the concept of gene action they use at the time of Morgan to gene networks so the initial view was that you have a gene and by some obscure ways in a given environment it contributes to give a protein and ten minutes. Okay so well and now oh sorry I read something in French I'm sorry but anyway it's the idea that we go through the well by recognizing they have like regulating and coding genes there is the idea of gene regulatory networks that have been very influential and very important here and I saw in this book so the Dalem workshop from 1891 in which many people like Davidson, Gould, Waldperth Bonner were involved so one of the first steps towards like integrating evolution and development, Davidson had a paper on so gene regulatory networks and it's clearly in the context of thinking about evolution so if we understand the genomic organization underlying this specific ontogenic regulatory pattern we could construct a theory of evolutionary invention at the DNA level and in a very recent paper that reconstruct the gene regulatory network of the skeletal maryclean genes they say so the gene regulatory network is supposed to be a total explanation of why the events of development occur and they are reusing the notion of genetic program actually they are reconstructing the network by perturbating gene expression and so they develop like several genomic programs for various stages and the functions of the development of the skeleton the skeletonic micromere and so we have like initiating specifications stabilization of the regulatory state and there are signal and repression of alternative fate but the point is that they are talking about the function of genes in the genomes but they are mostly not specific outputs they are transients because the function is depending upon at which stage of the activity of the gene regulatory networks you are looking at so the networks are very complex and what they are doing depends upon the state at which you are looking and so the complexity is such that you cannot focus on the specific causal action of a given allele and their context specific network defined effect rather than a specific causal role and the second thing that occurs in the post-genomic term even though it's more continuous process but while the modern synthesis was saying that it's mutation and recombination that gives you variation you have multiplication of kinds of variations so I switch that but so the c-value products or the g-values products are about a few genes can produce many variant transcripts and then many proteins so and then there is another shift the second one about the reverse one where many genomic networks can be actually like yell the same series of gene products of genomic expression profile so the number of genes on human genomes drops from millions to 20,000 after the completion of the human gene project and the linear DNA sequence gives rise to flexible interpretations through transplicing RNA, editing, translation and recording so have like many splice variants up to more than 3,000 so the abstract is that the gene to trade pathway is very much complicated so at the times of manual we have this idea that you go from genotype to phenotype in a many to one mapping and after the discovery of the structure of the DNA you have the idea that one DNA sequence is one gene but all those two equations are no more obvious and the genome is very flexible which gives rise to some challenges to the very notion of gene and philosophers are distinguishing very distinct ideas of gene whether it's nominal gene that doesn't sequence or the concept of gene that takes into account the splices and historically interestingly I think there was two important realizations that are quite inverse the first is the one that gave rise to the neutrality theory in population genetics where people realized when they got electrophoresis and they realized that the nucleotelic polymorphism was much higher than expected and they try to account for that in terms of neutral theory whereas in the 2000s it's like the reverse inverse phenomena the coding sequences are much less than expected and that gave rise to a criticism of the notion of gene and more to my point it implies that the variation is much more complex than mutation and recombination so that was the one which the two events that I was talking one that goes and favors explanation center of the chance and the recent increasing complexity of what is a gene that gave rise to that gave much weight to variation in evolution so I will be switching that and another point is that the genome functioning is much more extended than what used to be thought about genes so for example that in the genomic network and the gene regulatory network that's in a handbook on gene regulatory networks and they also says that the metabolites and proteins act as hidden variables there are nothing explicitly in the genes but the effect appears as edges between the observed variables so the general idea is that more weight is given to non-genomic elements to account for the dynamics of the cell basically and if we think of like pre-formationism and epigeneticism as two poles in the history of people who want to understand development so a pre-formation is that there is a form here that will be expressed and epigeneticism means there are physical processes that produce the outcome of development well the informational genes were pre-formationist clearly but what we know about the epigenetics goes into the size of epigeneticism so it fine-tunes the gene expression to set around the organismal environment but one of the lessons of what we know now is that genomes themselves are flexible and they are with all the genomic machinery that all those from various splicings and rene editing and so on they can be responding to state variation in the cell so the idea is that epigeneticism is much more than just epigenetics and Richard Burian was talking about molecular epigenetics just to describe how we know that the genome seems to function so I skipped that the two things I wanted to say the first one is that the new kinds of explanations I think so my point here is that it's not that we now we have the statistics we now know the mechanisms is that the very notion of mechanisms may be too cause going to understand what's going on here and our knowledge imply that we need new kind of explanation so I had this thing that we went from gene action to very complex gene networks exactly like in ecology we went to the notion of traffic chain to like wide traffic networks and Ulanovic who is an ecology specialist in traffic networks says something interesting he says network thinking has epistemic specificities and one of the specificity is that well the mechanism is too complex to identify all the causal role of all the nodes so what we want to do is to look at the topology and the general global topology properties and sometimes if we are looking from the topological properties we can know something about the dynamics so I think with the general regulatory networks there is something like this that goes that happens which is that topological explanations meaning inferring from topological properties of the network some consequences regarding the dynamics becomes more important and in those explanations we don't know exactly what the nodes are doing so what each gene is doing and what is its function in the same sense of its causal role in the network so I had an example about the reliability of information processing in network it's a paper by Clement Baldwin and they are showing that it's a very theoretical paper about network and they show how some topological structures of networks are more prone to robustness in terms of transmitting signals across variation and that's also a paper about Davidson and Peter from the Peter and Davidson about the regional which says that the rate of evolutionary change is tied to network topology that's the theory by Rue Allen and the fact that some very small motifs that are topological structure are highly conserved across many kinds of generator networks and so all these were going into the fact that it's not that we know more the mechanism that work is that we are using another kind of explanation that's more topological than mechanistic and the other point is that I think we are likely with the analysis we have to statistical analysis genome-wide analysis on genomes that we have to say more about the phenotypes even so that that was a paper about using whole genome sequence to predict quantitative trade phenotypes we are more able to predict phenotypes on the basis of the genome so and I think I'm almost done so the thing I wanted to say here is that even though at the level of mechanisms and causation you have traits are resulting from environment genes within environments so environment is crucially important in the projection of the phenotype in the mechanism to produce the phenotype here we may have statistical tools to from a very long sequence in high databases a very long sequence we can have tools to predict phenotype even if we don't know the causally interaction between phenotype and environment so the change here would be that we used to think about statistics in terms of we have statistical correlations and some of them are relevant important and significant and they are the ones that are about causation it might be that now we can have our best correlations with even if we don't know the causation that is underlying and we would be more predictive and actually I think that the same kind of turn is also going on in ecology but that was I just can discuss it afterwards but that was just a hint so and for me it's also the same thing is going well something also new is going on with metagenomics which is not decoupling the environment gene interactions in producing traits but decoupling genomes from species and organisms so but I'm very quick on that but and it can be very predictive to some extent so that was like the project of reconstructing the structure on the function of the global ocean microbiome by metagenomics of samples of ocean water okay so the claims here and just to summarize that the fine-graining of explanations it's not that we have a very good mechanistic science of variation and inheritance that Fisher Wright and Duprzynski didn't have it's we are fine-graining the explanation then we'll have a more complex image of what's going on and also a more diverse array of kinds of explanations so statistical topological accounts and mechanistic accounts the gene the causal roles the genes in the networks may not have identified functions either in the sense of selected effects of the genes or even in the sense of their their characteristic causal role in networks and this notion of gene itself many as many philosophers have said is not univocal and so and maybe this is the reason why we can integrate developmental biology within evolution then because we have one mod of mechanistic explanation into which we can integrate what was targeted by perpetuation genetics and the modern synthesis but because we have a variety of epistemic mods and that are that are arising now and we could think of a synthesis in terms of a synthesis between those epistemic mods okay and I stopped here because I was too long. Thank you very much. Thank you. Can I just push you on one point at the very end where you talked about statistical predictions of phenotype. And so clearly when we say we're going to predict the phenotype we would like to know can we use the information we have to say this is what the person or the fly is going to be like. When we have statistical predictions from quantitative data and GWAS data we're going to be wrong most of the time. Yeah. And so statistically on average we have an idea but I'm not sure that I would dress that up as statistically predicting a phenotype because it carries this weight I can statistically predict what your genotypes are just from looking at you would be right a lot. Those genotypes don't translate into phenotypes even though they are maybe influencing phenotype a little bit so I would just be careful about this new statistical predictions of phenotype because it implies a strength of association that may not really be there. Yeah. It is true that we can do it and we can actually mainly do it to find what genes might be associated with a phenotype even if they're really minimally associated. They actually just just wanted to but I mean you are perfectly right of course but I just wanted to talk about this paper that I've read they are contrasting their method to the GWAS methods and they are saying something like they they are saying something that they want to have an accurate prediction of quantitative trade phenotypes from the info so what they claim is something like they're going a step further to the GWAS analysis but I mean it's just a claim in this paper. The first statement would be really cool if we could do that. Yeah. I'm not sure. I'm not sure either. Their point is that they make they go further than the GWAS to do that because they capture effect of single variants that was wiped out in the GWAS method but then it's just one or two papers and they I used it because one of the if I do my parallel with ecology the thing that I think is going on is that we used to think that statistical predictions are reliable if actually there is causal relation behind that otherwise at some point they will be failing and in ecology some people are thinking that they are pushing the view of what they call predictive ecology so we don't have a theory of the mechanism that are going on who it's whom what are the the ecological characters of the species and what are the you know the generalized specialists and all the ecology of that we don't have a real theory of that we don't know the mechanisms but we might have reliable prediction for example if we have like two more degrees next year what would happen or if there is 10% of biodiversity loss what would happen we can have predictions and because even if we don't know the causes we have such a huge mass of data such a huge possibility of comparing this situation with other situations that are let's say our statistic will capture we don't know which but will capture real causal relations but we don't know which ones but we can know that we capture some real causal relations and then our collision will be reliable and our prediction will be reliable so in this case I think that what we are seeing is that we are moving away from the picture where the reliable correlations that you get from the statistics are the ones that are causal and we can attest it by seeing it pinpointing the causation to a view where actually we'll never be able to see what are the causal relations behind because they might be like a huge variety of distinct possible causal relations but that's about the same correlation that's something like this and we capture this I don't see them as major paradigms well actually that's I'm not sure I mean that's a possible interpretation of that and in the case of ecology I would say like you because I was involved in the study that actually you don't get so wide paradigmatic difference people who push for predictive ecology wants to get actually so I'm quite skeptical but that's the epistemic view that solved with this supposed change and I assume in genomics it might be the same so I have no idea if this is doing any good or not this microphone it's a placebo effect so two comments you made late one in response to Larry's question about diversity and generation of diversity and then on the structure and function of ocean microbiomes and I was particularly interested in the notion of the structure of the microbiome so my question is do you see significant implications for our approaches to systematics and taxonomy on the basis of what you're proposing here as an extended extension of the modern synthesis should we be thinking about about descent and classification in different ways as a result yeah actually let's say most of yeah I was saying there are many lines of challenges to the modern synthesis and what I was talking today was if we take this very simple distinction between patterns of evolution and processes of evolution was about what comes from the processes of evolution now what's going on here and that's exactly something people who are working on prokaryotes and the phallogen of prokaryotes are looking at is that things about the process are also very much more complex than what was expected so it's not like a nice tree but you have lateral gene transfer and it's reticulated at the level of prokaryotes at least or arches it's more as a the clear phallogenetic image is more like a net at least a bush or rather a network than a tree and that's something that I've been interested in because I'm working also with someone who is Eric Bethest who is really pushing for this view he's working on prokaryotes and their phallogeny but I think that something in metagenomics that is interesting is that it's not assuming the notions of organism or species I mean you start with the sample and you sample you sequence all the genes that are there independently of which gene belongs to whom and it means that you don't assume structures about what is an organism, what is an individual what is a species what are species genera and a tree and it's much more neutral regarding what's the pattern of evolution and so it might be that so I think I was citing this I was sorry I was so so quick but I was citing this paper for Doolittle and Zabayeva which says that metagenomics is not embedded within the view of species that is the classical systematic view I mean the units of biological organization is not assumed to be species genera and a lineage and they and this seems to be quite interesting and actually it might be the most genuine challenge today what in synthesis this I don't know whether I answer your question but can I push you on that and I'm going to give it to Adam of course so I think this is not read the paper and I kind of understand that what they're getting at is to understand the biotic system as a whole but I think metagenomics is a really awful model to deal with this right it's a tool that we use because we can and if Adam were able to do a survey of the ocean or wherever he wants to look and individually sequence the genome of every organism present by itself and also count how many of each one is present he would the problem is you can't do that economically right now and there the tool might or might not change this argument right so metagenomics we go and we scoop stuff up and we you're looking at genes in the system a gene in this organism and a gene in this organism look at them as somehow interacting without talking about the organism is the temporary artifact of when we do these things different from looking at the biotic whole and we just have this kind of weird tool that because we can because it's cool that we can use to get information that was invisible before at some point so that we understand the genomes of each of these organisms so I may be misinterpreting, I'm just reacting to the title in what I know of metagenomics which is this temporary messy tool which gives us stuff that we didn't know about you just want me to agree with you no there's nothing I agree it is currently messy and we were talking about this in the half an hour that we were discussing earlier you're missing the boundary of the cell wall that's compartmentalizing all of the biology that's happening within the cell and then seeing the interactions without recognizing that there's a boundary there but looking for pathways in a metagenome you could find a pathway and you don't know that the boundaries were there that might prevent it from happening but you saw it, it might give you a hypothesis you could go forward with but I agree entirely it's a messy view of the current state but yeah no I I'm not trying to tell you as a new way to interpret all of biology let's assume I know you're curious let's assume you find a couple of genes that say something about the way to mind the contrary but when you dig deeper you find out there an organism that was here and an organism that was here and they never beat and they never hear out and there might be 16 different spaces where this very weak connection metagenomically and when you find this you might put a lot of significance on that and you'd be wrong yeah no I perfectly agree I think that the one thing is the limits of the tool which are real and maybe overlooked and one other thing is like conceptually the fact that it's used to task that were not assumed to be part of genetics like for example in my in the ecology institute where I'm affiliated indirectly a lot of people are doing what they call environmental metagenomics so they are using it for example to metagenomic to say well you have a lake and you look at who is in the lake and then you use many other sources of information like the phylogenies you know and the systematics and it allows you to say something about the history of the lake and what's changing here is that you don't use genetics to understand how how the cell or the organism works you use it to describe something about to get informed about the history of a landscape and so that's the that's the shift that I wanted to emphasize I was looking at them as sensors exactly how it tells about what's going on it doesn't tell you the interesting thing is that there is something in the human genome project explaining why we are living the way we live so it's explaining the causes of organism development and functioning and those use of metagenomics it doesn't tell you anything about the causes it's just about describing what's there and because you know something about evolutionary history describing what has been there and so it shifts from like understanding the causes and the reasons to better describing things that are not necessarily organisms as a species or cells so yeah exactly yeah that's the point because the more you'll have your sample is from something that is structured the more your metagenomics analysis have chances to be flowed actually there's something like this because yeah I'm just going to say something quick and then we have to end it so one of the things that I would really like to see is this kind of conceptual discussion integrated with the tools and methods discussion a lot of this changes not necessarily because of the conceptual shift we have different better tools of investigating nature and I think that goes back to your point about fish are not as scalable so thank you very much well thank you