Pritchard Lab Research

Genetic variation within species is the ultimate driver of evolution and phenotypic variation. Most of our work uses statistical and computational methods to study aspects of genetic variation in human genetics and in evolutionary biology.

We often work on problems where we need to tackle new kinds of genomic data or new types of questions. Thus, a central part of our work involves developing appropriate statistical and computational approaches that can yield new insights into modern genome-scale data sets. Some of our main research interests are described below, with representative references.

Current active areas in the lab include:

  • Genetic architecture of complex traits, polygenic prediction, portability, and the omnigenic model.
  • Gene regulation and variation: cis and trans eQTLs, gene regulatory networks, inference and models.
  • Population genetics of complex traits, adaptation, and human history: theory, inference, data analysis.

  • Beyond these areas we have broad interests in human genetics and evolution, and are happy to collaborate on a wide range of topics.

    [Most recent papers from PubMed]
    [Google Scholar summary]

    Genetic architecture of complex traits

    One area of longterm interest in our lab is to understand the molecular processes that map genotypes to phenotypes. In particular, we are interested in understanding why so much of the genome affects any given complex trait. How should we understand recent observations from GWAS that typical complex traits are affected by 104 to 105 variants spread across the genome, likely mediated through a considerable fraction of all genes expressed in trait-relevant tissues?

    In a pair of papers in 2017 and 2019, we proposed a model in which any given trait is affected directly by a modest number of genes (perhaps hundreds), that we term "core genes". But since the core genes are plugged into gene regulatory networks, essentially any expressed gene can potentially exert nonzero trans effects on core gene regulation. (In practice, not all genes have regulatory variants, and the distribution of effect sizes is presumably centered around zero.)

    One prediction of the model is that anywhere from 70% to nearly 100% of the heritability for a complex trait (depending on the covariance of core genes) is due to weak trans effects mediated via peripheral genes (Liu et al, 2019).

    An Expanded View of Complex Traits: From Polygenic to Omnigenic. Boyle et al. 2017 Cell 169:1177-1186. [PDF]

    Trans Effects on Gene Expression Can Drive Omnigenic Inheritance. Liu et al. 2019 Cell 177(40):1022-1034 [PDF]

    In follow-up work we are exploring the architecture of a variety of model traits that can illustrate general principles of trait architecture. The first of these papers focuses on studying the nature of core genes, and how much these contribute to heritability. Meanwhile, important work from other groups has shown that in at least some cases polygenic risk is mediated via core genes: [link], [link].

    GWAS of three molecular traits highlights core genes and pathways alongside a highly polygenic background. Sinnott-Armstrong et al. 2021 eLife. 10:e58615 [PDF]

    And an important new focus relates to experimental measurement, inference, and theoretical modeling of gene regulatory networks, motivated by the prediction that most trait heritability flows through trans effects. Stay tuned in this space.

    Genetic variation in gene expression

    Genetic variants that impact gene regulation play central roles in the genetics of complex traits and adaptation. Yet we still have incomplete understanding of exactly how the genome sequence encodes regulatory information, or how that information is "read" by the cellular machinery in any given cell type or context. The challenge of distinguishing functional regulatory variants from the many millions of nonfunctional variants is a fundamental hurdle to understanding complex traits and evolution.

    Starting in 2008 our lab initiated a primary focus to understand the primary mechanisms by which variation impacts expression, and to predict which variants have regulatory activity. Working with Yoav Gilad's lab, our main approach was to perform QTL mapping of a large number of functional genomics phenotypes in the same panel of cell lines with known genomes (75 Yoruba lymphoblastoid cell lines).

    Our work over a ten year period helped to pioneer QTL mapping for RNA-seq and functional assays such as DNase, histone marks, and ribosome profiling. At the time, this was by far the most comprehensive data set of QTL mapping for many assays performed in the same panel of cell lines. Through our work we showed in 2012 that most cis-eQTLs are due to SNP effects on chromatin, and that these can be detected as DNase QTLs (ATAC-QTLs provide similar signal). We showed that few gene-level QTLs affect processes after transcription initiation, though small numbers affect processes such as mRNA decay and protein levels. Meanwhile, in a 2016 paper we showed that mRNA splicing QTLs are also more widespread than had been appreciated, and that these are also a significant contributor to complex traits.

    More recently, our main focus has turned to understanding trans-regulation through gene regulatory networks. This is motivated in part by the inference that most heritability for complex traits is mediated via trans-regulation. At present, methods for studying trans-regulation lag far behind the methods for studying and interpreting cis-regulation; however this is an area where modern CRISPR-based perturbations are starting to make key contributions.

    Systematic discovery and perturbation of regulatory genes in human T cells reveals the architecture of immune networks. Freimer et al 2021. bioRxiv [PDF]

    Landscape of stimulation-responsive chromatin across diverse human immune cells. Calderon et al 2019. Nature Genetics 51(10):1494-1505 [PDF]

    Annotation-free quantification of RNA splicing using LeafCutter. Li et al 2018. Nature Genetics 50(1):151-158. [PDF]

    RNA splicing is a primary link between genetic variation and disease. Li et al 2016. Science. 352:600-4. [PDF]

    WASP: allele-specific software for robust molecular quantitative trait locus discovery. van de Geijn et al 2015. Nature Methods. 12:1061-3. [PDF]

    Impact of regulatory variation from RNA to protein. Battle et al 2015. Science 347:664-7. [PDF]

    Identification of Genetic Variants That Affect Histone Modifications in Human Cells. McVicker et al 2013. Science 342:747-9. [PDF]

    Primate Transcript and Protein Expression Levels Evolve under Compensatory Selection Pressures. Khan et al 2013. Science 342:1100-4. [PDF]

    DNaseI sensitivity QTLs are a major determinant of human expression variation. Degner et al 2012. Nature 482:390-4. [PDF]

    Dissecting the regulatory architecture of gene expression QTLs. Gaffney et al 2012. Genome Biology 13(1):R7. [PDF]

    Understanding mechanisms underlying human gene expression variation with RNA sequencing. Pickrell et al 2010. Nature 464:768-72. [PDF]

    Inference of population structure and history

    A long-term interest for us has been in the development of methods for interpreting population structure and population history from genetic data. The history of a species is recorded in the patterns of genetic variation within and between individuals and populations.

    One class of methods that we have developed makes use of data from SNPs or other markers to study population structure and ancestry of a species. The Pritchard-Stephens-Donnelly algorithm was implemented in the program Structure algorithm. Structure views a sample of individuals as (potentially) representing a mixture from different genetic populations. It uses the marker data to infer both the overall genetic structure and the ancestry of individuals.

    This model has been very widely used in a wide variety of contexts, and the original paper has been highly cited [historical perspective by John Novembre 2016]. The Structure ancestry model underlies the ancestry reports used by 23andMe and The software is also widely used for applications in conservation biology and ecology. Closely related models - developed a couple of years later by David Blei and colleagues - have been very influential in the topic modeling literature.

    The image below is the first Structure ancestry plot, made by Noah Rosenberg and Jonathan in 2002. This shows clustering results for 1056 humans from 52 globally distributed populations.

    In other early work, we built on key papers from Simon Tavare and Gunther Weiss to develop the first application of Approximate Bayesian Computation, in this case to estimate human demography from Y chromosome data [PDF].

    With Joe Pickrell, we developed the TreeMix algorithm for inferring the relationships among modern populations, while allowing for pulses of gene flow between different clades in the tree (as illustrated at right).

    Our current work in this area has moved into using ancient DNA approaches to study changes in population ancestry over time, as well as adaptive shifts. In collaboration with Ron Pinhasi (Vienna) and Alfredo Coppa (Rome) we are conducting studies of ancient DNA in Rome and other regions of the Mediterranean. We are also interested in methods development for studying population genetics in temporal and spatial data.

    Ancient Rome: A genetic crossroads of Europe and the Mediterranean. Antonio et al 2019. Science. 366:708-714. [PDF] [Short Video] [Science news]

    fastSTRUCTURE: variational inference of population structure in large SNP data sets. Raj et al 2014. Genetics 197:573-89. [PDF]

    Inference of population splits and mixtures from genome-wide allele frequency data. Pickrell and Pritchard 2012. PLoS Genetics 8:e1002967 [PDF] [Software]

    Sequencing and Analysis of Neanderthal Genomic DNA. Noonan et al 2006. Science 314:1113-1118. [PDF]

    The genetic structure of human populations. Rosenberg et al 2002. Science 298: 2381-2385. [PDF]

    Inference of population structure using multilocus genotype data. Pritchard et al 2000. Genetics 155: 945-959. [PDF], [Software]

    Population growth of human Y chromosomes: a study of Y chromosome microsatellites. Pritchard et al 1999. Mol. Biol. Evol., 16:1791-1798. [PDF],

    Natural selection in human populations

    Another major area of interest is in understanding natural selection in human populations. We are interested in understanding both the typical modes by which selection acts, as well as the key genes and phenotypes that have been targets of adaptation in different human populations.

    In our early work on this problem our goal was to identify the strongest signals of selective sweeps in the genome (Voight et al 2006).

    However in 2009, with more extensive data, we argued that in fact there appear to be fewer signals of strong classical hard sweeps in recent human evolution than we had believed earlier (see also related work by our colleague Molly Przeworski).

    Instead we have proposed that most adaptation likely occurs through a process of "polygenic adapation" in which small allele frequencies at large numbers of quantitative trait loci allow very rapid phenotypic adaptation but are difficult to detect by standard tests. Consequently we are very interested in approaches for studying polygenic adaptation from modern and ancient DNA data. One example of this work was presented in Yair Field's 2016 paper using our SDS statistic to infer recent changes in allele frequencies. However, this work remains challenging due to concerns about subtle population structure confounding in GWAS.

    In other ongoing work we are also studying models of stabilizing selection in complex trait, as well as balancing selection in MHC.

    Reduced signal for polygenic adaptation of height in UK Biobank. Berg et al 2019. Elife. 8:e39725 [PDF]

    Evidence of weak selective constraint on human gene expression. Glassberg et al 2019. Genetics. 211(2):757-772 [PDF]

    Detection of human adaptation during the past 2000 years. Field et al 2016. Science 354:760-764. [PDF]

    The deleterious mutation load is insensitive to recent population history. Simons et al 2014. Nature Genetics 46:220-4. [PDF]

    The genetics of human adaptation: hard sweeps, soft sweeps, and polygenic adaptation. Pritchard et al 2010 Current Biology. 20:R208-15. [PDF]

    How we are evolving. Pritchard 2010 Scientific American. 301(10):41-47. [link]

    The role of geography in human adaptation. Coop et al 2009 PLoS Genetics 5:e1000500. [PDF]

    A Map of Recent Positive Selection in the Human Genome. Voight, et al 2006. PLoS Biol 4(3): e72 [PDF] [2013 Perspective on this work]

    Recombination, LD, deletions, association mapping...and other topics

    Beyond the areas listed above, we have broad interests in problems where our computational toolbox can provide biological insights; a handful of examples are given below.

    One area of interest has been in understanding linkage disequilibrium and recombination (much of this in collaboration with Molly Przeworski and Graham Coop), including providing evidence for variation in hotspot usage across individuals (now known to be due to variation at PRDM9).

    When Don Conrad was in the lab he provided one of the early genome-wide surveys of deletion polymormisms, when it was first becoming clear that copy number variation is an important aspect of genome variation (see figure at right).

    We also helped to introduce the idea of using genotype data to detect and controlling for the confounding effects of population structure in association mapping. More broadly we have been interested in population genetic models of complex traits, including work on the role of rare variants in disease.

    In more recent work, Xun Lan revisited old models concerning the evolution of gene duplications. In contrast to one prevailing model, he concluded that subfunctionalization is likely not the main process enabling the evolutionary survival of young duplicates. Instead he proposed a mechanism of rapid downregulation followed by dosage sharing (Lan and Pritchard 2016). In other work, Kelley Harris initiated a wonderful line of research studying evolution of the mutational spectrum in humans, which she has since taken on to her own lab at U Washington.

    A wild-derived antimutator drives germline mutation spectrum differences in a genetically diverse murine family. Sasani et al 2021. bioRxiv 2021.03.12.435196 [PDF]

    Rapid evolution of the human mutation spectrum. Harris and Pritchard 2017. Elife. 6: e24284 [PDF]

    Coregulation of tandem duplicate genes slows evolution of subfunctionalization in mammals. Lan and Pritchard 2016 Science. 352:1009-13. [PDF]

    Genetic variation in MHC proteins is associated with T cell receptor expression biases. Sharon et al 2016. Nature Genetics. 48:995-1002 [PDF]

    High-Resolution Mapping of Crossovers Reveals Extensive Variation in Fine-Scale Recombination Patterns Among Humans. Coop et al 2008. Science 319: 1395-1398. [PDF]

    A high-resolution survey of deletion polymorphism in the human genome. Conrad et al 2006. Nature Genetics 38:75-81. [PDF]

    Clonal origin and evolution of a transmissible cancer. Murgia, et al 2006. Cell 126:477-87. [PDF]

    Linkage disequilibrium in humans: models and data. Pritchard and Przeworski 2001. Am. J. Hum. Genet. 69:1-14 [PDF]

    Use of unlinked genetic markers to detect population stratification in association studies. Pritchard and Rosenberg 1999. Am. J. of Hum. Gen. 65: 220-228. [PDF]

    Research Funding.

    Our work has been generously supported by the National Institutes of Health, the Howard Hughes Medical Institute, the Packard Foundation, the Sloan Foundation, and Burroughs Wellcome Fund.