Spanning the last twenty years, from my undergraduate studies at the University of Texas to the last six years as a faculty member at the University of Colorado, I have been intrigued by evolution—how it succeeds, and fails, in creating the biological complexity surrounding us. I am especially fascinated by how we can understand heritable behavioral conditions, particularly psychiatric disorders, from an evolutionary perspective. Yet, as with many journeys of discovery, the journey to find my own academic niche—from a PhD in social psychology to post-doctoral training in latent variable modeling in behavioral genetics to an increasing awareness that statistical genetics is central to answering my questions of interest—has not been direct. While my core interests have remained stable, my understanding of the best ways to pursue those interests has evolved. In particular, I have come to believe that direct measurement of genomic variation itself, rather than indirect inferences about it via latent variable modeling, offers the most hope to eventually understanding why heritable disorders exist from an evolutionary perspective, and what the genetic architectures of disorders tell us about their evolutionary roots. Below, I discuss these areas of research. None of the work I describe below would have been possible without the fantastic graduate students, postdoctoral fellows, and mentors I have been fortunate enough to work with and learn from over the years.


Areas of research

Understanding the genetic architecture of traits

The last forty years have brought a wealth of information on the importance of genetic contributions to individual differences in human traits. Twin and family studies provide evidence that most human phenotypes studied to date are heritable at ~.50 +/- .20 (PLOMIN et al. 2001). More recently, GWAS’s have succeeded in finding over 2,000 SNPs reliably associated with human traits (VISSCHER et al. 2012), and future studies using whole-exome and whole-genome sequence data will allow researchers to test the importance of rare variants or rare variant burdens. In the midst of this deluge of data, however, fundamental questions about the genetic architecture of heritable traits remain unanswered or poorly characterized, and important decisions about how limited resources should be invested would be impacted by answers to these questions. In particular, how much genetic variation can actually be ascribed to additive effects of alleles (VA)? What are the “allelic spectra” of heritable phenotypes—that is, how much trait variation is explained by each causal variant minor allele frequency (MAF) bin? And finally, to what degree do genetic variants that predict a trait in one population or ethnicity also predict that trait in different populations or ethnicities? Much of the recent research in our lab has been focused on using family data and publicly available genome-wide SNP data to gain much better insight into these fundamental questions about the genetic architecture of heritable traits.

One central issue we have grappled with is the degree to which genetic variation in traits is additive (due to the average effects of alleles taken individually) or non-additive (due to interactions between two or more alleles). The most commonly used methods in behavioral genetics (e.g., twin-only designs), are very poor at uncovering the genetic architecture of traits (COVENTRY and KELLER 2005; KELLER and COVENTRY 2005; KELLER et al. 2005; MEDLAND and KELLER 2009). We have developed several extended twin family models that are much more better at dissecting the genetic architecture of complex traits (KELLER et al. 2010b; KELLER et al. 2009b; MAES et al. 2009). We have used these designs to better estimate additive vs. non-additive genetic effects in several traits (HATEMI et al. 2010; KELLER et al. 2013). However, the data necessary to fit these models is difficult to come by, and it does not exist for most psychiatric disorders.

An alternative approach to understanding the additive genetic variation of traits, at least that which is tagged by common SNPs, is to use Genome-wide Complex Trait Analysis (GCTA) (YANG et al. 2010). This method determines the extent to which genetic similarity at measured SNPs is related to phenotypic similarity between all pairs of individuals in a sample. By treating SNPs as random effects, GCTA derives unbiased estimates of VA associated with common SNPs. Because GCTA gives an estimate of the VA due to common SNPs (which tag common, not rare, risk alleles), this method also helps elucidate a second major issue we have grappled with: the degree to which common vs. rare alleles influence trait variation. We have used GCTA to estimate the VA caused by common alleles for schizophrenia (DE CANDIA et al. 2013; LEE et al. 2012), personality traits (VERWEIJ et al. 2012), cardiovascular disease (SIMONSON et al. 2011), and alcohol dependence (PALMER et al, in prep). These studies show that ~one-third to half of genetic variation is both additive and caused by common causal variants. And we have used this method to estimate the genetic correlations between various psychiatric diseases (WRAY et al, in press), between smoking initiation and addiction (WILLS & KELLER, in prep), and for schizophrenia risk between individuals of African and European descent (DE CANDIA et al. 2013). I believe this latter publication, led by graduate student Teresa De Candia, is particularly important because there has been concern that GWAS results might be European-centric, given that most samples are of European origin. Our paper not only shows that there is substantial overlap of schizophrenia risk alleles across European and African descent populations, it also demonstrates a method that can be used to get at this question between any ethnicities for any trait.

We were recently awarded an R01 to develop methodological tools that use haplotype information latent in SNP data to give direct estimates of the allelic spectra underlying traits, and also to provide unbiased estimates of the total amount of VA (and not that due solely to common causal variants).


Evolution and psychiatric disorders

We have used modern evolutionary genetics theory and empirical observations to try to understand why genetic variants that predispose to psychiatric disorders have persisted across evolutionary time in the population. In a series of articles (CANNON and KELLER 2006; KELLER 2007b; KELLER 2008a; KELLER and MILLER 2006a; KELLER and MILLER 2006b) and chapters (KELLER 2005; KELLER 2007a; KELLER 2008b; KELLER et al. 2010a; VERWEIJ et al. 2012), we have argued that the prevailing view—that risk alleles persist because they confer some kind of counterbalancing benefit (termed “balancing selection” in evolutionary genetics)—is difficult to reconcile with a number of empirical observations and theoretical insights. Instead, we have argued that much of the risk of these disorders stems from the conglomerate effects of numerous, small effect mutations that every individual is predicted to carry to varying degrees. This hypothesis—the polygenic mutation-selection balance hypothesis—seems consistent with the data on psychiatric disorder prevalence rates, fitness costs, their allelic spectra, and the increased risks of psychiatric disorders with brain trauma, inbreeding, paternal age, and ionizing radiation. Consistent with this view, we have found that alleles that increase schizophrenia risk (KELLER et al. 2012), decrease IQ (HOWRIGAN et al, in prep), and decrease attractiveness in personality (VERWEIJ et al. 2012), are shifted toward recessivity, a likely consequence of purifying selection against them.

However, and importantly, while much evidence appears consistent with a mutational role in the genetic variation of traits, we have recently reported several findings that at face value appear inconsistent with predictions of a mutational model. For example, we have reported that several predictions of a polygenic mutation hypothesis are not born out for homosexuality (ZIETSCH et al. 2008), suggesting that perhaps balancing selection accounts for heritable variation in this trait. Similarly, contrary to a mutational argument for potential sexually selected traits in humans, we found no relationship, genetic or otherwise, between IQ and facial attractiveness (MITCHEM et al, accepted). Furthermore, along with colleagues at the University of Queensland, we have found that common causal variants play an important role in schizophrenia (LEE et al. 2012), explaining ~1/3 of the genetic variation. Moreover, many schizophrenia risk alleles appear to have arisen before divergence between African and European population and have persisted ever since (DE CANDIA et al. 2013). These observations for schizophrenia may suggest a role for balancing selection, but a more parsimonious explanation in my view is that many schizophrenia risk alleles have such small effects on overall risk that they behave nearly neutrally, can drift to high frequencies, and can persist for a long time in the population.


The effects of distal inbreeding on psychiatric disorders

While introducing a theoretical framework into an area of research is important, it is critical that the theory make testable predictions. Much of our recent work has involved testing whether psychiatric disorders show inbreeding depression. Inbreeding depression, which refers to reduced fitness among offspring of related parents, provides clues to the evolutionary genetics of traits. The risk alleles for traits showing inbreeding depression tend to have been under negative (purifying) selection ancestrally. Inbreeding in humans and non-human animals has traditionally been studied using known pedigrees. In practice, pedigree information is difficult to obtain, potentially unreliable, and rarely assessed for inbreeding arising from common ancestors who lived more than two or three generations in the past.  In two simulation studies (HOWRIGAN et al. 2011; KELLER et al. 2011), we demonstrate that very distal inbreeding (e.g., from a common ancestors up to ~50 generations in the past) can be reliably measured in ostensibly outbred samples using modern, dense SNP arrays.

Our simulation work suggested that very large samples (e.g., 10,000+ individuals) would be required to pick up distal inbreeding effects using SNP arrays. These numbers are impractical for any single lab to collect given the base rate of many of the disorders we are interested in (e.g., ~1% for schizophrenia) and due to the cost of collecting genotyping and interview data on an individual. Thus, our lab has organized two consortia to study the effects of distal inbreeding on schizophrenia and cognitive ability. The PGC Schizophrenia ROH subgroup was composed of 20 scientists from around the world, all of whom are also members of the Psychiatric GWAS Consortium. We have recently completed a project analyzing a schizophrenia case-control dataset of ~22,000 individuals. This study represents the largest and most powerful test to date on the effects of distal inbreeding on schizophrenia risk. As predicted, we found that distal inbreeding is indeed a significant risk factor for schizophrenia (KELLER et al. 2012).We have also organized a similar albeit smaller consortium of 12 scientists for studying the effects of distal inbreeding on cognitive ability. We currently have information on ~9000 individuals from eight different labs. Early results indicate that the effect trends in the predicted direction but is not significant (HOWRIGAN et al. in prep), although, as an interesting aside, we documented the first case of non-pathological paternal isodisomy on chromosome 2 in this sample (KELLER et al. 2009a). Finally, in a study of >8000 individuals, we found evidence that distal inbreeding was associated with less desirable personality traits (VERWEIJ et al. 2012).


Candidate gene-by-environment research

Candidate gene-by-environment interaction (GxE) research tests the hypothesis that the effect of some environmental variable on some outcome measure (e.g., depression) depends on a particular genetic polymorphism. Previous graduate student Laramie Duncan and I critically reviewed all GxE research in psychiatry conducted to date (DUNCAN and KELLER 2011). We argued that the false positive rate of published findings in the field may be much higher than the nominal type I error rate of .05 because sample sizes have typically been small, the appropriateness of multiple testing corrections has been difficult to verify, and the unpublished ‘file drawer’ of negative findings may be large. Furthermore, I have recently described why the typical way that gene-by-environment research attempts to statistically control for covariates is incorrect, and that many alternative explanations for GxE findings that investigators had thought were eliminated have not been (KELLER, 2013).


Evolutionary significance of normal low mood

Another avenue of research centers around the possible evolutionary functions of depressive symptoms. Like fever and pain - both unpleasant yet nevertheless adaptive reactions - depressive symptoms (e.g., fatigue, crying, sadness) may be adaptive reactions crafted by natural selection to deal with difficult situations (KELLER et al. 2007; KELLER and NESSE 2005; KELLER and NESSE 2006). There is no claim that major depressive disorder (i.e., severe and protracted depressive symptoms) is adaptive, only that the symptoms of depression, expressed in the appropriate contexts and at normal levels, may have served functions in ancestral environments.


Impact of Research

The Brain and Behavioral Sciences article published by Keller & Miller (2006) was named an “Editor’s Choice” article, a distinction given to one or two articles that the editors of the journal consider particularly important each year. Our article that appeared in the American Journal of Psychiatry (KELLER et al. 2007) was selected by the editors of that journal as a “Notable Paper of 2007.” The article “Are extended twin family designs worth the trouble? A comparison of the bias, precision, and accuracy of parameters estimated in four twin family models” (KELLER et al. 2010b) won the Fulker Award for best paper published in Behavior Genetics in 2010. Duncan & Keller (2011) has garnered much attention, and two journals (Behavior Genetics and the Journal of Abnormal Child Psychology) have used our findings among others to justify new policies outlining stricter criteria that must be met before manuscripts reporting candidate gene main effects or interactions will be considered for review. In 2012, I was awarded the Fuller/Scott early career award by the Behavior Genetics Association. We have published papers in many of the top journals in genetics, including PLoS Genetics (impact factor, IF=8.7), American Journal of Human Genetics (IF=11.2), Genetics (IF=4.2), Nature Genetics (IF=35.2), Evolution (IF=4.9), American Journal of Psychiatry (IF=14.7), Biological Psychiatry (IF=9.8), Psychological Science (IF=4.5), and Brain and Behavioral Sciences (IF=18.6). Overall, 21 of my articles have been cited 21 or more times (h-index=21).


Grant Support

NIMH K01 MH085812 (PI: Keller)   1/01/2010 – 12/31/2014  6.00 academic
NIH/NIMH     $824,000 (total direct) 3.00 summer
Evolutionary Roles of Homozygosity & Copy Number Variation in Mental Disorders
This application proposes to use dense whole-genome SNP data to detect distal inbreeding effects on the risk for psychiatric disorders.


NIMH 1R01MH100141 (PI: Keller)   2/1/2014-1/31/2018  2.25 academic
NIH/NIMH     $1,605,455 (total direct) 3.00 summer
Estimating the frequencies and population specificities of risk alleles
This application develops methods that use haplotypic information from genome-wide data to estimate the additive genetic variation and allelic spectra underlying complex traits.




CANNON, T. D., and M. C. KELLER, 2006 Endophenotypes in genetic analyses of mental disorders. Annual Review of Clinical Psychology 2: 267-290.

COVENTRY, W. L., and M. C. KELLER, 2005 Estimating the extent of parameter bias in the classical twin design: A comparison of parameter estimates from extended twin-family and classical twin designs. Twin Research and Human Genetics 8: 214-223.

DE CANDIA, T., S. H. LEE, N. R. YANG, B. L. BROWNING, P. V. GEJMAN, D. F. LEVINSON, . . . M. C. KELLER, 2013 Additive genetic variation in schizophrenia risk is shared by populations of African and European descent. American Journal of Human Genetics.

DUNCAN, L. E., and M. C. KELLER, 2011 A critical review of the first ten years of candidate gene-by-environment interaction research in psychiatry. American Journal of Psychiatry 168: 1041-1049.

HATEMI, P. K., J. R. HIBBING, S. E. MEDLAND, M. C. KELLER, J. R. ALFORD, K. B. SMITH, . . . L. J. EAVES, 2010 Not by Twins Alone: Using the Extended Family Design to Investigate Genetic Influence on Political Beliefs. American Journal of Political Science 54: 798-814.

HOWRIGAN, D. P., G. DAVIES, B. M. NEALE, A. F. MCRAE, S. E. HARRIS, N. G. MARTIN, . . . M. C. KELLER, in prep The effects of genome-wide autozygosity on cognitive ability.

HOWRIGAN, D. P., M. A. SIMONSON and M. C. KELLER, 2011 Detecting autozygosity through runs of homozygosity: A comparison of three autozygosity detection algorithms. BMC Genomics 12: 460-475.

KELLER, M. C., 2004 Evolutionary theories of schizophrenia must ultimately explain the genes that predispose to it. Behavioral and Brain Sciences, 27, 861-862.

KELLER, M. C. Gene-by-environment interaction studies have not properly controlled for potential confounders: The problem and the (simple) solution. Biological Psychiatric accepted.

KELLER, M. C., 2007a The role of mutations in human mating in Mating intelligence: Theoretical, experimental, and differential perspectives, edited by G. GEHER and G. F. MILLER. Erlbaum, Mahwah, NJ.

KELLER, M. C., 2007b Standards of evidence in the nascent field of evolutionary behavioral genetics. European Journal of Personality 21: 608-610.

KELLER, M. C., 2008a The evolutionary persistence of genes that increase mental disorders risk. Current Directions in Psychological Science 17: 395-399.

KELLER, M. C., 2008b Problems with the imprinting hypothesis of schizophrenia [commentary]. Behavioral and Brain Sciences, 31, 241-320.

KELLER, M. C., and W. L. COVENTRY, 2005 Quantifying and addressing parameter indeterminacy in the classical twin design. Twin Research and Human Genetics 8: 201-213.

KELLER, M. C., W. L. COVENTRY, A. C. HEATH and N. G. MARTIN, 2005 Widespread evidence for genetic non-additivity in Cloninger's and Eysenck's Personality dimensions using a twins plus sibling design. Behavior Genetics, 35, 707-721.

KELLER, M. C., C. E. GARVER-APGAR, M. J. WRIGHT, N. G. MARTIN, R. P. CORLEY, M. C. STALLINGS, . . . B. P. ZIETSCH, 2013 The genetic correlation between height and IQ: shared genes or assortative mating? PLoS Genetics 9.

KELLER, M. C., D. P. HOWRIGAN and M. A. SIMONSON, 2010a Theory and methods in evolutionary behavioral genetics in Evolution of personality and individual differences, edited by D. BUSS and P. HAWLEY. Oxford Press, Oxford.

KELLER, M. C., A. F. MCRAE, J. M. MCGAUGHRAN, P. M. VISSCHER, N. G. MARTIN and G. W. MONTGOMERY, 2009a Non-pathological paternal isodisomy of chromosome 2 detected from a genome-wide SNP scan. Am J Med Genet A 149A: 1823-1826.

KELLER, M. C., S. E. MEDLAND and L. E. DUNCAN, 2010b Are extended twin family designs worth the trouble? A comparison of the bias, precision, and accuracy of parameters estimated in four twin family models. Behav Genet 40: 377-393.

KELLER, M. C., S. E. MEDLAND, L. E. DUNCAN, P. K. HATEMI, M. C. NEALE, H. M. M. MAES and L. J. EAVES, 2009b Modeling extended twin family data I: Description of the Cascade model. Twin Res Hum Genet 12: 8-18.

KELLER, M. C., and G. F. MILLER, 2006a An evolutionary framework for mental disorders: Integrating adaptationist and evolutionary genetic models. Behavioral and Brain Sciences 29: 429-441.

KELLER, M. C., and G. F. MILLER, 2006b Resolving the paradox of common, harmful, heritable mental disorders: Which evolutionary genetic models work best? Behavioral and Brain Sciences 29: 385-452.

KELLER, M. C., M. C. NEALE and K. S. KENDLER, 2007 Association of diffrent adverse life events with distinct patterns of depressive symptoms. American Journal of Psychiatry 164: 1521-1529.

KELLER, M. C., and R. M. NESSE, 2005 Is low mood an adaptation? Evidence for subtypes with symptoms that match precipitants. Journal of Affective Disorders 86: 27-35.

KELLER, M. C., and R. M. NESSE, 2006 The evolutionary significance of depressive symptoms: Different adverse situations lead to different depressive symptoms patterns. Journal of Personality and Social Psychology 91: 316-330.

KELLER, M. C., M. A. SIMONSON, S. RIPKE, B. M. NEALE, P. V. GEJMAN, D. P. HOWRIGAN, . . . P. F. SULLIVAN, 2012 Runs of homozygosity implicate autozygosity as a schizophrenia risk factor. PLoS Genet 8: e1002656.

KELLER, M. C., P. M. VISSCHER and M. E. GODDARD, 2011 Quantification of inbreeding due to distant ancestors and its detection using dense SNP data. Genetics 189: 237-249.

LEE, S. H., T. R. DECANDIA, S. RIPKE, J. YANG, P. F. SULLIVAN, M. E. GODDARD, . . . N. R. WRAY, 2012 Estimating the proportion of variation in susceptibility to schizophrenia captured by common SNPs. Nat Genet 44: 247-250.

MAES, H. M. M., M. C. NEALE, S. E. MEDLAND, M. C. KELLER, N. G. MARTIN, A. C. HEATH and L. J. EAVES, 2009 Flexible Mx specifications of various extended twin kinship designs. Twin Res Hum Genet 12: 26-34.

MEDLAND, S. E., and M. C. KELLER, 2009 Modeling extended twin family data II: Power associated with different family structures. Twin Res Hum Genet 12: 19-25.

PLOMIN, R., J. C. DEFRIES, G. E. MCCLEARN and P. MCGUFFIN, 2001 Behavioral genetics. Worth Publishers, New York.

SIMONSON, M. A., A. G. WILLS, M. C. KELLER and M. B. MCQUEEN, 2011 Recent methods for polygenic analysis of genome-wide data implicate an important effect of common variants on cardiovascular disease risk. BMC Med Genet 12: 146.

VERWEIJ, K. J., J. YANG, J. LAHTI, J. VEIJOLA, M. HINTSANEN, L. PULKKI-RABACK, . . . B. P. ZIETSCH, 2012 Maintenance of genetic variation in human personality: testing evolutionary models by estimating heritability due to common causal variants and investigating the effect of distant inbreeding. Evolution 66: 3238-3251.

VISSCHER, P. M., M. A. BROWN, M. I. MCCARTHY and J. YANG, 2012 Five years of GWAS discovery. Am J Hum Genet 90: 7-24.

YANG, J., B. BENYAMIN, B. P. MCEVOY, S. GORDON, A. K. HENDERS, D. R. NYHOLT, . . . P. M. VISSCHER, 2010 Common SNPs explain a large proportion of the heritability for human height. Nat Genet 42: 565-569.

ZIETSCH, B. P., K. I. MORLEY, S. N. SHEKAR, K. J. H. VERWEIJ, M. C. KELLER, S. MACGREGOR, . . . N. G. MARTIN, 2008 Genetic factors predisposing to homosexuality may increase mating success in heterosexuals. Evolution and Human Behavior 29: 424-433.







[Matthew C Keller's Home Page] [Biosketch] [Vita] [Publications] [Keller Lab] [Program Code] [R] [Courses] [Links]