The Genealogical Adam and Eve

by S. Joshua Swamidass

  Olson teamed up with Chang to study genealogical ancestry further. With a graduate student, they performed large-scale simulations of human history that included geographical barriers. Within geographic areas, populations were further divided into smaller tribes. Known estimates of population size in the past were used to fix population sizes in the simulation. Access to different geographic areas opened up in accordance with known history. Individual people were modeled in these simulations, billions of them, and the software they wrote kept track of their parents, children, spouse, and birth and death dates.

  They stacked the simulation against recent common ancestry (fig. 4.1). For example, migration and interbreeding rates were very low. Most people in the simulation never left the community in which they were born. Those that did change communities only did so once in their lifetime. Couples stayed together for life, totally monogamous. Trade routes, like the Silk Road, were not included in the simulation, even though they would have pushed the estimates to more recent times. Large population migrations, military movements, and empires were not included in the simulation either, even though they also would have pushed the common ancestor more recently. Chang’s original work assumed that there were no barriers to mixing. This simulation took the opposite approach, modeling much higher barriers than we observed.

  Figure 4.1. Simulating universal ancestry on a world map. Universal common ancestry has been studied both analyticallya and with simulations.b A 2004 study in Nature simulated world history on the world map shown here. Each square is a country with many “towns.” Small amounts of migration were enough for universal ancestors to arise in about three thousand years. The arrows show some of the migration routes used in the simulation, labeled with the year they are opened up. The insets note four reasons that estimates from this simulation might be larger than a more accurate simulation, not smaller.

  aJ. Kelleher and others, “Spread of Pedigree Versus Genetic Ancestry in Spatially Distributed Populations,” Theoretical Population Biology 108 (2016): 1-12, https://doi.org/10.1016/j.tpb.2015.10.008; Joseph Lachance, “Inbreeding, Pedigree Size, and the Most Recent Common Ancestor of Humanity,” Journal of Theoretical Biology 261 (2009): 238-47, https://doi.org/10.1016/j.jtbi.2009.08.006; Chang, “Recent Common Ancestors.”

  b Douglas L. T. Rohde, Steve Olson, and Joseph T. Chang, “Modelling the Recent Common Ancestry of All Living Humans,” Nature 431 (2004): 562-66, https://doi.org/10.1038/nature02842; Douglas L. T. Rohde, “On the Common Ancestors of All Living Humans,” unpublished paper, November 11, 2003, https://tedlab.mit.edu/~dr/Papers/Rohde-MRCA-two.pdf.

  What were the results? In these simulations, common ancestors arose very recently, matching Olson’s guess of just two to three thousand years ago. In 2004, the results of these simulations were published in the journal Nature. These results are surprising. The journal editors invited comment from another scientist,9 who agreed with its findings, and eagerly looked to ancient genomes to further improve our knowledge of universal ancestry. One way we assess the strength of a finding, also, is by observing how it fares over time in the scientific community. Increasing our confidence, the surprising results of this 2004 Nature study stood uncontested in the literature for fifteen years. Several follow-up studies, moreover, increased our understanding of how different factors influence universal genealogical ancestry.

  As a preview of what is to come in the following chapters, my contribution to this work begins in 2017,10 when I first presented the genealogical hypothesis. Then, in a peer-reviewed article in 2018, I worked out more of the mathematical details, extrapolating the findings to the right time period in the past. I also examined the current state of evidence for and against isolation of populations, and what we might expect in the future. Perhaps most surprisingly, I found that there is no evidence for or against the de novo creation of Adam and Eve within a larger population.11

  Before explaining this, however, a review of what was already known is in order. Universal genealogical ancestors are among the most nonintuitive facts of human origins. In this chapter, I review four surprising findings. As we will see, (1) universal ancestors arise recently in our past; (2) this finding is robust, not tightly dependent on any strong assumptions; (3) universal ancestors are common, arising everywhere; but (4) most of them are hidden, leaving no evidence of their existence in our genomes. Though these are surprising facts, everything in this chapter is well established and agreed on among population geneticists.12

  SURPRISINGLY RECENT

  The first surprise is that universal genealogical ancestry is very recent. We share genetic common ancestors in the distant past, over one hundred thousand years ago. However, the most recent universal genealogical ancestor (MRUGA) of all living humans could have been situated as recently as three thousand years ago.13 The math from Chang’s 1999 study illuminates the difference. In a random mating model, universal genetic ancestors (like Y-adam and Mt-eve) appear in N generations, where N is proportional to the population size. But MRUGA comes in merely log2N generations.14 For a population of one million, universal ancestors arise in about twenty generations. For population one thousand times larger, one billion, universal ancestors arise in about thirty generations. Going back just a little bit further, we reach the identical ancestor point (IAP), where everyone at that point and more ancient is either ancestor of everyone alive today or ancestor of no one.15 We hit the IAP in just about 1.77 log2N generations, a little less than twice as long. For a population of one million, for example, the IAP would be at about thirty-five generations. Using a generation time of twenty-five years, that is less than 900 years ago.

  We can build intuition about this by comparing the total population on Earth with the number of ancestors we expect using a naive calculation. Going back each generation, we have two parents, then four grandparents, then eight great-grandparents; the number of ancestors appears to increase exponentially as we go back. The population size in past generations, however, either stays comparatively constant in much of paleo-history or decreases exponentially over the last ten thousand years.

  For example, there are about 160 generations between 10,000 and 5,000 years ago. Naively (and falsely) assuming there is no overlap in our family trees, we can compute the number of ancestors alive 10,000 years ago from the population at 5,000 years ago, 18 million people;16 we arrive at about 2.6 x 1055 ancestors. This is more ancestors than the number of stars in the visible universe. However, there were just 2 million people alive 10,000 years ago. How is this possible? The ratio between these numbers is 1.3 x 1049. This is how many times ancestors at 10,000 years are being double counted in this naive calculation, and it is an astronomical number of times.

  How do we reconcile this gap between the naive estimate and reality? Very quickly, all our genealogies begin to “collapse” by sharing more and more of the same ancestors.17 This is why the naive calculation double counts so much. It did not take into account that we are all related to one another, in multiple ways. It is this collapse of our family trees into the same large tree that gives rise to universal ancestors. As we go back in time, the first universal genealogical ancestor appears quickly. In a random mating model, this high degree of double counting gives rise to a MRUGA in about seven hundred years. In the 2004 Nature study, the MRUGA arises in just a few thousand years, and the IAP is just a little more ancient than this.

  SURPRISINGLY ROBUST

  The second surprise is that universal ancestry arises very reliably. Since the seminal study in 2004, scientists have continued to explore how genetic and genealogical ancestry interact. Understanding the difference between the two has clarified why these results should be trusted. Genetic ancestry transmits unreliably, but genealogical ancestry is consistent. This makes genealogical ancestry easier to predict. It arises robustly, with far less uncertainty than genetic ancestry.

  A visual picture might help. Genetic ancestry spreads feebly, like a drop of dye diffusing in cup of water. The DNA passed from an ancest
or dilutes as it slowly spreads from its starting point. In contrast, genealogical ancestry spreads like an explosion, in a chain reaction that does not dilute, but only grows. The number of people brought into a genealogical lineage grows at an increasing rate as time goes on. Certain factors can slow or hasten the rate of the explosion, but either way the explosion will still spread very fast. A set of simulations on a two-dimensional map demonstrates this difference. Genetic ancestry spreads in a dissipating wave that slows down with time and disappears, just like a drop of dye. In contrast, like an explosive chain-reaction, the wave of genealogical ancestry propagates across the map at a constant speed, taking over more territory each generation as it spreads.18 The number of genealogical descendants grow exponentially, picking up numbers more rapidly with time. This all takes place in a simulation that does not include migration between distant places. Long-range migration even more dramatically accelerates the spread of genealogical ancestry.

  This pattern holds up as more complexity is included in simulations. When migration is restricted to the idealized geography of a graph (a set of nodes connected together by edges), the time to universal ancestors is increased by a constant factor that only linearly depends on the size of graph.19 Moreover, time to universal ancestry does not depend on high migration rates between nodes in the graph; less than a single migrant per generation in the distant past robustly yields recent universal ancestors.20 Likewise, increasing inbreeding increases time to universal ancestors by a small, constant factor.21 Different factors can increase or decrease the time to universal ancestry, but not by much.

  How do these simplified models extrapolate to more realistic simulations of human history? As we already discussed, the 2004 Nature simulated the ancestry of present-day humans across the globe. This simulation took several factors into account. The simulation took place on a world map. Migration pathways opened in dates consistent with what is known of world history. In local regions, populations were divided into groups that interbred less frequently. Surprising even experts,22 these barriers only triple or quadruple time to universal ancestry from the seven hundred years computed in Chang’s first calculation. With low levels of migration, universal ancestors of all living humans still seem to arise in as few as three thousand years. In some simulations, they arise in as few as two thousand years.

  Figure 4.2. The contrast between genetic and genealogical ancestry. The top right panel shows the full genealogy of the population, with men (squares) and women (circles) connected to their children (lines). The top left and middle panels show two different genetic ancestries (black), converging to different ancestors, Y-adam (large square) and Mt-eve (large circle). The bottom panels show the lineages (black) of three different universal ancestors, shading in gray the generations of which they are universal ancestors. Every individual with a black border is a universal ancestor of everyone in the present.

  SURPRISINGLY COMMON

  The third surprise is that universal genealogical ancestors arise everywhere, stretching into the distant past. At the IAP, everyone across the globe who leaves ancestors eventually becomes a universal genealogical ancestor. Many individuals are each individually ancestors of “all the living.” All humans alive descend from each of these universal ancestors.

  Intuition can be built by considering a group of grandchildren that share the same grandfather. The grandfather is their common genealogical ancestor, but so also is every ancestor of the grandfather. Considering the distant ancestors shared by their parents, we find even more genealogical ancestors. Unlike genetic ancestors, genealogical ancestors are very numerous (fig. 4.2). They arise in a sudden cloud of individuals that quickly grows as we look back in time. All our different lineages quickly collapse into one family.

  Two of the objections to identifying Y-adam and Mt-eve with Adam and Eve do not apply to universal genealogical ancestors. First, it is not likely that Y-adam and Mt-eve were a paired couple. They are two unique individuals, most likely separated in time and space. In contrast, universal genealogical ancestors are so common that we expect couples. Consider a man who is monogamous and a universal genealogical ancestor. His partner would also be a universal genealogical ancestor. There is no reason to doubt that even the most recent universal genealogical ancestors included many paired couples.

  Second, Y-adam and Mt-eve most likely did not live in the Middle East, but probably lived in northern Africa. Universal genealogical ancestors, however, arise everywhere. They are not restricted to a single geographical area. Whatever we observe in the genetic data, universal genealogical ancestors are expected wherever we might think Adam and Eve would have lived.

  The ubiquity of universal genealogical ancestors raises a new objection. If universal ancestors are so common, they are not special; why, then, would we associate Adam and Eve with a universal ancestor? This misses the point of our inquiry. We are testing a hypothesis with evidence. Adam and Eve would not be special merely because they are universal ancestors, but for other reasons. In our hypothesis, Adam and Eve are not remarkable for unique biology or any scientific reasons. They are the same biological species of everyone outside the Garden. Abraham Lincoln, Albert Einstein, Rosalind Franklin, and Martin Luther King Jr. are all special, but not because of their genomes. In the same way, anything special about Adam and Eve would arise for reasons other than their genome.

  SURPRISINGLY HIDDEN

  The third surprise is that genealogical relationships are usually undetectable in genetic data. The same is true of universal genealogical ancestors. Detectable ancestors must (1) pass DNA to at least some descendants that live till today, and (2) this DNA be must identifiable as coming from them.

  Figure 4.3. Y-adam and Mt-eve only trace a subset of our genealogical ancestors. In this figure, the ancestry of a single individual (central circle) is displayed. Usually circles are women, but here it is a man. Each shell moving outward is a different generation. The first generation back has two ancestors, then four, eight, sixteen, and so on. At each generation, there is only one ancestor on the lineage leading to Y-adam and another one ancestor on the lineage leading to Mt-eve. The rest of the ancestors are invisible to Y-DNA and Mt-DNA.

  Genealogical ancestors in the distant past, however, are only rarely genetic ancestors; they usually leave their descendants no DNA.23 As one study explains, commonly universal ancestors are genetic ghosts who leave DNA to only some of their descendants, not all. Many of our ancestors are genetic super-ghosts “who are simultaneously (i) genealogical ancestors of each of the individuals at the present, and (ii) genetic ancestors to none of the individuals at the present.”24 Even when they do leave us DNA, we usually cannot identify this DNA with specific individuals in the past. Most of our genomes are identical to other humans, even though we inherit DNA from different sources.

  Figure 4.4. Genetic ghosts arise quickly in our past. In this figure, the ancestry of a single individual (central circle) is displayed. Each shell moving outward is a different generation. The first generation back has two ancestors, then four, eight, sixteen, and so on. Ten generations back there are 1,024 ancestors. In the top diagram (A), each ancestor is colored by the amount of DNA shared with the central individual. Diluting each generation back, the shade grows lighter each generation. Soon genetic ghosts arise, colored black to make them clearly visible. In the bottom diagram (B), the same data is displayed a different way. Here, the shading is scaled so that the highest genetic contribution of each generation is colored black, giving a clearer view of the variation within each generation. (C) Genetic ghosts arise very quickly. They are surprisingly common. Ten generations back (just 250 years ago), about two-thirds of the central individual’s ancestors are genetic ghosts.

  The DNA of individual ancestors is rapidly lost every generation (fig. 4.4). We can build our intuition about this by comparing the number of genetic and genealogical ancestors each generation, g. Just we already discussed, going back in time, the number of our ancestors grows exponentially:
two, four, eight, and so on, according to 2g. Our genomes are billions of bases long, but these bases are grouped into a much smaller number of pieces. We start out with 47 chromosomes (including mitochondrial DNA), each of which is a single piece of DNA. Going back each generation, a process called “recombination” shuffles chromosomes in a way that increases the number of “pieces” of our genome by about 60, so there are about 47 + 60g pieces of our genome as we move back in time. The ratio between the number of ancestors and the number of pieces of our genome each generation, then, is (47 + 60g)/2g. This ratio becomes very small very quickly, though it never quite reaches zero. When this ratio is small, it corresponds approximately to the fraction of our ancestors that pass DNA to us in generation g. At ten generations, just 250 years in the past, about two-thirds of our ancestors are genetic ghosts. At fifteen generations, about 98% of our ancestors are genetic ghosts, and only about 2% of ancestors leave us any DNA. At twenty generations, merely 500 years ago, about 99.9% of our ancestors are genetic ghosts (fig. 4.3 and 4.4), and only one out of a thousand ancestors leave us any DNA. These number are approximate. A better estimate takes into account interbreeding, and more details of genetic inheritance. This exercise, nonetheless, correctly builds our intuition, showing why most of our ancestors are genetic ghosts.

  This is a critically important point. Since most of our ancestors leave us no DNA at all, let alone identifiable DNA, genealogical relationships are “essentially unobservable” in genetic data past about fifteen generations.25 Recall, only a small amount of migration is required for universal genealogical ancestors to arise. The low level of ancient migration required for recent genealogical ancestry is undetectable in genetic data too.26 A single migrant per generation to an isolated population is enough to reliably give rise to recent genealogical ancestors. Even when migrants do leave DNA, it is not usually identifiable as a different population. The nearby migrants would be likely to have DNA very similar to the isolated population. This further reduces the likelihood that small amounts of migration would be detected.