AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 108:137–146 (1999)

Genealogy Reconstruction From Short Tandem Repeat

Genotypes in an Amazonian Population
FRANCESC CALAFELL,1 AUDREY SHUSTER,1 WILLIAM C. SPEED,1
JUDITH R. KIDD,1 FRANCIS L. BLACK,2 AND KENNETH K. KIDD1,3*
1Department of Genetics, Yale University School of Medicine,

New Haven, Connecticut 06520-8005

2Department of Epidemiology and Public Health, Yale University School

of Medicine, New Haven, Connecticut 06520-8005

3Department of Psychiatry, Yale University School of Medicine,

New Haven, Connecticut 06520-8005

KEY WORDS microsatellites; Surui; parentage

ABSTRACT We have reconstructed partial genealogies in a sample of 44

SW Amazonian Rondonian Surui, in which 45 dinucleotide short tandem
repeat polymorphisms had previously been typed. The genotypes of 488 pairs
of individuals having an age difference of 13 or greater were compared, and
parentage was excluded if a pair failed to share an allele at more than one
locus. In order to test the power of this method, we computed the expected
distribution of the number of exclusionary loci for such pairs of unrelated
individuals, as well as that for individuals with different degrees of related-
ness, and compared it to the observed distribution. We estimated that the
pairs compared contained ⬃20% of individual pairs with a first-cousin
relation or closer. A total of 25 pairs were identified as possible parent-child. In
three instances, we could identify two or more children having a common
parent; we computed a relatedness coefficient in order to establish whether
the children were full or half sibs. The genealogies inferred show that
instances of polygyny and polyandry (or, alternatively, serial mating), in
addition to apparent monogamy, can be found among the Surui. The Surui
sample can be used as a model for paleoanthropological populations, in which
the determination of relatedness can provide further insights into the social
structure of past populations. We estimate that, depending on the history of
the populations and the degree of inbreeding, 10–20 highly informative
nuclear loci should be typed in order to infer genealogies with acceptable
confidence. Am J Phys Anthropol 108:137–146, 1999. r 1999 Wiley-Liss, Inc.

The advent of PCR and automated typing terms of relatedness, genealogy, mating
of short tandem repeat polymorphisms structure, the differential genetic contribu-
(STRPs) (also known as microsatellites) has tion of some individuals to the next genera-
allowed the typing of large numbers of indi- tion, population substructure, or recent ad-
viduals for numerous genetic markers with
relative ease and speed. Thus, genotypic
information for a large number of highly Grant sponsor: USPHS; Grant numbers: MH39329, MH50390;
Grant sponsor: NSF; Grant number: SBR-9632509; Grant spon-
informative genetic markers has become sor: CIRIT.
available for population samples (Bowcock Francesc Calafell’s current address is Unitat d’Antropologia,
Facultat de Biologia, Universitat de Barcelona, Diagonal 645,
et al., 1994; Deka et al., 1995; Jorde et al., 08028 Barcelona, Catalonia, Spain.
1995, 1997; Pérez-Lezaun et al., 1997). This *Correspondence to: Kenneth K. Kidd, Department of Genetics,
Yale University School of Medicine, 333 Cedar St., New Haven,
information can be used to explore the inter- CT 06520-8005. E-mail: [email protected]
nal structure of the population sampled in Received 2 September 1997; accepted 17 October 1998.

r 1999 WILEY-LISS, INC.

138 F. CALAFELL ET AL.

mixture. Intrapopulation analysis (Queller They also have been included in global sur-
et al., 1993) based on STRPs was carried out veys of allele frequency variation at indi-
by Morin et al. (1995) in samples of chimpan- vidual loci: DRD2 (Castiglione et al., 1995),
zees, and it is increasingly common in other CD4 (Tishkoff et al., 1996a), PLAT (Tishkoff
wild populations of animals (many articles et al., 1996b), and DRD4 (Chang et al., 1996).
in recent issues of Molecular Ecology and
other journals are devoted to genealogy re- MATERIALS AND METHODS
construction and relatedness in a number of The Surui sample comprised 44 individu-
species, usually for applications in conserva- als for whom age (ranging from 2–50 years)
tion biology), but, to the best of our knowl- and sex was known, and the ethnographic
edge, no equivalent analysis has been under- information collected indicated the possibil-
taken in humans. Usually information ity that the sample included several nuclear
gathered by field workers from the subjects families. The genealogical data collected was
sampled is sufficient to establish the related- of a social nature: the fieldworkers were
ness among the individuals in the sample, informed about who a man’s wives and
and demographic studies can provide accu- children were but not which, if any, wife was
rate descriptions of mating patterns, en- the mother of which child or children. It was
dogamy, population substructure, and migra- not inquired whether the children were off-
tion. However, this is not always the case. spring from an earlier marriage of the wife.
Populations in which socially defined rela- If a son or daughter were married and had
tions are markedly different from biological his or her own family, he or she would never
relations (e.g., due to frequent adoption or be described a someone’s offspring.
multiple sexual partners), may not have an Forty-five STRP markers from ABI Prism
accurate record of biological relations among Linkage Mapping Sets, panels 13–16, map-
its members. In different scenarios, samples ping to chromosomes 9, 10, and 11, were
may be available without accompanying ge- typed as described in Calafell et al. (1997).
nealogical data (e.g., archival samples), and Allele frequencies can be retrieved from the
paleoanthropological samples do not com- Web site 7http://info.med.yale.edu/genetics/
monly display unequivocal indicators of bio- kkidd8. Genotype frequencies (inferred as-
logical relationship among individuals. suming codominant inheritance and no null
In the present paper, we demonstrate the alleles) did not show statistically significant
feasibility of partial genealogy reconstruc- departure from Hardy-Weinberg expecta-
tion in a traditional population (the SW tions. Mean heterozygosity was 0.5488, with
Amazonian Surui), which can also serve as a a range from 0.0503–0.8103.
model for a paleoanthropological popula- In the absence of complete and accurate
tion, in which 45 dinucleotide STRPs were genealogical information, the genotypes of
typed. The Surui of Rondônia (they should all 488 pairs of individuals with an age
not be confused with the Surui-Aikewara of difference of 13 or greater were compared,
Pará) constituted a population of 500 indi- and parentage was excluded if more than
viduals around the time of fieldwork (1986) one marker showed exclusionary genotypes
(Santos and Coimbra, 1996). They speak a (i.e., the pair of individuals did not share at
language of the Tupi stock and are polygy- least one allele). If exactly one incompatible
nous and patrilocal (Mindlin, 1985). The locus was found, allele calling was re-
sample analyzed was from a single, 85 inhab- checked for possible allele miscalling. If the
itant village. Previous genetic studies on allele calling was confirmed, a pair not shar-
this Surui sample include the major histo- ing one allele at exactly one locus was still
compatibility complex (HLA) and immuno- considered nonexclusionary, given the possi-
globulin allotypes (Black, 1991; Bhatia et bility of mutation.
al., 1995), serum cholinesterases (Guerreiro et In order to assess the power of the loci
al., 1989), 31 blood groups and protein poly- analyzed in positively identifying parent-
morphisms (Callegari-Jacques et al., 1994), child pairs, we computed the probability
and 30 nuclear restriction fragment length that a pair of unrelated individuals does not
polymorphisms (RFLPs) (Kidd et al., 1991). show any exclusionary autosomal geno-
 GENEALOGY RECONSTRUCTION FROM STRP GENOTYPES 139

types, assuming Hardy-Weinberg propor- ull (1976). In particular, the expectation and
tions and independence across loci, as variance of the number of exclusionary loci
k are E(X) ⫽ ⌺pex and Var(X) ⫽ ⌺pex(1 ⫺ pex ).
pne,ur ⫽ 兿 (1 ⫺ p
l⫽1
exl(ur)) (1) The distribution of the number of exclusion-
ary loci among all pairs of individuals in the
Surui sample having an age difference 13 or
where
greater was generated and compared to that
n(l)
expected under different degrees of related-
pexl(ur) ⫽ 兺p
i⫽1
2
il (1 ⫺ pil)2 ness: unrelated individuals, first cousins,
half sibs, and full sibs. The expected distribu-
n(l) (2)
tions were obtained through a Monte Carlo
⫹2 兺p p (1 ⫺ p p )
i⫽j
il jl il jl
2
procedure: one million pairs of individuals
i⬎j were generated at random and were as-
signed genotypes at 45 loci according to the
and k is the number of loci, n(l) is the allele frequencies observed in 45 STRP loci
number of alleles at locus l, and pil and pjl in the Surui. From those, and from extra
are the relative frequencies, respectively, of simulated individuals when necessary, one
the ith and jth allele at locus l. The sub- million pairs each of full sibs, half sibs, and
scripts ne, exl, r, and ur refer to nonexclu-
first cousins were generated. The genotypes
sion, exclusion, related, and unrelated, re-
within each pair were compared and the
spectively. The chance that two unrelated
number of loci at which the pair did not
individuals are not excluded from being
share at least one allele (i.e., the exclusion-
parent and child (equation 1) is simply the
ary loci) were counted.
product of the probabilities of each indi-
Next, we tested whether the groups of
vidual locus showing no exclusion. The
individuals for whom one individual had
chance that a locus shows exclusion (equa-
been identified as their common parent were
tion 2) is the probability that the two indi-
viduals are homozygotes for different alleles full or half sibs. For every pair of individuals
or that one of the two is a heterozygote and (x and y) sharing a parent, we computed the
the other does not have these two alleles. relatedness coefficient r (slightly modified
Chakraborty and Jin (1993a) presented a from Queller and Goodnight, 1989):
formally different but equivalent formula- k 2

tion for equation 2. If the individuals in the

pair are related, the probability that they do
兺 兺 (l
l⫽1 a⫽1
y ⫺ p*)
r⫽ (4)
not share any allele identical by state (IBS) k 2
is 兺 兺 (1 ⫺ p*)
l⫽1 a⫽1
k

pne,r ⫽
l⫽1
兿 (1 ⫺ f (r)p 0 exl(ur)) (3) where a is an index for allelic position, p* is
the population frequency of the allele found
where f0 (r) is the expected fraction of pairs in individual x at allelic position a, corrected
presenting no alleles identical by descent for by dropping individuals x and y from the
a particular degree of relatedness r. In par- sample, and Iy is an indicator variable that
ticular, f0 ⫽ 0.25 for full sibs, f0 ⫽ 0.5 for half is set to 1 if y presents the same allele as x at
sibs or uncle-niece pairs (throughout this position a and becomes 0 if x and y have
article, we use the phrase uncle-niece as different alleles at that position. r is an
shorthand for all possible gender combina- estimate of the probability that one random
tions between an individual and his or her allele from x and another random allele from
full sibs’ children), and f0 ⫽ 0.75 for simple y are identical by descent. For full sibs of
first cousins (Chakraborty and Jin, 1993b). unrelated parents, that probability is 0.5,
We can extend this analysis to predict the and for half sibs with unrelated, nonshared
distribution of pairs showing a particular parents, r ⫽ 0.25. However, the estimation
number of exclusionary loci, following the of r through equation 4 carries a relatively
method described by Chakraborty and Sch- large standard deviation (Queller and Good-
140 F. CALAFELL ET AL.

night, 1989). We estimated the empirical and 22% of full sibs would not be excluded as
distribution of r from 10,000 simulated pairs parent-offspring dyads (Fig. 1).
each of random individuals, half sibs, and Twenty-five pairs of individuals met the
full sibs, with genotypes drawn randomly criteria for being possible parent-offspring
with replacement with alleles at the frequen- pairs: they had an age difference of 13 years
cies observed in the Surui sample (Blouin et or greater, and they shared an allele at a
al., 1996). From those distributions, we esti- minimum of 44 of the 45 loci typed. How-
mated confidence intervals for r in unrelated ever, as we have estimated above, full sibs
individuals, half sibs, and full sibs. The have a relatively high probability of meeting
empirical distributions allow for testing the genetic criterion for parent-offspring
whether an actual r value falls within the pairs. To address this issue, we analyzed the
range expected for half sibs or for full sibs. distribution of the number of nonexclusion-
ary loci per individual pair in order to esti-
RESULTS mate the expected number of putative par-
The genotypes at 45 STRP loci of all pairs ent-child pairs that have in fact other degrees
of individuals in the Surui sample having an of relatedness. The distribution of the num-
age difference greater or equal to 13 years ber of expected exclusionary loci given the
were compared. Thirty-five individuals were observed allele frequencies was estimated
entered as possible children (i.e., there was as described in Materials and Methods for
at least one other individual among the 44 in unrelated individuals as well as for different
the sample who was 13 or more years older), degrees of relatedness: first cousins, half
15 as candidate fathers (i.e., at least one sibs or uncle-niece, and full sibs.
individual among the 44 in the sample was The distribution observed for pairs of indi-
13 or more years younger), and 12 as candi- viduals in the Surui sample having an age
date mothers. Each individual ‘‘child’’ was difference of 13 years or greater plus the
tested against a mean of 10.03 candidate expected distributions under different de-
fathers (range: 3–15) and against a mean of grees of relatedness are shown in Figure 2.
3.94 candidate mothers (range: 0–12). For a The observed distribution is roughly bell-
total of 25 of the possible children, one shaped, with a mode at eight loci, a maxi-
individual or two individuals of different sex mum of 19, and a secondary mode at zero
among the possible parents showed no exclu- exclusionary loci. The mean number of exclu-
sionary loci. We did not find any case in sionary loci observed was 8.20, with a vari-
which a child failed to show incompatible ance of 9.98; the values expected if all the
genotypes with more than two individuals or individuals were unrelated were 8.29 and
with two individuals of the same sex. Given 6.18. If all the pairs of individuals analyzed
the allele frequencies observed (Calafell et had the same degree of relatedness, the
al., 1997) and from equations 1–3, the prob- mode of the number of exclusionary loci
ability of not observing any exclusionary would increase as the degree of relatedness
locus in a pair of unrelated individuals is decreased (Fig. 2): full sibs have a mode at
5.8 ⫻ 10⫺5 or 1/17,248, for a pair of simple two loci, half sib or uncle-niece pairs peak at
first cousins it is 8.92 ⫻ 10⫺4 (1/1,121), for four exclusionary loci, first cousins peak at
half sib or uncle-niece pairs it is 0.011 (1/91), six, and unrelated individuals show a mode
and for full sib pairs it is 0.113, or ⬃1/9. at eight loci, as the observed distribution
Thus, and considering the nature of the did. Presumably the kinship within every
sample, we cannot exclude that some of the pair of individuals we compared varies; more
pairs of individuals that share at least one remote multiple relationships would result
allele at every locus are in fact full sibs who in essentially a continuum of possible kin-
happen to be 13 or more years apart in age. ship values.
With the 21 most informative loci (Fig. 1), We used the four distributions of number
the probability of not excluding an unrelated of exclusionary loci with different degrees of
individual as a parent (i.e., the type I error relatedness to estimate the proportions of
rate) is slightly below 0.001; with those loci, the different relations found within the pairs
0.7% of first-cousin pairs, 4.3% of half sibs, of individuals by finding the proportions
 GENEALOGY RECONSTRUCTION FROM STRP GENOTYPES 141

Fig. 1. Probability of not excluding a pair of Surui individuals as parent-child given different degrees
of relatedness, estimated from equations 1–3, using the observed allele frequencies for the 45 STRP loci
typed. Loci are sorted by decreasing informativeness ( pex in equation 2).

that would fit best (by a least-square crite- to create a bias towards parent-child pairs
rion) the observed distribution, conditional and against sib pairs. Moreover, the mean
on the frequency of nonexclusions being age difference between nonexclusionary pairs
equal to that observed. The best fit was is 27.60 ⫾ 1.82 years, with only four cases
obtained when the pairs in the sample com- below 20 years, and significantly larger than
prised 79.6% unrelated pairs, 12.4% half the age difference within exclusionary pairs
sibs or uncle-niece, 3.4% full sibs, 4.6% (23.18 ⫾ 0.38 years, P ⫽ 0.006, Mann-Whit-
parent-child pairs, and less than 0.1% first ney’s U test). Therefore, in our next analyses
cousins. However, the mixed distribution we will treat all nonexclusionary pairs as
(Fig. 2) was statistically significantly differ- parent-child pairs.
ent from the observed distribution (␹2 ⫽ Ignoring for the moment redundancy of
43.88, 13 d.f., P ⯝ 3 ⫻ 10⫺5), probably due to individuals in the various pairs, we found
the heavier right tail of the observed distri- the following nonexclusionary pairs. Consid-
bution. These estimated proportions imply ering the 35 individuals young enough to
that the 25 nonexclusionary pairs are ex- have a potential parent in the sample, four
pected to comprise 22.45 parent-child pairs, had both a mother and a father, 14 had only
1.87 full sibs, 0.66 half sib or uncle-niece a father, and three only a mother. Out of the
pairs, and 0.02 unrelated pairs. However, 15 men old enough to be fathers, eight had
the fact that we imposed an age difference of children in the sample, as did five out of 12
at least 13 years within the pairs is expected women. Three men had six, five, and two
142 F. CALAFELL ET AL.

Fig. 2. Number of loci showing an exclusion in a butions if all pairs had the same degree of relatedness.
putative parent-child pair (i.e., the two genotypes do not The best-fit mix is a linear combination of the distribu-
present any shared allele) from 45 STRP loci. Bars tions for unrelated individuals, first cousins, half sibs,
represent the observed number of exclusionary loci in and full sibs, obtained by minimizing the squared
488 pairs of Surui individuals; lines are expected distri- difference with the observed distribution.

children each; one woman had three chil- share the same father, r ranged from 0.120–
dren in the sample, and five men and four 0.319, with one exception (JK1476-JK1504).
women had one child each. Thus, four sib- Again with that exception, we could reject
ships, with six, five, three, and four sibs, with P ⬍ 0.02 that r was within the distribu-
respectively, were recognized; we will label tion for full sibs, but we could not reject that
them A, B, C, and D (Fig. 3). Sibship C con- r was within the range for half sibs. For
sists of three children who shared a mother, JK1476-JK1504, r ⫽ 0.458, with P ⫽ 0.277
but we identified a different putative father for r ⫽ 0.5 and P ⫽ 0.006 for r ⫽ 0.25. The
for each child. One of those men was also the three individuals JK1511, JK1514, and
putative father of five other children in JK1519 had among them more than two
sibship A. With that exception, we identified different nonpaternal alleles at five loci. The
the putative fathers but not the mothers of overall evidence points to JK1476 and
children in sibships A, B, and D. JK1504 sharing a mother and JK1511,
The empirical distributions of r for ran- JK1514, and JK1519 being children of three
dom pairs, half sibs, and full sibs are shown different women. In sibship B (Fig. 3), all but
in Figure 4. We computed r relatedness one r value fell in the range 0.405–0.583,
coefficients for all pairs of individuals in with P ⬍ 0.05 for belonging to the empirical
sibships A, B, and D. For pairs of offspring distribution for half sibs and P ⬎ 0.05 for
from sibship A, all of whom are inferred to belonging to that of full sibs. The pair
 GENEALOGY RECONSTRUCTION FROM STRP GENOTYPES 143

Fig. 3. Inferred relations among the Surui sample. Crossed symbols represent inferred individuals not
present in the sample. Not shown: three mother-child pairs, one father-child, and one father-mother-child.
Pedigree A refers to the children of JK1506, and pedigree C is defined by the children of JK1468.
Pedigrees B and D refer to the only other inferred pedigrees with more than one child.

JK1492-JK1505 showed a paradoxical r ⫽ 45 genotypes of all five sibs taken together.

0.235, which is lower than 0.5 with a nomi- Thus, and considering the multiple tests
nal P ⫽ 0.008. Obviously, JK1492 and done, the most likely explanation for the low
JK1505 cannot be half sibs to each other and relatedness coefficient between JK1492 and
share three full sibs. At all loci, we were able JK1505 is a chance effect. JK1502 and
to reconstruct a maternal genotype from the JK1527 (sibship D, Fig. 3) present r ⫽ 0.453,
144 F. CALAFELL ET AL.

Fig. 4. Frequency distributions of the r relatedness coefficient (Queller and Goodnight, 1989) obtained
from 10,000 simulated pairs of, from left to right, unrelated individuals, half sibs, and full sibs. Individual
genotypes were simulated by drawing randomly with replacement from the allele frequencies observed in
45 loci in the Surui. Ninety-five percent empirical confidence intervals are, respectively, [⫺0.153, 0.178],
[0.093, 0.416], and [0.340, 0.662].

which is larger than 0.25 with P ⫽ 0.007, Surui showed also the lowest heterozygosity
and we can accept that they share a mother. in a survey of HLA class I antigens among 22
native South American populations (Bhatia
DISCUSSION
et al., 1995). The type I error for unrelated
We have identified 25 parent-offspring individuals (i.e., the probability that a pair
dyads in a sample of 44 Rondônian Surui of unrelated individuals would not show
individuals, in which 45 STRP loci had been exclusionary genotypes) is 439 to over 1010
typed. Overall, we were able to place 34 of times lower in the other populations studied
the 44 individuals in family groups. Even for the same set of 45 loci. Using the 21 most
with the sparse genealogical information informative loci, the type I error for unre-
collected from the field, our genetic results lated individuals in the Surui is roughly
were confirmed by at least 13 socially de- 0.001; only the top seven loci would be
fined relationships. The same loci had also needed in African populations in order to
been typed in nine other populations from achieve the same error rate, (Biaka and
worldwide samples (Calafell et al., 1997). Mbuti Pygmies) seven to eight in Europeans
The Surui sample showed the lowest aver- (Danes and Druze), 10–11 in East Asians
age heterozygosity among the populations (Chinese, Japanese, and Yakut), 13 in
typed, which results in reduced informative- Melanesians (Nasioi), and 11 in the Maya.
ness for parentage testing. Moreover, the The low informativeness of the loci typed,
 GENEALOGY RECONSTRUCTION FROM STRP GENOTYPES 145

which is likely to have been caused by method, we could confidently classify all sib
bottleneck events in the initial colonization pairs tested into half or full sibs. To the
of South America (Calafell et al., 1997) as extent that our sample represents the Surui
well as in subsequent demographic crises, community it was drawn from, we can ob-
and the presence of individuals with rela- serve different mating patterns: a man had
tions other than parent-child implies that five children with the same woman, whereas
additional caution should be used in infer- another man had six children with five
ring genealogies. However, the large age different women, one of whom had two other
difference observed in nonexclusionary pairs children with two different men. The latter
is reassuring, especially because it is signifi- cases can be attributed to polygamy and/or
cantly higher than the age difference among to serial mating. This kind of information
exclusionary pairs. can be incorporated into any genetic model
The distribution of the number of exclu- in which effective population size is re-
sionary loci among pairs of individuals with quired. The difference in relatedness be-
an age difference of 13 years or greater was tween full and half sibs is maximal when the
closest to that of a sample consisting of nonshared parents of half sibs are unrelated
mostly unrelated individuals, with ⬃20% of to each other. However, if the unshared
pairs having first cousin relatedness or parents are full sibs to each other (and in
closer. However, the distribution of exclusion- fact the preferred wives for a Surui man are
ary loci showed a right tail that was heavier his sisters’ daughters [Mindlin, 1985]), r
than expected based on the best-fit mix of would become 0.375. For a pair of half sibs in
the distributions for unrelated individuals that situation, it would be often difficult to
or for individuals with various degrees of discern their immediate relation, and they
relatedness. A few individuals were overrep- could be easily deemed full sibs.
resented in the pairs showing the highest The genetic determination of parentage
number of exclusionary loci. Those individu- and other degrees of relatedness has an
als clearly clustered with the Surui when a important application in archeology and pa-
neighbor-joining tree based on the propor- leoanthropology. The extent of the related-
tion of shared alleles (computed as described ness of individuals interred in the same
by Bowcock et al., 1994) was constructed cemetery or in the same grave can provide
with over 500 individuals from ten world- important clues in the reconstruction of the
wide populations (data not shown); thus, it social structure of ancient populations. Popu-
is more likely that those individuals (two lation analysis based on ancient mtDNA has
women) represent recent migrants from been undertaken in New World populations
neighboring populations rather than non– (Stone and Stoneking, 1993; Parr et al.,
Native American admixture. 1996). However, autosomal nuclear markers
Although the relatedness coefficient r are needed to establish with a reasonable
(Queller and Goodnight, 1989) has a large level of certainty the degree of relatedness
variance, it takes into account information between individuals. Genotyping of STRPs
on allele frequencies, whereas other meth- from ancient DNA has been attempted with
ods, based on the extent of allele sharing mixed results (Hauswirth et al., 1994; Ra-
(Morin et al., 1995), probably do not extract mos et al., 1995; Zierdt et al., 1996), and
all the information available (Blouin et al., relatively large amounts of undegraded
1996). Given the possible complex pattern of DNA would be required for reliable genotyp-
relatedness among the individuals, the rela- ing of STRPs. These restrictions would prob-
tively broad overlap between the confidence ably impose a relatively recent age for the
intervals of r for different degrees of related- samples that can be typed, and thus it is
ness, and the low informativeness of genetic likely that the populations studied would be
markers in the Surui, we limited our analy- close temporally and genetically to contem-
sis of relatedness to those individuals for porary populations. The markers typed for
whom one common parent had been identi- an intrapopulation genetic analysis should
fied and tested whether they were more be selected carefully. Given that ancient
likely to be half or full sibs. Using this DNA analysis and in particular ancient
146 F. CALAFELL ET AL.

nuclear DNA typing can be extremely time cies at the human dopamine D4 receptor locus. Hum
Genet 98:91–101.
consuming, it is paramount to select those Deka R, Shriver MD, Yu LM, DeCroo S, Hundrieser J,
markers that are expected to provide the Bunker CH, Ferrell RE, Chakraborty R. 1995. Popula-
most information. Data from modern popula- tion genetics of dinucleotide (dC-dA)n (dG-dT)n poly-
tions can be used to infer the expected morphisms in world populations. Am J Hum Genet
56:461–474.
informativeness of the loci and in turn can Guerreiro JF, Santos SEB, Aguiar GFS. 1989. Serum
provide a framework against which to test cholinesterase polymorphisms (CHE1 and CHE2 loci)
for genetic continuity through time. among several Indian regions from Amazon region of
Brazil, and segregation of the C5 variant in families.
Gene Geogr 3:11–20.
ACKNOWLEDGMENTS Hauswirth WW, Dickel CD, Rowold DJ, Hauswirth MA.
1994. Inter- and intrapopulation studies of ancient
The Applied Biosystems Division of Per- humans. Experientia 50:585–591.
kin Elmer provided Linkage Mapping Sets Jorde LB, Bamshad MJ, Watkins WS, Zenger R, Fraley
AE, Krakowiak PA, Carpenter KD, Soodyall H, Jen-
and other materials for this study, which kins T, Rogers AR. 1995. Origins and affinities of
was also supported in part by USPHS grants modern humans: a comparison of mitochondrial and
MH39329 and MH50390 (to K.K.K.) and nuclear genetic data. Am J Hum Genet 57:523–538.
NSF grant SBR-9632509 (to J.R.K.). F.C. is Jorde LB, Rogers AR, Bamshad MJ, Watkins WS, Kra-
kowiak P, Sung S, Kere J, Harpending HC. 1997.
the recipient of a postdoctoral fellowship Microsatellite diversity and the demographic history
from CIRIT, autonomous government of of modern humans. Proc Natl Acad Sci USA 94:3100–
Catalonia. 3103.
Kidd JR, Black FL, Weiss KM, Balazs I, Kidd KK. 1991.
Studies of three Amerindian populations using nuclear
LITERATURE CITED DNA polymorphisms. Hum Biol 63:775–794.
Mindlin B. 1985. Nós paiter, os Surui de Rondônia.
Bhatia KK, Black FL, Smith TA, Prasad ML, Koki GN. Petrópolis, Brasil: Vozes.
1995. Class I HLA antigens in two long-separated Morin PA, Moore JJ, Chakraborty R, Jin L, Goodall J,
populations: Melanesians and South Amerinds. Am J Woodruff DS. 1995. Kin selection, social structure,
Phys Anthropol 97:291–305. gene flow, and the evolution of chimpanzees. Science
Black FL. 1991. Reasons for failure of genetic classifica- 265:1193–1201.
tions of South Amerind populations. Hum Biol 63:763– Parr RL, Carlyle SW, O’Rourke DH. 1996. Ancient DNA
774. analysis of Fremont Amerindians of the Great Salt
Blouin MS, Parsons M, Lacaille V, Lotz S. 1996. Use of Lake wetlands. Am J Phys Anthropol 99:507–518.
microsatellite loci to classify individuals by related-
Pérez-Lezaun A, Calafell F, Mateu E, Comas D, Ruiz-
ness. Mol Ecol 5:393–401.
Bowcock AM, Ruiz-Linares A, Tomfohrde J, Minch E, Pacheco R, Bertranpetit J. 1997. Microsatellite varia-
Kidd JR, Cavalli-Sforza LL. 1994. High resolution of tion and the differentiation of modern humans. Hum
human evolutionary trees with polymorphic microsat- Genet 99:1–7.
ellites. Nature 368:455–457. Queller DC, Goodnight KF. 1989. Estimating related-
Calafell F, Shuster A, Speed WC, Kidd JR, and Kidd KK. ness using genetic markers. Evolution 43:258–275.
1997. Short tandem repeat polymorphism in humans. Queller DC, Strassmann JE, Hughes CR. 1993. Microsat-
Eur J Hum Genet 6:38–49. ellites and kinship. Tree 8:285–288.
Callegari-Jacques SM, Salzano FM, Weimer TA, Hutz Ramos MD, Lalueza C, Girbau E, Pérez-Pérez A, Que-
MH, Black FL, Santos SEB, Guerreiro JF, Mestriner vedo S, Turbón D, Estivill X. 1995. Amplifying dinucle-
MA, Pandey JP. 1994. Further blood genetic studies tode microsatellite loci from bone and tooth samples of
on Amazonian diversity—data from Indian groups. up to 5000 years of age: more inconsistency than
Ann Hum Biol 21:465–481. usefulness. Hum Genet 96:205–212.
Castiglione CM, Deinard AS, Speed WC, Sirugo G, Santos RV, Coimbra CEA Jr. 1996. Socioeconomic differ-
Rosenbaum HC, Zhang Y, Grandy DK, Grigorenko entiation and body morphology in the Suruı́ of South-
EL, Bonné-Tamir B, Pakstis AJ, Kidd JR, Kidd KK. western Amazonia. Curr Anthropol 37:851–856.
1995. Evolution of haplotypes at the DRD2 locus. Am Stone AC, Stoneking M. 1993. Ancient DNA from a
J Hum Genet 57:1445–1456. pre-Columbian Amerindian population. Am J Phys
Chakraborty R, Jin L. 1993a. Determination of related- Anthropol 92:463–471.
ness between individuals using DNA fingerprinting. Tishkoff SA, Dietzch E, Speed W, Pakstis AJ, Kidd JR,
Hum Biol 65:875–895.
Cheung K, Bonné-Tamir B, Santachiara-Benerecetti
Chakraborty R, Jin L. 1993b. A unified approach to
study hypervariable polymorphisms: statistical consid- S, Moral P, Krings M, Pääbo S, Watson E, Risch N,
erations of determining relatedness and population Jenkins T, Kidd KK. 1996a. Global patterns of linkage
distances. In: Pena SDJ, Chakraborty R, Epplen JT, disequilibrium at the CD4 locus and modern human
Jeffreys AJ, (editors). DNA fingerprinting: state of the origins. Science 271:1380–1387.
science. Basel, Switzerland: Birkhäuser Verlag, pp. Tishkoff SA, Ruano G, Kidd JR, Kidd KK. 1996b.
153–175. Distribution and frequency of a polymorphic Alu
Chakraborty R, Schull WJ. 1976. A note on the distribu- insertion at the plasminogen activator locus. Hum
tion of the number of exclusions to be expected in Genet 97:759–764.
paternity testing. Am J Hum Genet 28:615–618. Zierdt H, Hummel S, Herrmann B. 1996. Amplification
Chang F-M, Kidd JR, Livak KJ, Pakstis AJ, Kidd KK. of human short tandem repeats from medieval teeth
1996. The world-wide distribution of allele frequen- and bone samples. Hum Biol 68:185–199.