The Date of Interbreeding between Neandertals and Modern Humans

Neandertal DNA is present in the genomes of people alive today. That much has been established. However, for years, scientists struggled to prove that this DNA arrived through actual interbreeding rather than from a shared ancestor who lived in Africa long before any migration occurred. The evidence looked the same either way—more Neandertal-like genetic variants are found in non-Africans than in Africans, with two completely different explanations that could account for it. Sankararaman and colleagues discovered a method to differentiate between these scenarios. Their approach was not based on mutations but on how chromosomes break apart. The core issue is this: when researchers compared the draft Neandertal genome to those of present-day humans, non-Africans consistently showed more genetic variants in common with Neandertals than sub-Saharan Africans did. Two hypotheses fit that observation equally well. The first is recent interbreeding—modern humans leaving Africa encountered Neandertals in Europe or western Asia, had children with them, and passed Neandertal DNA into the ancestors of everyone outside Africa today. The second is ancient population structure—the ancestral population that gave rise to both Neandertals and modern humans was already divided deep in Africa, and the branch that eventually expanded outside Africa happened to be genetically closer to Neandertals from the start. No interbreeding would be needed. Both scenarios result in the same excess of shared variants, but they do not produce the same timing patterns. This is where linkage disequilibrium comes into play. Linkage disequilibrium, or LD, refers to the tendency for nearby genetic variants to be inherited together on the same chromosome. When two populations intermingle, DNA from one enters the other in long, unbroken segments. With each successive generation, recombination splits those segments into shorter pieces. The older the mixing event, the shorter the surviving fragments and the weaker the correlation between distant variants. This fragmentation serves as a clock. Sankararaman and colleagues utilized it by defining a distance-dependent statistic, D of x, that measures the average correlation between pairs of Neandertal-associated variants as a function of the genetic distance between them. Under a gene-flow model, that correlation diminishes exponentially with distance—the rate of decay encodes the time since the gene flow occurred. The approach taken to identify relevant data was critical. To enhance the admixture signal and reduce background noise, the team focused their analysis on single nucleotide polymorphisms where the Neandertal genome carries the derived allele—the newer, non-ancestral version—and where that derived allele is rare in the modern population, appearing at less than ten percent frequency. This filtering step significantly diminishes contamination from linkage disequilibrium that would exist even without any interbreeding, as confirmed by their coalescent simulations across various demographic models. They then fitted an exponential curve to the decay of D of x for genetic distances ranging from 0.02 to 1 centimorgan, using increments of 0.001 centimorgans. One significant technical issue arose: errors in genetic maps—these are the charts that indicate how often recombination occurs at each position in the genome—could distort the date estimate. A map that consistently overestimates recombination rates would make the linkage disequilibrium decay appear faster, suggesting a more recent admixture event than actually took place. The team addressed this by introducing a map-precision parameter. They estimated this parameter by comparing the deCODE genetic map against seven hundred twenty-eight crossover events from a Hutterite pedigree study and corrected it using Gibbs sampling within a Bayesian framework. This correction carried map uncertainty through to the final date estimate. When applied to the 1000 Genomes Pilot 1 European data, which included sixty European Americans, the analysis produced an exponential decay parameter of approximately one thousand one hundred seventy-nine to one thousand two hundred thirty-three using the deCODE map. After correcting for map uncertainty and converting generations to years with a baseline of twenty-five to thirty-three years per generation, this translates to a most likely window of forty-seven thousand to sixty-five thousand years ago. The conservative range, accounting for both maps used in the study, is thirty-seven thousand to eighty-six thousand years ago. Simulations demonstrated that the estimator is never more than fifteen percent off from the true value across various demographic models. In scenarios without gene flow, including ancient subdivision alignments with real data, the inferred dates always traced back at least five thousand generations—much older than what was observed. The empirical results are not an artifact of the methodology. Two alternative explanations still needed to be directly addressed. The first is natural selection. If selection has actively removed or favored certain variants near genes, it could generate linkage disequilibrium that mimics the admixture signal, making the inferred date appear more recent than it truly is. To investigate this, the team categorized single nucleotide polymorphisms by distance to the nearest exon, dividing the data into five groups. The decay parameter across all five groups ranged from one thousand one hundred forty-five to one thousand three hundred one, which is completely consistent with the unstratified estimate of one thousand two hundred one. Therefore, natural selection near genes is not responsible for the observed result. The second alternative is ancient African structure. This hypothesis is more conceptually complex as it isn't merely a statistical artifact—it provides a genuine alternative history of human origins. Sankararaman and colleagues argue it leads to a distinct prediction: under ancient structure, the last common ancestry between the relevant populations should date to at least two hundred thirty thousand years ago, based on the Neandertal fossil record. In contrast, under recent interbreeding, the last gene exchange should occur within roughly one hundred thousand years of today. The linkage disequilibrium decay signature observed—exponential at the measured scale—aligns with what admixture linkage disequilibrium produces, not with background linkage disequilibrium from ancient shared ancestry and genetic drift. A date of forty-seven thousand to sixty-five thousand years ago contradicts a purely ancient-structure explanation. So what does this window actually indicate? It aligns with a particular chapter of human prehistory. Modern humans appear in the Middle East before one hundred thousand years ago at archaeological sites like Skhul and Qafzeh. Neandertals are recorded expanding into the same region around seventy thousand years ago, including at Tabun Cave. Modern humans reappear there around fifty thousand years ago, this time equipped with Upper Paleolithic technologies—the new toolkit associated with behavioral modernity. The forty-seven thousand to sixty-five thousand year window places the genetic contact squarely within that overlap period. The authors emphasize what the data can and cannot clarify. The date is clearest under a single-episode admixture model. If gene flow occurred in multiple pulses or as a prolonged, low-level trickle, the estimate becomes an average of those events and an upper limit on the most recent one. The geographic location of the contact also remains uncertain—the genetic data alone cannot pinpoint the exact location of the mixing, only provide an approximate time frame. The paper also mentions that the source population need not have been Neandertals themselves; it could have been a population more closely related to Neandertals than any group alive today. What the data concretely support is this: the linkage disequilibrium pattern in living Europeans is inconsistent with Neandertal DNA arriving through ancient African substructure. It is consistent with a real biological event—people meeting, reproducing, and passing on genes that occurred roughly between forty-seven thousand and sixty-five thousand years ago. The Neandertal-derived DNA that non-Africans carry today is not an abstraction or a statistical residue. It represents the genomic trace of specific encounters during a specific period that has left a mark still detectable in the chromosomes of anyone outside Africa walking around today. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.