GENOMICS North Indians and the other more ancestral

to modern South Indians. The ancestry of

New middle chapter in the most present-day Indians is probably com-

posed of these two populations along with a
handful of others (6–8). The ancestral South
story of human evolution Indian population formed as individuals
typical of the Indus Valley Civilization con-
tinued to mix with a population related to
Analyzing genomic data from ancient humans illuminates today’s Andaman Islanders; this genomic
South Asian ancestry mixing likely persisted until as late as 1700
to 400 BCE.
The ancestral North Indian population,
By Nathan K. Schaefer1,2 and Beth Shapiro1,2 side of, South Asia between 3300 and 2000 by contrast, contains ancestry from mi-
BCE and whose genetic signatures indicate grants thought to have moved south from

B
y comparing genetic information that they were migrants from the south. the Eurasian steppe—a flat, unforested
from extant humans worldwide, Each of these migrant genomes was a mix- grassland—but the source of this ancestry
researchers have painted a broad- ture of one population comprising mostly is controversial (6, 8). During the Bronze
strokes picture of human prehistory. early Iranian farmers and another resem- Age, a culture of steppe pastoralists called
However, these data reveal only the bling hunter-gatherers from the Andaman the Yamnaya spread ancestry and probably
oldest and most recent chapters of Islands south of India. These two groups technology and Indo-European languages

Downloaded from http://science.sciencemag.org/ on September 11, 2019

the story of human evolution. On page 999 mixed between 5400 and 3700 BCE, form- as far west as Spain and as far east as the
of this issue, Narasimhan et al. (1) capital- ing a genetic signature detectable in South Altai Mountains in Siberia (4, 5). Whereas
ize on recent advances in high-throughput Asian genomes today. Because these South some have suggested that the Yamnaya ar-
genome sequencing techniques and access Asian migrants were contemporaneous rived in North India around 3000 to 2500
to well-preserved ancient human remains with the Indus Valley Civilization, Narasim- BCE (8), others argue that steppe ancestry
to write a key middle chapter of this story. han et al. inferred that the migrants’ genetic in North India also contains a European
Over the past 100,000 years or so, hu- profile was probably typical of members of Neolithic component, which implies that
mans dispersed across Africa and then the Indus Valley Civilization. Thus, their it came from a later eastward expansion
throughout the rest of the world, conse- study provides evidence of the genomic ori- around 2300 to 1200 BCE (6). Narasimhan
quently adapting to a wide variety of lands, gins of the Indus Valley Civilization despite et al. date the arrival of steppe ancestry to
climates, and ecosystems (2). This basic nar- not including any ancient genomes from 1900 to 1500 BCE, which supports the lat-
rative illuminates questions—about local the Indus Valley Civilization itself. ter hypothesis. Thus, steppe ancestry and
adaptation, ecosystem disruption, linguis- These new data also suggest that, after the associated Indo-European languages
tic evolution, and human prehistory—that the fall of the Indus Valley Civilization, arrived later in India than in Europe and
genomic data from extant humans fails to several migrations into South Asia led to likely arrived by way of Europe.
answer. To address this deficiency, Narasim- the formation of two distinct populations The amount of genome-wide data avail-
han et al. built on previous large-scale stud- therein, one more ancestral to modern able from ancient humans has exploded in
ies of human migration history conducted
with ancient DNA obtained from human
remains across Eurasia (3–6) (see the fig- Ancient human nuclear genomes published since 2010
ure). The authors sequenced more than 500 Research articles returned in a PubMed search for “ancient genomes” were analyzed for newly published,
genomes of humans belonging to ancient genome-scale DNA sequencing data (single-nucleotide polymorphism or shotgun) from anatomically modern
cultures from archaeological sites across humans. Cited articles contributed the most ancient genomes in each year going back to 2014.
a large part of Asia. They then used an ar-
ray of allele frequency–based statistics and Years (thousands) covered by study
algorithms to model human populations 0 10 20 30
across time as mixtures of other, earlier
populations and investigated outstanding
1 26 genomes 2 101 genomes 3 44 genomes
GRAPHIC: V. ALTOUNIAN/SCIENCE FROM N. K. SCHAEFER AND B. SHAPIR0

questions about human dispersals in South

and Central Asia. Although this effort does 1000 North America Europe and Asia Southwest Asia
Raghavan et al. (9) Allentoft et al. (5) Lazaridis et al. (10)
Number of ancient genomes

not rewrite the history of humans in South

Asia, it does fill in several missing pieces. 4 130 genomes 5 400 genomes 6 523 genomes
Despite the archaeological and historical Europe Europe THIS STUDY

importance of the Copper Age Indus Val- Lipson et al. (11) Olalde et al. (12) Asia
Narasimhan et al. (1)
ley Civilization in the northwest regions of
South Asia, the genetic origins of the peo- 500
ple of this culture remain uncharacterized.
Narasimhan et al. identified 11 individuals
1 2 3 4 5 6
who lived in regions adjacent to, but out-

1
Howard Hughes Medical Institute, University of California, Santa
Cruz, Santa Cruz, CA 95064, USA. 2Department of Ecology and 0
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Evolutionary Biology, University of California, Santa Cruz, Santa (Jan.−June)
Cruz, CA 95064, USA. Email: [email protected] Year

SCIENCE sciencemag.org 6 SEP TEMBER 2019 • VOL 365 ISSUE 6457 981
Published by AAAS
INSIGHTS | P E R S P E C T I V E S

recent years, and the study by Narasimhan COMPUTER SCIENCE

et al. is the largest contribution to date, with
more than 500 ancient human genomes se-
quenced (see the figure). This scale of ge-
nomic data enabled the authors to compare
Machine learning transforms
genomes across a large number of locations
and time points and to home in on increas- how microstates are sampled
ingly specific questions that would have
been unanswerable even a few years ago. A deep neural network is trained to optimally explore
The human story revealed by Narasim- rugged potential energy landscapes in simulations
han et al. in South Asia is similar to those

By Mark E. Tuckerman ged “landscape” in this high-dimensional

space characterized by an exponentially

A
tomistic simulations of complex large number of low-energy regions or
molecular systems can provide key minima, all separated by ridges whose ener-
microscopic insights not easily ac- gies are typically several to many times kBT
cessible to experiments, such as fold- above these minima at room temperature.
ing of proteins (see the first figure), In a Boltzmann distribution, the minima
binding of small-molecule drug can- and ridges represent microstates of high

Downloaded from http://science.sciencemag.org/ on September 11, 2019

didates or peptide therapeutics to selected and low probability, respectively. It is gener-
targets, and formation of different poly- ally not possible to sample the distribution
morphs of molecular crystals, provided that directly, hence the challenge is to devise an
two major hurdles are overcome. First, an efficient algorithm to explore the U(x) land-
accurate description of the interatomic in- scape in order to locate these low-energy re-
teractions is needed, which is captured in a gions and identify the ridges that constitute

CREDITS: (PHOTO) ROBERTO MICHELI AND MASSIMO VIDALE/ISMEO - ITALIAN ARCHAEOLOGICAL MISSION IN PAKISTAN; (GRAPHIC) V. ALTOUNIAN/SCIENCE FROM M. E. TUCKERMAN
single potential energy function U(x), where the most likely transition pathways between
x denotes the full set of atomic coordinates. the minima, much as a hiker tries to identify
Second, given U(x), predicting thermody- the easiest mountain pass to cross from one
Narasimhan et al. studied the DNA from namic and other equilibrium properties of valley to another. Physically, minima largely
more than 100 ancient humans unearthed from the system or estimating kinetics requires determine thermodynamics, whereas kinet-
northernmost South Asia. sampling a statistically sufficient num- ics depend mostly on the ridges.
ber of realizations of x from the so-called Numerous approaches have been proposed
from elsewhere in Eurasia: Successive waves Boltzmann probability distribution P(x), to perform an efficient exploration of U(x).
of migration altered the genetic makeup of, which is proportional to exp[–U(x)/kBT], Some methods bias the search (2); others tar-
but did not completely replace, preexisting where T is the system temperature and kB get a few preselected functions of x, known
groups (4–6). Modern South Asians appear is Boltzmann’s constant. On page 1001 of as “reaction coordinates,” that are assumed
to be a mixture of Iranian-like hunter-gather- this issue, Noé et al. (1) introduce a machine to capture the most relevant features of the
ers, a population ancestral to the Andaman learning–based approach to address the lat- Boltzmann distribution (3–9). Such tech-
Islands, and Eurasian steppe herders who ter of these two challenges. niques are often quite sensitive to how these
first settled in Europe. Some South Asian The primary difficulty in sampling physi- biases or reaction coordinates are chosen, and
populations later received immigrants from cal realizations x of the system (also called poor choices can waste many hours of com-
other outside groups. As more genomes be- “microstates”) from the Boltzmann distribu- putation searching irrelevant regions of U(x).
come available from previously unexamined tion lies in the nature of U(x) itself. In large, The ruggedness of U(x) that makes sam-
historical cultures around the world, stories complex systems, x holds the positions of pling the Boltzmann distribution so chal-
like this one will fill other middle chapters in hundreds of thousands to millions of atoms. lenging originates in the way that physical
the book of human history. j Thus, U(x) should be viewed as a vast, rug- microstates (coordinates x) are represented.
REFERENCES AND NOTES
1. V. M. Narasimhan et al., Science 365, eaat7487 (2019).
2. M. Slatkin, F. Racimo, Proc. Natl. Acad. Sci. U.S.A. 113, Peptide folding as a search problem
6380 (2016). Hypothetical peptide configurations that might arise in energy or example training. Images adapted from (13).
3. I. Lazaridis et al., Nature 513, 409 (2014).
4. W. Haak et al., Nature 522, 207 (2015).
5. M. E. Allentoft et al., Nature 522, 167 (2015).
6. P. de Barros Damgaard et al., Science 360, eaar7711
(2018).
7. D. Reich et al., Nature 461, 489 (2009).
8. A. Basu et al., Proc. Natl. Acad. Sci. U.S.A. 113, 1594 (2016).
9. M. Raghavan et al., Science 345, 1255832 (2014).
10. I. Lazaridis et al., Nature 536, 419 (2016).
11. M. Lipson et al., Nature 551, 368 (2017).
12. I. Olalde et al., Nature 555, 190 (2018).

ACKNOWLEDGMENTS
We thank R. E. Green for helpful comments on and discussion
about this manuscript.
X Z
10.1126/science.aay3550 Fxz

982 6 SEP TEMBER 2019 • VOL 365 ISSUE 6457 sciencemag.org SCIENCE

Published by AAAS
New middle chapter in the story of human evolution
Nathan K. Schaefer and Beth Shapiro

Science 365 (6457), 981-982.

DOI: 10.1126/science.aay3550

Downloaded from http://science.sciencemag.org/ on September 11, 2019

ARTICLE TOOLS http://science.sciencemag.org/content/365/6457/981

RELATED http://science.sciencemag.org/content/sci/365/6457/eaat7487.full
CONTENT

REFERENCES This article cites 12 articles, 5 of which you can access for free
http://science.sciencemag.org/content/365/6457/981#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title
Science is a registered trademark of AAAS.