Am. J. Hum. Genet. 65:1493–1500, 1999

1493

Long Homozygous Chromosomal Segments in Reference Families from the
Centre d’Étude du Polymorphisme Humain
Karl W. Broman and James L. Weber
Marshfield Medical Research Foundation, Marshfield, WI

Summary

Using genotypes from nearly 8,000 short tandem-repeat
polymorphisms typed in eight of the reference families
from the Centre d’Étude du Polymorphisme Humain
(CEPH), we identified numerous long chromosomal seg-
ments of marker homozygosity in many CEPH individ-
uals. These segments are likely to represent autozygosity,
the result of the mating of related individuals. Confi-
dence that the complete segment is homozygous is gained
only with markers of high density. The longest segment
in the eight families spanned 77 cM and included 118
homozygous markers. All individuals in family 884
showed at least one segment of homozygosity: the father
and mother were homozygous in 8 and 10 segments with
an average length of 13 and 16 cM, respectively, and
covering a total of 105 and 160 cM, respectively. The
progeny in family 884 were homozygous over 5–16 seg-
ments with average length 11 cM. The progeny in family
102 were homozygous over 4–12 segments with average
length 19 cM. Of the 100 individuals in the other six
families, 1 had especially long homozygous segments,
and 19 had short but significant homozygous segments.
Our results indicate that long homozygous segments are
common in humans and that these segments could have
a substantial impact on gene mapping and health.

Introduction

The reference families from the Centre d’Étude du Poly-
morphisme Humain (CEPH) (White and Lalouel 1988;
Dausset et al. 1990) were recruited in the effort to con-
struct the first human genetic maps. Lymphoblastoid cell
lines derived from the CEPH individuals have provided

Received May 19, 1999; accepted for publication September 28,
1999; electronically published November 5, 1999.

Address for correspondence and reprints: Dr. Karl W. Broman,
Department of Biostatistics, School of Hygiene and Public Health,
Johns Hopkins University, 615 North Wolfe Street, Baltimore, MD
21205–2179. E-mail: kbroman@jhsph.edu

q 1999 by The American Society of Human Genetics. All rights reserved.
0002-9297/1999/6506-0004$02.00

a nearly limitless supply of DNA, making these families
available for genotyping by investigators around the
world. Many thousands of short tandem-repeat poly-
morphisms (STRPs) have been genotyped within a subset
of eight of the CEPH families. These data provide a
uniquely comprehensive view of the genomes of these
individuals, which allows analyses that would not be
possible on the basis of data from a more typical genome
scan of 400 markers.

We recently constructed new genetic maps based on
these families (Broman et al. 1998). As part of that work,
we screened the data for apparent tight double-recom-
bination events indicative of genotyping errors or mu-
tations. In the process, we identified several long seg-
ments of noninformative markers in family 884, caused
by long stretches of homozygous markers in the parents
of that family.

These segments likely represent autozygosity. An in-
dividual is autozygous at a locus if he or she received
two copies of a single ancestral allele at the locus—that
is, if the two alleles are identical by descent (Hartl and
Clark 1997). Autozygosity occurs when a couple shares
a chromosomal segment identical by descent and both
transmit the segment to one of their offspring.

Autozygosity may be detected at a single locus if ge-
netic data are available on many individuals in an inbred
pedigree. More generally, one needs to look at sets of
linked markers: if an individual is homozygous at a large
number of contiguous markers, it is likely that he or she
is autozygous for the segment (i.e., that the two chro-
mosomal segments have a common origin).

We developed a LOD score to characterize the sig-
nificance of a segment of homozygosity, to distinguish
autozygosity from chance homozygosity. We calculated
this score for all possible subsets of contiguous auto-
somal markers in all 134 individuals from the eight
CEPH families, in order to discover autozygous seg-
ments. All individuals in family 884 showed at least one
segment of homozygosity, as did all of the progeny in
family 102. Total lengths of homozygous segments in
these individuals were 62–253 cM. Twenty percent of
the individuals in the other six families also had signif-
icant homozygous segments. These unexpected findings



1494 Am. J. Hum. Genet. 65:1493–1500, 1999

Table 1

Probabilities of Observed Genotype at a Marker, Given the
Autozygosity Status of the Surrounding Segment

OBSERVED
GENOTYPE

PROBABILITY THAT
SEGMENT IS

RATIOAutozygous
Not

Autozygous

AA 2(1 2 «)p 1 «pA A
2pA (1 2 «)/p 1 «A

AB 2«pA pB 2pApB «

NOTE.—« denotes the combined rate of genotyping errors
and mutations.

indicate that long homozygous segments are common in
humans.

Material And Methods

Genetic Markers and Genotype Data

We used the data described by Broman et al. (1998);
we summarize them again here. We considered eight of
the CEPH families (102, 884, 1331, 1332, 1347, 1362,
1413, and 1416), excluding individuals 1332-09 and
1416-10, on whom few genotype data were available.
These families contain a total of 134 individuals, in-
cluding 92 progeny, 16 parents, and 26 grandparents.
We considered genotypes for 7,746 autosomal STRPs,
including 5,064 from Généthon (Dib et al. 1996), 1,334
from CHLC (Sheffield et al. 1995; Sunden et al. 1996),
849 from the Utah Marker Development Group (1995),
285 from the Center for Medical Genetics, Marshfield
Medical Research Foundation, 35 telomeric markers
(Rosenberg et al. 1997), and 179 miscellaneous markers.
The genotypes for families 884, 1331, 1332, and 1362
were 96% complete. Families 102, 1347, 1413, and
1416 were not typed for the 849 Utah markers, but their
genotypes were 94% complete for the other 6,897 mark-
ers. All the data are available at the Center for Medical
Genetics, Marshfield Medical Research Foundation Web
site.

The data had previously been cleaned of genotypes
causing apparent tight double-recombination events in-
dicative of genotyping errors or mutations. The genetic
maps and marker order were as determined by Broman
et al. (1998). The total sex-averaged genetic length of
the 22 autosomes was 3,488 cM, corresponding to ap-
proximately one marker per 0.5 cM. In the following,
all reported genetic distances are sex averaged. Allele
frequencies were estimated on the basis of the 28 found-
ing individuals in the eight families. The average (SD)
of the marker heterozygosities was .68 (.13). Cytogenetic
locations of the markers, from Collins et al. (1996), were
obtained from the Genetic Location Database at the Uni-
versity of Southampton.

Identification of Long Homozygous Segments

Our goal was to identify subsets of contiguous mark-
ers for which an individual showed an unusual propor-
tion of homozygosity. For each individual, we looked at
all possible subsets of contiguous markers. For each such
subset, we calculated a LOD score comparing the hy-
potheses that the individual was or was not autozygous
(i.e., homozygous by descent) at all markers in the sub-
set. Large values of the statistic indicate segments in
which the individual showed a significant amount of
homozygosity.

In forming the LOD score, we assumed that the mark-
ers were in both linkage and Hardy-Weinberg equilib-

rium, and we used a simple model for genotyping errors
and mutations, much like that used by Broman and We-
ber (1998) in the context of the inference of pairwise
relationships. The LOD score has the following form:

k
Pr(g Fautozygous at i)iLOD(j,k) = log ,O 10 [ ]Pr(g Fnot autozygous at i)i=j i

where is the individual’s observed genotype at markergi
i. The assumed genotype probabilities are displayed in
table 1. In the model underlying the LOD score, it is
assumed that errors and mutations occur with proba-
bility « and that, when such events occur, genotypes are
obtained according to their population frequencies. This
model serves as a device for obtaining a statistic that
facilitates the detection of homozygous segments while
both taking account of the varying informativeness of
the markers and allowing the presence of a small pro-
portion of heterozygous markers within each segment;
it was not intended to reflect the details of the error and
mutation processes.

In calculating the LOD score, we assumed that the
combined rate of genotyping errors and mutations was

. A smaller value of « would give greater weight« = .001
to heterozygous markers, resulting in shorter identified
segments. If « were made larger (smaller) by a factor of
5, the LOD score for a segment would be increased
(decreased) by 0.7 for each heterozygous marker that it
contained; for example, if a segment containing a single
heterozygous marker had a LOD score of 7.9 when

, the LOD score would be ∼7.2 and ∼8.6 when« = .001
and .005, respectively.« = .0002

The above-described LOD score was calculated for
each possible subset of contiguous markers in each in-
dividual. When, for a particular individual, two subsets
of markers with large LOD scores overlapped, we re-
tained information only on that subset with the larger
LOD score.

Simulations under Linkage Equilibrium

Even under complete linkage equilibrium, individuals
may be homozygous for a number of contiguous mark-



Broman and Weber: Autozygosity in CEPH Families 1495

Figure 1 Minimum detectable length (in cM) of an autozygous
region for a genome of length 3,488 cM, with equally spaced markers
of constant heterozygosity.

Table 2

Homozygous Segments for Individual 884-01

Chromosome (Markers) Cytogenetic Band(s)
Length
(cM)

Proportion
Homozygous LOD Score

4 (D4S3021–D4S1597) q28-q31 14.8 49/50 26.58
6 (D6S1017–D6S1623) p21-p11 14.5 36/36 21.74
6 (D6S1631–D6S1580) q16-q22 13.2 49/49 27.85
7 (D7S2486–D7S800) q31-q32 12.8 38/38 20.18
11 (DllS1794–DllS4138) p15 1.6 15/16 5.46
12 (D12S354–D12S386) q24 23.4 36/38 17.08
19 (D19S605–D19S890) q13 9.8 11/11 5.37
21 (D21S1891–D21S1897) q22 15.0 20/20 12.78

ers somewhere in the genome, completely by chance. In
order to estimate the distribution of the above-described
LOD score under linkage equilibrium, we simulated the
genetic data for unrelated individuals134 # 20 = 2,680
having the same pattern of missing data as were seen in
the 134 CEPH individuals whom we studied. We as-
sumed Hardy-Weinberg and linkage equilibrium and
used the allele frequencies estimated on the basis of the
28 founding individuals in the eight families. For each
simulated genome, we calculated the maximum LOD
score, among all possible subsets of contiguous markers,
and the length of the subset with the maximum LOD
score.

To determine the effects of marker density and marker
informativeness on the minimum detectable length of an
autozygous segment, we simulated genomes that were
of length 3,488 cM and had markers that were equally
spaced, had constant heterozygosity, and were in linkage
equilibrium. To simplify the simulation, we assumed that
the genome consisted of a single long chromosome. For
each of various marker densities and heterozygosities,
we simulated the genetic data for 10,000 genomes and
calculated the 95th percentile of the maximum length

of contiguous homozygous markers. This percentile is
the minimum detectable length of an autozygous seg-
ment when a 5% genomewide type I–error rate is used.

Results

The minimum detectable length of an autozygous seg-
ment, as a function of marker density, in the case of
equally spaced markers with constant heterozygosity, is
displayed in figure 1. For markers with .7 heterozygosity,
an autozygous segment as short as 9 cM may be detected
when the markers are 1 cM apart, whereas the segment
must be >32 cM long if there are markers every 4 cM.
At the density of a more typical genome scan (i.e., 10
cM), only very long autozygous segments can be
detected.

The simulations under linkage equilibrium that made
use of the genetic maps and allele frequencies estimated
from the data indicated that, ∼5% of the time, an in-
dividual will have a segment of at least nine contiguous
homozygous markers, with a LOD score 14.67, some-
where in the genome. In the following, we present only
those segments that had LOD scores 14.67.

All individuals in family 884 had at least one ho-
mozygous segment with a LOD score 14.67. The father,
individual 884-01, had eight homozygous segments,
which are summarized in table 2. LOD scores for these
homozygous segments were as high as 27.85. One seg-
ment was ∼2 cM long; the others were 10–23 cM long.
Overall, these segments covered a total of 105 cM (3.0%
of the autosomal genome). The mother, individual
884–02, had 10 homozygous segments, which are sum-
marized in table 3. Five segments were !10 cM long;
three were 30–40 cM long. Overall, these segments cov-
ered a total of 160 cM (4.6% of the autosomal genome).

Table 4 contains a summary of the homozygous seg-
ments in the other members of family 884. Three of the
four grandparents in the family had multiple long ho-
mozygous segments; the other grandparent showed just
one homozygous segment of 11 markers. The 12 prog-
eny in the family showed 5–16 homozygous segments
with average length 11.0 cM and covering, on average,
97 cM (2.8% of the autosomal genome). These homo-



1496 Am. J. Hum. Genet. 65:1493–1500, 1999

Table 3

Homozygous Segments for Individual 884-02

Chromosome (Markers) Cytogenetic Band(s)
Length
(cM)

Proportion
Homozygous LOD Score

3 (D3S1571–D3S1617) q28 4.9 9/9 5.53
4 (GATA144E02–D4S189) p11-q12 11.1 21/21 12.26
5 (D5S398–D5S401) q11-q14 29.8 77/77 46.21
6 (D6S1711–D6S278) q11-q22 35.3 109/113 48.12
8 (D8S506–D8S385) q22-q23 8.0 28/30 12.35
9 (D9S1802–D9S250) q33 6.5 18/18 9.53
12 (D12S103–D12S1680) q13-q21 11.3 43/43 21.82
16 (D16S494–D16S3107) q21-q22 8.8 26/26 17.23
16 (D18S450–GATA51E05) q21-q22 40.3 84/84 49.79
22 (D22S1156–D22Sl179) q13 3.9 21/21 15.81

Table 4

Homozygous Segments for Progeny and Grandparents in Family
884

Individual
No. of

Segments
Length
(cM)

No. of
Markers

Total
(Average)
Length

Progeny:
3 9 2.8–39.1 14–93 111.5 (12.4)
4 5 2.8–25.8 15–77 61.7 (12.3)
5 5 11.7–15.7 14–58 65.7 (13.1)
6 9 3.2–17.9 9–64 93.1 (10.3)
7 9 3.8–17.9 13–39 84.0 (9.3)
8 16 1.7–36.2 9–87 195.8 (12.2)
9 8 4.2–19.5 9–71 77.6 (9.7)
10 6 5.9–12.9 13–35 57.0 (9.5)
11 10 2.1–36.2 15–90 124.8 (12.5)
12 7 1.7–16.5 7–32 66.4 (9.5)
13 9 4.2–30.0 9–64 118.0 (13.1)
14 13 2.7–15.9 14–45 109.7 (8.4)

Grandparents:
15 4 4.8–16.0 6–45 35.5 (8.9)
16 1 6.5 11 6.5 (6.5)
17 7 1.6–17.2 8–45 57.4 (8.2)
18 9 1.7–29.3 10–51 116.2 (12.9)

zygous segments derived from shared haplotypes in the
parents. The grandpaternal chromosome in the father
matched the grandmaternal chromosome in the mother
in 12 segments covering 130 cM (3.7% of the autosomal
genome); the grandmaternal chromosome in the father
matched the grandpaternal chromosome in the mother
in 5 segments covering 82 cM (2.4% of the autosomal
genome); the grandpaternal chromosomes in the father
and mother were identical in 9 chromosomal segments
covering 87 cM (2.5% of the autosomal genome); and
the grandmaternal chromosomes in the father and
mother were identical in 6 segments covering 65 cM
(1.9% of the autosomal genome).

The 14 progeny in family 102 showed 4–12 homo-
zygous segments with average length 18.5 cM and cov-
ering, on average, 155 cM (4.4% of the autosomal ge-
nome) (table 5). Neither parent in family 102 showed
any significant homozygous segments; genotype data on
the grandparents in this family were not available.

A 31-cM segment of chromosome 4, containing the
89 markers from D4S1604 to D4S3030, is of special
interest with regard to family 102. All 14 progeny were
homozygous for at least one portion of this segment.
Child 102-16 was homozygous for 12 markers in this
segment, spanning 7 cM; children 102-10 and 102-12
were homozygous for the entire 31-cM segment. The
other 11 progeny were homozygous for a segment of
intermediate size. An inspection of the inferred haplo-
types in the two parents shows that they share a common
haplotype on one pair of their chromosomes over this
entire segment; on their other chromosome pair, they
share a common haplotype for an 18-cM segment con-
taining 44 markers. The homozygosity in 8 of the 14
progeny derives from the longer haplotype shared by the
two parents; the homozygosity in the other 6 progeny
derives from the other shared haplotype.

Of the 100 individuals in the other six families, 20
had at least one homozygous segment with LOD score
14.67. Individual 1416-14 (a grandparent) showed four
homozygous segments, including three segments with

LOD scores 115, containing 125 markers and spanning
19–29 cM. Nine individuals were homozygous for seg-
ments !2 cM long, including one individual homozygous
for two short but significant segments. Ten other indi-
viduals were homozygous for segments 3–7 cM long.
One of these contained 27 markers; the others contained
6–11 markers. A complete list of significant homozygous
segments for all individuals in the eight CEPH families
is available with the electronic version of this article, at
the Journal’s web site.

Many of the identified homozygous segments con-
tained one or more heterozygous markers. These het-
erozygous markers could be due to errors in marker
order, genotyping errors, mutations, or gene conver-
sions. Of the 286 homozygous segments identified, 41
contained one heterozygous marker, and 8 contained
two to four heterozygous markers. The other 237 seg-



Broman and Weber: Autozygosity in CEPH Families 1497

Table 5

Homozygous Segments for Progeny in Family 102

Individual
No. of

Segments
Length
(cM)

No. of
Markers

Total
(Average)
Length

3 10 4.1–23.1 11–69 157.9 (15.8)
4 6 8.2–28.2 9–70 105.9 (17.7)
5 10 7.5–38.0 9–61 198.4 (19.8)
6 6 3.3–27.0 9–55 98.6 (16.4)
7 4 8.7–74.1 27–118 123.5 (30.9)
8 12 5.1–38.4 8–68 200.7 (16.7)
9 12 2.5–35.7 16–95 253.1 (21.0)
10 9 8.2–38.6 10–87 200.7 (22.3)
11 9 3.9–38.4 7–60 158.3 (17.6)
12 5 7.6–31.4 19–76 106.9 (21.4)
13 11 5.2–27.2 7–77 175.1 (15.9)
14 7 1.7–31.8 10–71 128.7 (18.4)
15 12 3.2–32.9 8–69 183.5 (15.3)
16 4 9.0–37.5 17–69 77.7 (19.4)

ments contained no heterozygous markers. Among the
eight segments with two or more heterozygous markers,
there were three cases in which two heterozygous mark-
ers were adjacent to each other, separated by 1 cM.
(These occurred in a segment on chromosome 13 in
progeny from family 102.) In all other cases, the het-
erozygous markers were relatively isolated from each
other.

The lengths of the homozygous segments that we iden-
tified, as well as the proportion of the genome that they
cover, are likely to be underestimates. The markers were
ordered on the basis of the 184 meioses in these families,
and so the order is generally correct only to ∼1–3 cM.
At the ends of a homozygous segment, the presence of
a few heterozygous markers, caused by mutation or in-
correct marker order, may cause an early truncation of
the identified segment. When one assumes that « =

, a heterozygous genotype subtracts 3.0 from the.001
LOD score, whereas homozygous genotypes for alleles
with frequencies .1 and .25 add 1.0 and 0.6, respectively,
to the LOD score. Thus, for a homozygous segment to
be extended past a heterozygous genotype, one must see
three to five additional homozygous genotypes. Of the
18 homozygous segments described in tables 2 and 3, 2
contained a single heterozygous genotype, 2 contained
two heterozygous genotypes, and 1 contained four het-
erozygous genotypes; in all of these cases, there were at
least five homozygous genotypes between the hetero-
zygous marker and the end of the identified segment.

It is likely that a number of autozygous segments were
missed, since the segments need to be of a sufficient size,
containing a sufficient number of markers, to allow de-
tection (see fig. 1). Although the marker data that we
considered were of unusually high density, marker den-
sity varied considerably across the genome, with some
regions containing 15 markers in !1 cM and others con-

taining gaps as long as 8 cM with no markers. Marker
heterozygosity varied considerably as well. Although the
average heterozygosity was .68, 43% of the markers had
heterozygosities <.60 or >.80.

Discussion

Autozygosity occurs when two copies of an ancestral
haplotype come together in an individual. This may be
the result of the mating of close relatives or of linkage
disequilibrium in a population, but these are just the
two extremes in a continuum of ancestral sharing. It is
customary to distinguish between linkage disequilibrium
and inbreeding: homozygous segments that are the result
of linkage disequilibrium would not ordinarily be called
autozygous. However, insofar as linkage disequilibrium
is the result of the persistence of ancestral haplotypes in
a finite population, homozygosity as the result of linkage
disequilibrium is indeed the result of the mating of (very
distantly) related individuals. Linkage disequilibrium is
a local phenomenon and thus would cause only short
homozygous segments. Long homozygous segments,
such as that on chromosome 18 in individual 884-02,
which is 40 cM in length and contains 84 markers, can-
not be explained by linkage disequilibrium.

The length of an autozygous segment reflects its age:
haplotypes are broken up by recombination at meiosis,
and so a short autozygous region is likely to be of distant
origin. (Haplotypes can also be brought together by re-
combination; if such an event were to occur twice in-
dependently, the result could be an autozygous segment
composed of portions derived from two distinct ances-
tors.) The proportion of an individual’s genome that is
autozygous, the expected value of which is the inbreed-
ing coefficient (Hartl and Clark 1997), is a measure of
the degree of relationship between his or her parents.

Family 102 was of Venezuelan origin. The long seg-
ments of homozygosity in the progeny of this family,
covering on average 4.4% of the autosomal genome,
indicates that the parents were relatively closely related.
Six of the 14 progeny had homozygous segments cov-
ering 15% of their genome. Although neither parent
showed any inbreeding, the parental haplotypes corre-
sponding to the homozygous segments on chromosome
4 imply that the homozygosity in this family is the result
of relatedness between at least two pairs of the four
grandparents. For example, the grandparents could have
been composed of two sets of first cousins, making the
parents in the family double–second cousins. If that were
true, then one would expect the progeny in the family
to be autozygous at an average of 3.1% of the autosomal
genome.

Family 884 is from the Old Order Amish in Penn-
sylvania. The progeny in this family were autozygous at
a smaller proportion of the genome (2.8%), and the



1498 Am. J. Hum. Genet. 65:1493–1500, 1999

autozygous segments were somewhat smaller, than those
in family 102. The parents and grandparents in this fam-
ily, however, also showed a large proportion of autozy-
gosity. The grandparental origins of the haplotypes
shared between the two parents indicates that each of
the four grandparents was related to each of the others.
The genealogy of this family (Egeland 1972) indicates
that the grandparents have no ancestors in common in
the three prior generations. For example, if one looks
at the five generations preceding father 884-01, he is
indicated to be the child of third cousins: one of his
father’s great-grandmothers is the sister of one of his
mother’s great-grandfathers. The parents of mother 884-
02 show a similar relationship. The child of third cousins
would be expected to be autozygous at, on average,
0.4% of the genome. Individuals 884-01 and 884-02,
however, were autozygous at 3.0% and 4.6% of their
genomes, respectively. Similarly, this part of the gene-
alogy, on its own, would imply an average of 0.7%
autozygosity in the progeny, one-fourth the observed av-
erage amount of autozygosity in the 12 progeny. Thus
the homozygosity in family 884 cannot be explained by
the inbreeding in four generations preceding the family’s
grandparents. Prior generations likely showed much
more inbreeding, so that the observed autozygosity was
the result of drawing from a limited pool of DNA five
or more generations back in time.

The presence of long homozygous segments in the
parents of family 884 make this family a less than ideal
choice as a reference for genetic maps. Future analyses
of the CEPH data should treat this family with special
care. For example, Bugge et al. (1998) used the CEPH
genetic data to calculate crossover counts for each au-
tosome. Family 884 should have been excluded from
those counts for chromosomes for which a parent was
homozygous—and, therefore, not informative—for a
large portion of the chromosome, since several double
crossovers were likely missed.

Of the 100 individuals in the other six CEPH families,
all from Utah, 1 individual, grandparent 1416-14, had
one homozygous segment 6 cM long and three segments
19–25 cM long, the four segments together covering
2.2% of the autosomal genome, indicating a close re-
lationship between her parents; 19 of the other individ-
uals in these families had quite small homozygous seg-
ments. These six families clearly do not display the sort
of inbreeding observed in families 102 and 884, but the
results cannot reasonably be ascribed to chance homo-
zygosity; rather, we conclude that the families were sub-
ject to a small amount of inbreeding, at least in the past,
and that the markers studied show considerable linkage
disequilibrium, a modern trace of older relationships in
the population from which these families were drawn.

The long homozygous segments provide opportunity
for study of mutation. Heterozygous markers within

these homozygous regions are likely to be the result of
a mutation event in one of the generations between the
last common ancestor and the autozygous individual.
An example is marker GATA44 within the long ho-
mozygous region on chromosome 6 in 884-02 (genotype
152,148). These heterozygous markers will provide val-
uable information about mutation rate—and, when
combined with data on the relevant haplotype or allele
frequencies, about the nature of the nucleotide change.

The presence of long homozygous regions has at least
two implications for gene mapping in genetically isolated
populations. On the positive side, it may be possible to
extend the widely used homozygosity (autozygosity)
mapping for rare recessive disorders (Smith 1953;
Lander and Botstein 1987), to include genetically more
complex disorders. The presence of two copies of a rel-
atively common allele often increases disease risk com-
pared with the presence of a single copy (e.g., see Farrer
et al. 1997; Rosendaal 1997). By searching for shared
homozygous regions among individuals from an isolated
population with extreme phenotypes, it may be possible
to map responsible genes. Furthermore, a modification
of the LOD score described here could be used to identify
haplotypes shared by distantly related diseased individ-
uals. In this case, one would search for segments of many
contiguous markers for which the pair shared at least
one allele identical by state.

On the negative side, if a large sibship contributes
substantially to the power of a gene-mapping study and
one (or both) of the parents is homozygous at a major
gene locus, then evidence in support of linkage may be
reduced below the detection threshold. In this respect,
it is interesting to note that indications of a gene, on
chromosome 18, predisposing to bipolar affective dis-
order have been reported in some studies (Berrettini et
al. 1994; Stine et al. 1995; Freimer et al. 1996; Mc-
Mahon et al. 1997)—but not in the Old Order Amish
(Pauls et al. 1995; Polymeropoulos and Schaffer 1996;
Ginns et al. 1996, 1998)—and that the mother of Amish
family 884 (a family with bipolar affective disorder) is
homozygous for a long segment of chromosome 18.

Although it is impossible to accurately extrapolate
from data on only eight families, we suspect that long
homozygous segments are common in many human pop-
ulations. The Old Order Amish are a clearly recognized
genetic isolate, but neither the Venezuelan kindred, 102,
nor the remaining six, Utah Mormon, kindreds derived
from any obvious genetic isolates. Furthermore, as de-
scribed above, the grandparents in Amish family 884
were not particularly closely related, and many human
populations are probably as isolated genetically as the
Amish. The total lengths of autozygous segments within
individuals from a specific population may provide a
useful measure of the isolation of that population.

Clinical implications of the long homozygous seg-



Broman and Weber: Autozygosity in CEPH Families 1499

ments are substantial. First-cousin matings will produce,
on average, autozygosity of 6%. Fitness and health of
the progeny of first-cousins pairs are well known to be
reduced (Morton 1978; Jaber et al. 1992; Stoll et al.
1994; Vogel and Motulsky 1997; Ober et al. 1999). Yet,
our results indicate that many individuals who are not
the offspring of obviously consanguineous matings have
degrees of autozygosity near or even exceeding that of
the offspring of first-cousin matings. Autozygosity may
play a much larger role in morbidity than previously has
been suspected.

Analogous to a more powerful telescope that makes
possible study of new (and often more ancient) astro-
nomic phenomena, the high-density whole-genome poly-
morphism scan opens to examination a whole new ge-
netics realm. The results presented in this report could
not have been obtained with low-density scans. Current
technology does not permit 0.5-cM-density scans to be
easily completed today, but many developments are un-
derway that should eventually open this new realm to
routine exploration (e.g., see Wang et al. 1998).

It is interesting to speculate what would be observed,
in terms of autozygosity, with an even-higher-densi-
ty polymorphism scan; undoubtedly, at least some
shorter—and, hence, older—homozygous regions would
be detected. The many reported examples of strong link-
age disequilibrium in humans virtually guarantee that
these short autozygous segments exist (although their
number and length are unknown). Mutation, especially
for STRPs, will obscure detection of these more ancient
segments. For efficient detection, it may become neces-
sary to turn to diallelic polymorphisms with lower mu-
tation rates. Homologous human chromosome pairs
may even prove to be a patchwork of alternating short
segments that are either largely homozygous or largely
heterozygous for low-mutation-rate polymorphisms.

Acknowledgments

Mark Neff and Andrew Paterson generously provided com-
ments for improvement of the manuscript. Janice Egeland and
colleagues generously provided their time and expertise to trace
the genealogy of family 884. This work was supported in part
by a John Wasmuth Fellowship in Genomic Analysis, National
Human Genome Research Institute grant F32-HG000198 (to
K.W.B.), and by National Heart, Lung, and Blood Institute
contract N01HV48141 for the Mammalian Genotyping Ser-
vice (to J.L.W.).

Electronic-Database Information

URLs for data in this article are as follows:

Center for Medical Genetics, Marshfield Medical Research
Foundation, http://www.marshmed.org/genetics

Genetic Location Database, http://cedar.genetics.soton.ac.uk/
public_html/ldb.html

References

Berrettini WH, Ferraro TN, Goldin LR, Weeks DE, Detera-
Wadleigh S, Nurnberger JI, Gershon ES (1994) Chromo-
some 18 DNA markers and manic-depressive illness: evi-
dence for a susceptibility gene. Proc Natl Acad Sci USA 91:
5918–5921

Broman KW, Murray JC, Sheffield VC, White RL, Weber JL
(1998) Comprehensive human genetic maps: Individual and
sex-specific variation in recombination. Am J Hum Genet
63:861–869

Broman KW, Weber JL (1998) Estimation of pairwise rela-
tionships in the presence of genotyping errors. Am J Hum
Genet 63:1563–1564

Bugge M, Collins A, Petersen MB, Fisher J, Brandt C, Hertz
JM, Tranebjærg L, et al (1998) Non-disjunction of chro-
mosome 18. Hum Mol Genet 7:661–669

Collins A, Frezal J, Teague J, Morton NE (1996) A metric map
of humans: 23,500 loci in 850 bands. Proc Natl Acad Sci
USA 93:14771–14775

Dausset J, Cann H, Cohen D, Lathrop M, Lalouel JM, White
R (1990) Centre d’Étude du Polymorphisme Humain
(CEPH): collaborative genetic mapping of the human ge-
nome. Genomics 6:575–577

Dib C, Fauré S, Fizames C, Samson D, Drouot N, Vignal A,
Millasseau P, et al (1996) A comprehensive genetic map of
the human genome based on 5,264 microsatellites. Nature
380:152–154

Egeland JA (ed) (1972) Descendents of Christian Fisher and
other Amish-Mennonite pioneer families. Johns Hopkins
Hospital, Baltimore

Farrer LA, Cupples LA, Haines JL, Hyman B, Kukull WA,
Mayeux R, Myers RH, et al (1997) Effects of age, sex, and
ethnicity on the association between apolipoprotein E ge-
notype and Alzheimer disease: a meta-analysis. JAMA 278:
1349–1356

Freimer NB, Reus VI, Escamilla MA, McInnes LA, Spesny M,
Leon P, Service SK, et al (1996) Genetic mapping using hap-
lotype, association and linkage methods suggests a locus for
severe bipolar disorder (BPI) at 18q22-q23. Nat Genet 12:
436–441

Ginns EI, Ott J, Egeland JA, Allen CR, Fann CSJ, Pauls DL,
Weissenbach J, et al (1996) A genome-wide search for chro-
mosomal loci linked to bipolar affective disorder in the Old
Order Amish. Nat Genet 12:431–435

Ginns EI, St Jean P, Philibert RA, Galdzicka M, Damschroder-
Williams P, Thiel B, Long RT, et al (1998) A genome-wide
search for chromosomal loci linked to mental health well-
ness in relatives at high risk for bipolar affective disorder
among the Old Order Amish. Proc Natl Acad Sci USA 95:
15531–15536

Hartl DL, Clark AG (1997) Principles of population genetics,
3d ed. Sinauer Associates, Sunderland, MA

Jaber L, Merlob P, Bu X, Rotter JI, Shohat M (1992) Marked
parental consanguinity as a cause for increased major mal-
formations in an Israeli Arab community. Am J Med Genet
44:1–6



1500 Am. J. Hum. Genet. 65:1493–1500, 1999

Lander ES, Botstein D (1987) Homozygosity mapping: a way
to map human recessive traits with the DNA of inbred chil-
dren. Science 236:1567–1570

McMahon FJ, Hopkins PJ, Xu J, McInnis MG, Shaw S, Car-
don L, Simpson SG, et al (1997) Linkage of bipolar affective
disorder to chromosome 18 markers in a new pedigree series.
Am J Hum Genet 61:1397–1404

Morton NE (1978) Effect of inbreeding on IQ and mental
retardation. Proc Natl Acad Sci USA 75:3906–3908

Ober C, Hyslop T, Hauck WW (1999) Inbreeding effects on
fertility in humans: evidence for reproductive compensation.
Am J Hum Genet 64:225–231

Pauls DL, Ott J, Paul SM, Allen CR, Fann CSJ, Carulli JP,
Falls KM, et al (1995) Linkage analysis of chromosome 18
markers do not identify a major susceptibility locus for bi-
polar affective disorder in the Old Order Amish. Am J Hum
Genet 57:636–643

Polymeropoulos MH, Schaffer AA (1996) Scanning the ge-
nome with 1772 microsatellite markers in search of a bipolar
disorder susceptibility gene. Mol Psychiatry 1:404–407

Rosenberg M, Hui L, Ma J, Nusbaum HC, Clark K, Robertson
L, Dziadzio L, et al (1997) Characterization of short tandem
repeats from thirty-one human telomeres. Genome Res 7:
917–923

Rosendaal FR (1997) Risk factors for venous thrombosis:
prevalence, risk, and interaction. Semin Hematol 34:
171–187

Sheffield VC, Weber JL, Buetow KH, Murray JC, Even DA,
Wiles K, Gastier JM, et al (1995) A collection of tri- and

tetranucleotide repeat markers used to generate high quality,
high resolution human genome-wide linkage maps. Hum
Mol Genet 4:1837–1844

Smith CAB (1953) Detection of linkage in human genetics. J
R Stat Soc B 15:153–192

Stine OC, Xu J, Koskela R, McMahon FJ, Gschwend M, Frid-
dle C, Clark CD, et al (1995) Evidence for linkage of bipolar
disorder to chromosome 18 with a parent-of-origin effect.
Am J Hum Genet 57:1384–1394

Stoll C, Alembik Y, Dott B, Feingold J (1994) Parental con-
sanguinity as a cause of increased incidence of birth defects
in a study of 131,760 consecutive births. Am J Med Genet
49:114–117

Sunden SLF, Businga T, Beck J, McClain A, Gastier JM, Pulido
JC, Yandava CN, et al (1996) Chromosomal assignment of
2900 tri- and tetranucleotide repeat markers using NIGMS
somatic cell hybrid panel #2. Genomics 32:15–20

Utah Marker Development Group (1995) A collection of or-
dered tetranucleotide-repeat markers from the human ge-
nome. Am J Hum Genet 57:619–628

Vogel F, Motulsky AG (1997) Human genetics: problems and
approaches, 3d ed. Springer-Verlag, Berlin

Wang DG, Fan JB, Siao CJ, Berno A, Young P, Sapolsky R,
Ghandour G, et al (1998) Large-scale identification, map-
ping, and genotyping of single-nucleotide polymorphisms in
the human genome. Science 280:1077–1082

White R, Lalouel JM (1988) Sets of linked genetic markers for
human chromosomes. Annu Rev Genet 22:259–279