Human ancestors interbred with two distinct populations of distant relatives

By analyzing the distribution of derived alleles, this study provides evidence that human ancestors interbred with two distinct superarchaic populations: an earlier-diverging group that contributed to Neanderthals and Denisovans, and a later-diverging group that contributed to early modern humans in Africa.

The Gist

Imagine human history not as a single, straight line of evolution, but as a massive, tangled family reunion where distant relatives kept showing up at different parties and swapping recipes.

For a long time, scientists thought we had the guest list mostly figured out: Modern humans, Neanderthals, and Denisovans. We knew these groups had met and had children together (interbred). But there was a mystery: DNA evidence suggested that all of these groups had also mixed with an even older, stranger group of ancestors that we call "Superarchaics."

The big question was: Was there just one "Superarchaic" family that visited everyone, or were there two different families?

This new paper says: It was two different families.

Here is the story broken down with some simple analogies:

The Two "Ghost" Families

Imagine the human family tree as a big river that splits into different streams.

1. The "Deep River" Family (Population S): A long time ago (over a million years), one group of humans split off and went their own way. They became very isolated. Later, they met up with the ancestors of Neanderthals and Denisovans (who were living in Eurasia) and mixed with them. Think of this as a mysterious great-uncle who showed up at the Eurasian family reunion and left some of his DNA behind.
2. The "Shallow River" Family (Population Z): At the same time, another group of humans was evolving in Africa. They also split off from the main human line, but they didn't split off as long ago as the first group. They stayed in Africa. Later, the ancestors of all modern humans (who were also in Africa) met this second group and mixed with them. This is like a different, slightly less ancient great-aunt who showed up at the African family reunion.

The Big Discovery: Before this paper, scientists thought it might be the same great-uncle visiting both the Eurasian and African parties. This paper proves that it was two different people. The "Eurasian Superarchaic" and the "African Superarchaic" were distinct groups that had been separated for a very long time.

The Detective Work: How Did They Know?

The authors didn't have time machines. Instead, they used a digital magnifying glass called Legofit.

Think of DNA as a giant puzzle made of millions of tiny pieces (letters). When two groups mix, they leave specific patterns in the puzzle.

• If Group A mixes with Group B, you see a specific pattern of letters.
• If Group A mixes with Group C, you see a different pattern.

The researchers looked at the DNA of:

• Modern people from West Africa and Europe.
• Ancient Neanderthals and Denisovans.

They ran thousands of computer simulations, testing different family trees. They asked: "Does the data look better if we assume one Superarchaic family visited everyone, or if we assume two different families visited different groups?"

The Result: The computer said, "The two-family story fits the puzzle pieces perfectly. The one-family story leaves gaps."

Why Does This Matter?

This changes how we picture the ancient world:

• Africa was a Melting Pot: It wasn't just one group of "Modern Humans" waiting to leave Africa. It was a complex mix of different ancient groups that had been separated for hundreds of thousands of years, finally coming together.
• The "Missing" Ancestors: We now know that modern humans carry DNA from a "Superarchaic" group that lived in Africa and is completely different from the Neanderthals and Denisovans we usually hear about.
• Solving a Math Problem: Previous studies had a weird glitch. When they tried to calculate when Europeans and Africans split apart, the math gave them a silly answer (like "they split only 5,000 years ago," which we know is wrong because archaeology says otherwise). By adding this "Second Superarchaic" into the model, the math suddenly worked perfectly, and the dates made sense again.

The Takeaway

Human history is messier and more fascinating than a simple tree. It's more like a braid.

We used to think we were a single thread that occasionally tangled with Neanderthals. Now we know that our thread was actually woven from three different strands:

1. The main modern human line.
2. A very ancient "Superarchaic" line that mixed with Neanderthals/Denisovans.
3. A different ancient "Superarchaic" line that mixed with our African ancestors.

So, when you look in the mirror, you aren't just looking at a modern human. You are looking at a unique blend of moderns, Neanderthals, and two different kinds of ancient "ghosts" from the deep past.

Technical Summary

1. Problem Statement

Recent ancient DNA studies have established that modern humans interbred with Neanderthals and Denisovans, and that these archaic groups themselves interbred with a "superarchaic" population that diverged early in the Pleistocene. Additionally, evidence suggests that early modern human ancestors in Africa also received gene flow from a superarchaic source.

However, a critical ambiguity remained: Did the superarchaic gene flow into Eurasian archaics (Neanderthals/Denisovans) and the superarchaic gene flow into African modern humans originate from the same population, or were there two distinct superarchaic populations? Previous models often assumed a single superarchaic source, but this assumption led to statistical inconsistencies and biologically implausible parameter estimates (specifically regarding the separation time of modern African and European populations).

2. Methodology

The authors employed a model-based approach using nucleotide site patterns and the Legofit software to evaluate competing hypotheses of population history.

• Data Source: The analysis utilized 12 high-coverage published genomes grouped into five subsets:
  • X: 3 West African (Yoruba) modern humans.
  • Y: 5 European (French/English) modern humans.
  • R: 2 recent Neanderthals (Vindija, Chagyrskaya).
  • A: 1 older Neanderthal (Altai).
  • D: 1 Denisovan.
• Statistical Framework:
  • Site Patterns: The method analyzes the frequency of specific allele configurations (site patterns) across the four sampled populations (X, Y, N, D). For example, a mutation on a specific branch of the genealogy creates a pattern where only one population carries the derived allele.
  • Model Comparison: The authors compared complex demographic models involving multiple episodes of gene flow (admixture), denoted by Greek letters (e.g., $\alpha, \beta, \gamma, \delta, \zeta$).
  • Model Selection: They used the Bootstrap Estimate of Predictive Error (bepe) to assess model fit. To determine the robustness of the preferred model, they applied Bootstrap Model Averaging (booma), which calculates the fraction of bootstrap replicates in which a specific model yields the lowest bepe.
  • Parameter Estimation: Legofit minimizes the Kullback-Leibler (KL) divergence between observed and expected site-pattern frequencies using a differential evolution algorithm. The analysis involved a multi-stage optimization process (stochastic to deterministic) and Principal Component Analysis (PCA) to reduce dimensionality.
  • Validation: The authors conducted simulations using msprime under the preferred model to assess bias and confidence intervals, particularly for population size and admixture parameters.

3. Key Contributions

• Identification of Two Distinct Superarchaics: The study provides strong statistical evidence that there were at least two distinct superarchaic populations involved in hominin history, rather than one.
  • Population S: A deeply divergent superarchaic that contributed ancestry to the ancestors of Neanderthals and Denisovans (and later to Denisovans specifically).
  • Population Z: A second, distinct superarchaic population that diverged more recently than S but still prior to the split of modern humans from archaics. This population contributed ancestry specifically to the ancestors of modern humans in Africa.
• Resolution of Model Misspecification: The paper explains why previous models (assuming a single superarchaic) produced absurd estimates for the separation time ($T_{XY}$) between West Africans and Europeans (suggesting a split of only a few thousand years). The new model corrects this bias by accounting for the unique shared derived alleles between modern Africans and Eurasians caused by the distinct superarchaic admixture.

4. Key Results

• Model Selection: The preferred model is $\alpha\beta\gamma\delta\zeta$, which received 98% of the bootstrap model average weight.
  • This model includes gene flow from Population S to Neanderthal-Denisovan ancestors ($\delta$) and a separate gene flow event ($\zeta$) from Population Z into the ancestors of modern humans ($XY$).
  • The alternative model ($\alpha\beta\gamma\delta\epsilon$), which assumes the same superarchaic contributed to both archaics and moderns, received 0% weight.
• Parameter Estimates:
  • Divergence Time ($T_{XY}$): Under the new model, the separation time between West Africans and Europeans is estimated with an upper confidence interval of ~44 kya, aligning with archaeological consensus (previously, the single-superarchaic model forced this to implausibly low values).
  • Admixture Fraction ($m_\zeta$): The study estimates that ~19.6% (95% CI: 12.4–26.4%) of the ancestry of early modern humans derives from Population Z.
  • Population Z Divergence: Population Z is estimated to have diverged from the modern-archaic trunk approximately 1.3 million years ago (assuming 29-year generations).
• Bias Analysis: Simulations revealed that while admixture and time parameters are largely unbiased, population size estimates for large populations (like superarchaics) tend to be upwardly biased, while smaller populations are downwardly biased. However, the true values generally fall within the estimated confidence intervals.

5. Significance

• Revising Pleistocene Structure: The findings reveal previously unrecognized structure among Pleistocene hominins. It suggests that the "superarchaic" label encompasses at least two distinct lineages with different divergence times and interaction histories.
• African Evolutionary History: The evidence supports the hypothesis that Population Z was likely an African population that remained largely isolated from the ancestors of modern humans for roughly a million years before interbreeding. This challenges the view of a panmictic African population during the Middle Pleistocene.
• Methodological Robustness: The study demonstrates the power of using site-pattern frequencies and rigorous model selection (booma) to resolve complex demographic histories that are difficult to infer from simple phylogenetic trees.
• Synthesis of Conflicting Data: The results reconcile previous findings regarding the "deep" haplotype at the MUC7 locus in Africans and other studies suggesting deep ancestry in modern humans, confirming that modern humans are a mixture of multiple deep lineages.

In conclusion, Rogers et al. demonstrate that the genetic history of humans is more complex than a simple tree with a single "superarchaic" branch; instead, it involves at least two distinct, deeply divergent superarchaic populations that contributed differentially to the gene pools of Eurasian archaics and African modern humans.