Clin Genet 2000: 58: 375–385

Printed in Ireland. All rights reser6ed

Original Article

A multivariate analysis of 59 candidate genes

in personality traits: the temperament and
character inventory
Comings DE, Gade-Andavolu R, Gonzalez N, Wu S, Muhleman D, DE Comingsa,
Blake H, Mann MB, Dietz G, Saucier G, MacMurray JP. A multivari- R Gade-Andavolua,
ate analysis of 59 candidate genes in personality traits: the tempera- N Gonzaleza, S Wua,
ment and character inventory. D Muhlemana, H Blakea,
Clin Genet 2000: 58: 375 – 385. © Munksgaard, 2000 MB Manna, G Dietza,
G Saucierb and JP MacMurrayc
Cloninger (Cloninger CR. Neurogenetic adaptive mechanisms in alco- a
Department of Medical Genetics, City of
holism. Science 1987: 236: 410 – 416) proposed three basic personality Hope Medical Center, Duarte, CA;
dimensions for temperament: novelty seeking, harm avoidance, and re- b
Department of Psychology, University of
ward dependence. He suggested that novelty seeking primarily utilized Oregon, Eugene, OR; c Department of
dopamine pathways, harm avoidance utilized serotonin pathways, and Psychiatry, Loma Linda University School of
reward dependence utilized norepinephrine pathways. Subsequently, one Medicine, Loma Linda, CA, USA
additional temperament dimension (persistence) and three character di-
mensions (cooperativeness, self-directedness, and self-transcendence)
were added to form the temperament and character inventory (TCI).
We have utilized a previously described multivariate analysis technique
(Comings DE, Gade-Andavolu R, Gonzalez N et al. Comparison of
the role of dopamine, serotonin, and noradrenergic genes in ADHD, Key words: behavior – cooperativeness –
ODD and conduct disorder. Multivariate regression analysis of 20 genes – harm avoidance – multifactorial –
genes. Clin Genet 2000: 57: 178 – 196; Comings DD, Gade-Andavolu R, novelty seeking – persistence – polygenic
Gonzalez N et al. Multivariate analysis of associations of 42 genes in – psychiatric – reward dependence –
ADHD, ODD and conduct disorder. Clin Genet 2000: in press) to self-directedness – self-transcendence
examine the relative role of 59 candidate genes in the seven TCI traits Corresponding author: DE Comings, De-
and test the hypothesis that specific personality traits were associated partment of Medical Genetics, City of Hope
with specific genes. While there was some tendency for this to be true, Medical Center, Duarte, CA 91010, USA
a more important trend was the involvement of different ratios of
functionally related groups of genes, and of different genotypes of the Received 1 June 2000, revised and ac-
same genes, for different traits. cepted for publication 1 August 2000

The identification of the genes involved in the
common, complex, polygenic disorders has proven
to be difficult. Family-based screening techniques,
such as lod score or sib-pair linkage analyses,
generally lack the power to detect genes with a
small effect (1, 2). Our studies have shown that for
behavioral variables the percentage of the variance
accounted for by each gene is usually quite small at
less than 1.5% (3, 4). Although association studies
have the power to detect these small effects, it has
been suggested that genome screening using associ-
ation studies and random single nucleotide poly-
morphisms (SNPs) may require a very large
number of polymorphisms (5). These observations
suggest that the most efficient approach to study-
ing disorders that are due to the additive effect of
multiple genes is to examine the additive effect of
multiple candidate genes.
A number of studies have reported the associa-
tion of individual genes with various personality
traits. For example, the DRD4 gene has been re-
ported to be associated with novelty seeking (6–
10), the DRD2 and DAT1 genes have been
reported to be associated with novelty seeking (9)
and schizoid avoidant behavior (11), and several
serotonin genes have been reported to be associ-
ated with a range of traits (12–14). However, as is
generally typical of reports of single genes with
complex polygenic traits, there are also many nega-
tive reports (8, 15–19). We have previously sug-
gested that this is the expected outcome of studies
of polygenic inheritance, where each gene accounts

375
Comings et al.
for only a small percentage of the total variance and
where there is a great degree of genetic heterogene-
ity (20, 21). We have proposed that the most
efficient method of examining polygenic disorders
that are due to the additive effect of multiple genes
is to examine the additive effect of multiple candi-
date genes and to examine the additive effect of
functionally related groups of genes (3).
We have previously described a multivariate
analysis of associations (MAA) technique to assist
in this goal (3, 4). In the present study, we utilized
the MAA technique to examine the relative effect of
different functional groups of genes for the person-
ality traits in the temperament and character inven-
tory (TCI). Cloninger (22) initially proposed three
basic temperament personality dimensions: novelty
seeking, harm avoidance, and reward dependence.
He suggested that novelty seeking primarily utilized
dopamine pathways, harm avoidance utilized sero-
tonin pathways, and reward dependence utilized
norepinephrine pathways. Twin studies have shown
that approximately 50% of the variance of personal-
ity traits is genetic (23 – 26). Since these traits play
an important role in behavior, including substance
abuse, pathological gambling, other risk-taking be-
haviors, and behavior in general, the identification
of the genes involved will be important for under-
standing human behavior.
Because of its emphasis on examining the effect
of groups of functionally related genes, the MAA
technique was well suited to examine the relati6e
role of dopamine, serotonin, and norepinephrine
genes in these traits. To determine if these genes
actually accounted for the majority of the genetic
variance, we have also examined genes affecting the
other functional groups of genes, such as GABA,
opioid, other neurotransmitter, sex hormone, and
other hormone genes (4). The study was carried out
in a series of male subjects consisting of 81 college
students and 123 individuals with a range of sub-
stance abuse disorders. The mean age of the student
group was 32.9 years (SD 7.3) and of the substance
abuse group was 40.8 years (SD 7.4). All subjects
were administered the TCI (27, 28). Since both age
and substance abuse can affect the total trait scores,
the dependent variable consisted of the residual
scores after covarying for age and diagnostic group.
We have tested the following three hypotheses: 1)
The Cloninger hypothesis (22). This proposes that
genes affecting dopamine, serotonin, and nore-
pinephrine are predominately and respectively in-
volved in novelty seeking, harm avoidance, and
reward dependence. For the sake of the present
study, we will assume predominant means 51% or
more of the genetic variance or approximately 25%
or more of the total variance. 2) A modified
376
Cloninger hypothesis. As mentioned above, we
have found that it is rare for a single gene to account
for more than 1.5% of the variance of a given
behavioral trait. Since there are a limited number of
specific dopamine, serotonin, and norepinephrine
genes, this observation might make it unlikely that
a single neurotransmitter pathway would account
for more than 25% of the variance of any personal-
ity trait. Thus, a modified Cloninger hypothesis is
that novelty seeking would have the highest opti-
mized total variance for the dopamine genes, re-
ward dependence would have the highest for the
serotonin genes, and harm avoidance would have
the highest for norepinephrine genes, but none of
the totals would exceed 25%. 3) A general polygenic
hypothesis. The final hypothesis tested is that for
each of the three traits other groups of genes will
play as much or more a role than dopamine,
serotonin, or norepinephrine genes. In addition, the
use of different genotypes of the same genes may be
just as important as the use of different genes (29).
Methods
Subjects
We collected DNA samples and performed exten-
sive personality testing on a series of students from
a local university (California State University at
San Bernardino) and subjects from a veterans ad-
ministration hospital addiction treatment unit
(ATU). To avoid gender as a confounding variable
and to allow the assessment of X-linked genes
without the complication of heterozygotes in only
some subjects, we have examined only males in both
groups. To avoid the confounding variable of race,
the study was restricted to non-Hispanic Cau-
casians. This produced a final group of 81 students
and 123 ATU subjects, 204 in all. The mean age of
the control group was 32.9 years (SD 7.3) and of the
ATU group was 40.8 years (SD 7.4). All subjects
were administered the TCI (27). The Levene test for
homogeneity of variance for comparing the stu-
dents and the ATU subjects was non-significant for
all but cooperativeness. The SPSS statistical pack-
age was used (SPSS, Inc, Chicago, IL) for the
ANOVA, test of homogeneity, and multivariate
analyses. The study and the consent forms were
approved by the Internal Review Boards of the City
of Hope and Loma Linda University.
Test instrument
After obtaining written informed consent, a blood
sample was obtained for genetic studies, and the
subjects were administered the TCI (27). This con-

sists of seven main personality traits, some of
which were divided into sub-traits. In the present
study, we have examined only the major traits.
Genes and polymorphisms
The majority of the genes and polymorphisms ex-
amined have been reported previously (3, 4). The
following are the additional genes used in this
study and references for the polymorphisms used.
For SNPs, 11 = 1, 12 =2, and 22 =3. The coding
is given for the repeat polymorphisms. The gene
symbols are listed at www.gdb.org. Those that are
X-linked are so indicated. More than one polymor-
phism was examined for some genes (DRD2, HTT,
TDO2, HTR2A, and AR). Dopamine genes: DRD2
ins/del (30). Serotonin genes: HTT (SLC6A4)
VNTR (31); HTR2C (X-linked) Hinf I (32); TDO2
G/T Dpn II (33). Norepinephrine genes: ADRB1
Eco 0109 I (34); MAOA (X-linked) Fnu I (35).
GABA genes: GABRA1 dinucleotide repeat (36)
B185/ B185=1, het= 2, ] 185/ ] 185=3;
GABRA3 (X-linked) dinucleotide repeat (37)
B168/ B168=1, het= 2, ] 168/ ] 168=3;
GABRA5 dinucleotide repeat (38) 5 284/ 5 284 =
1, het 2, \ 284/ \284= 3; GABRA6 dinucleotide
repeat (39) B202/ B202= 1, het = 2, ]202/
]202 =3; GABBR dinucleotide repeat (40) B214/
B241=1, het=2, ]141/ ] 241= 3; GAD2 dinu-
cleotide repeat (41) B173/ B173= 1, het =2,
] 173/ ]173= 3. Opioid genes: OPRMI Drd I
(42); PENK dinucleotide repeat (43) B 78/ B78 =
1, het=2, ]78/ ] 78. Other neurotransmitter
genes: no new genes over prior references. Other
genes: APOE Hba I (44); CD4 pentanucleotide
repeat (45), 593/ 593= 1, het= 2, ] 95/ ]95 =
3. Sex hormone genes: AR trinucleotide GGC re-
peat (46); AR trinucleotide CAG repeat (47). Other
hormone genes: LEP (48) B203/ B 203=1, het =
2, ] 203/ ]203=3; MC3R dinucleotide repeat
(49) B33 repeat/ B 33= 1, het = 2, ]33/ ]33 =
3; OXY (50, 51). This represents 63 polymorphisms
and 59 genes. We find that as long as two polymor-
phisms at a single locus are not in significant
linkage disequilibrium, the results with two poly-
morphisms will be evaluated as though they were
separate genes, even though they are listed for a
single gene. In general, only one of the two was
included in the regression equation.
Results
The MAA technique provides six pieces of infor-
mation about the genes involved in a complex
trait: 1) which genes are included in the equation;
2) the fraction of the variance attributed to each
Multiple candidate genes
included gene; 3) the level of significance for each
included gene; 4) the relative involvement of the
different genotypes of each involved gene for each
trait; 5) the sum of the variance of functional
groups of genes for each trait; 6) the ratio of the
variance for each gene group to all the other gene
groups for each trait. The first four pieces of infor-
mation are shown in Fig. 2A,B, the fifth is shown
in Fig. 3, and the sixth is shown in Fig. 4. The
method of diagrammatically representing the gene
codes is shown in Fig. 1. As discussed previously
(3), for autosomal genes, there are 12 modes of
inheritance. These include two modes each for
codominant, dominant, and recessive inheritance
and six modes for heterosis. There are two modes
for X-linked genes in males. As an example of the
diagrammatic representation of these modes, as
shown in Fig. 1 for codominant inheritance, those
carrying the 11 genotype may show the lowest
phenotypic score (open circle set at the bottom),
those carrying the 12 genotype may show an inter-
mediate phenotypic score (gray circle set in the
middle), and those carrying the 22 genotype may
show the highest phenotype score (black circle set
at the top). Such a gene score would be 012. A 210
score would show the reverse.
Individual genes
The results for the temperament traits are shown in
Fig. 2A. For novelty seeking, three genes, TPH,
GABRB1, and OXYR, were significant at p 50.01,
and 11 were significant at p 50.05. For harm
avoidance, 5 genes, TPH, ADRA2A, GABRB3,
CNRA4, and ADOR2A, were significant at p5
0.01, and 12 were significant at p 50.05. For re-
ward dependence, four genes, DRD4, PNMT,
APOE, and CYP19, were significant at p50.1,
and 11 were significant at p 50.05. For persis-
Fig. 1. Diagrammatic representation of the gene codes repre-
senting the 12 modes of inheritance for autosomal dominant
genes and the two modes for X-linked genes in males. See text
under Results for details.
377
Comings et al.

tence, only one gene, HTR2A, was significant at PNMT, GABRA6, NMDAR1, CNRA4, CYP19,
p 50.01, and eight were significant at p 50.05. SBP, and OXYR, were significant at p50.01, and
The results for the character traits are shown in 17 were significant at p50.05. For self-directed-
Fig. 2B. For cooperativeness, eight genes, DAT1, ness, two genes, CNRA4 and ADOR2A, were sig-

Fig. 2. These diagrams show the results for 59 genes in nine groups with the temperament traits shown in Fig. 2A and the character
traits in Fig. 2B. The names of the individual genes are shown on the abscissa, separated into nine groups. The height of the bars
corresponds to the r2 values for the gene. The diagram for the gene coding is at the top of each bar. The asterisks show the level
of significance for each gene. If there is no asterisk, the presence of a bar indicates the gene was included in the equation at p 50.10.

378
 Multiple candidate genes

Fig. 3. Percentage of the total variance for all genes accounted for by the total variance for each functional group of genes, i.e.,
dopamine, serotonin, norepinephrine, GABA, opioid, other neurotransmitters, other, sex hormone, and other hormone genes. This
shows that for reward dependence the total variance for the norepinephrine genes was greater than for any other group of genes.
It also illustrates that for novelty seeking the total variance for serotonin, GABA, and other hormone genes was greater than for
dopamine genes. Finally, it shows that for harm avoidance all the contribution of the dopamine, serotonin, and norepinephrine
genes is approximately comparable.

nificant at p50.01, and five were significant at
p 50.05. Finally, for self-transcendence, four
genes, DRD4, GABRA1, CD4, and AR, were sig-
nificant at p50.01, and 16 were significant at
p50.05. Thus, for the seven traits, the number of
genes significant at p5 0.01 ranged from one to
seven, and the number significant at p 50.05
ranged from eight to 13. This is consistent with a
modest, but certainly not dramatic, difference in
the number of significant genes for the seven traits.
Of the genes significant a p5 0.01, seven genes,
DRD4, TPH, PNMT, CNRA4, ADOR2A, CYP19,
and OXYR, were involved in more than one trait.
These repeated genes were evenly distributed
among the different functional groups of genes.
r2 for individual genes
As shown in Fig. 2A,B, even among those genes
that were significant at p 50.001, they rarely ac-
counted for more than 5% of the total variance of
a given trait. On average, the significant associa-
tions accounted for only 1.5 – 3.5% of the variance.
Total r2 for functional groups of genes
By contrast to individual genes shown in Fig.
2A,B, Fig. 3 shows the relative contribution of
each of the functional groups of genes. This is
expressed as the percentage of the total r2 for all of
the genes for a given trait divided by the total r2
for a given functional group of genes. Thus, the
total percentage for the nine functional groups of
genes is 100. This showed that the total r2 for a
functional group commonly accounted for 15–35%
of the total explained variance and averaged 5–
10%. Self-transcendence had the highest total vari-
ance for dopamine genes, persistence had the
highest for serotonin genes, reward dependence
had the highest for NE genes, novelty seeking and
self-transcendence had the highest for GABA
genes, reward dependence had the highest for opi-
oid genes, self-directedness had the highest for
other neurotransmitter genes, reward dependence
and self-transcendence had the highest for other
genes, cooperativeness had the highest for sex hor-
mone genes, and reward dependence had the
highest for other hormone genes. However, the
characterization of the different traits is far more
complex than simply identifying which traits show
the highest total variance for different functional
sets of genes.
Ratio of total r2 for different functional groups of genes
Fig. 4 expands on Fig. 3 by showing the ratio of
the total variance for each group of genes com-
pared to the total variance for each of the remain-
ing groups of genes for each trait. Ratios that are
infinite have been left out. Thus, instead of just
showing the total variance for each functional
group of genes by itself (as in Fig. 3). This shows
the magnitude of the total variance in relation to
each of the other groups of genes; the higher the
379
Comings et al.

Fig. 4. The ratios of the total variance for each of the nine different functional groups of genes to the total variance for the
remaining groups of gene. Among other things, this shows that the ratio for the dopamine gene compared to the other gene groups
tended to be highest for novelty seeking, persistence, and self-transcendence. The ratio for the serotonin genes compared to the
other gene groups was highest for novelty seeking, persistence, and self-directedness. The ratio for the norepinephrine gene
compared to the other groups of genes was highest for reward dependence, persistence, and cooperativeness.

380

bars, the greater the ratio. As an example, for
persistence, serotonin genes accounted for 34.5%
of the total explained variance of all the genes. The
ratio of serotonin to serotonin is 1. Thus, the gray
bar (serotonin) for the serotonin panel for persis-
tence (Fig. 4) shows the ratio is 1. For persistence,
the other NT genes accounted for 2.7% of the total
explained variance for all the genes. The ratio of
other NT to serotonin was 12.6 (Fig. 4). Thus, for
persistence, serotonin accounted for 12.6 times
more of the variance than other NT genes.
To examine the Cloninger hypothesis, we can
restrict this type of analysis to just dopamine,
serotonin, and norepinephrine genes for novelty
seeking, harm avoidance, and reward dependence.
Doing this shows that the ratio of the dopamine
genes to the other groups of genes was highest for
novelty seeking. The ratio of the serotonin genes to
the other groups of genes was also highest for
novelty seeking. By contrast, the ratio of the nore-
pinephrine genes to the other groups of genes was
highest for reward dependence. When all of the
groups of genes and all of the traits are examined,
other features are of interest. For example, the
ratio of the serotonin genes to the other groups of
genes was highest for novelty seeking and persis-
tence, the ratio of the GABA genes was highest for
novelty seeking, the ratio of the other NT genes
was highest for self-directedness, the ratio of the
other genes was highest for reward dependence and
self-directedness, the ratio for the sex hormone
genes was highest for cooperativeness, and the
ratio for other hormone genes was highest for
novelty seeking.
Total r2 for all genes
Although it is a maximized estimate, the total r2
accounted for by the 59 genes examined is of some
interest. This varied over a relatively narrow range
from 32.5% of the variance for harm avoidance to
41.3% for reward dependence, with p values rang-
ing from 1.9 ×10 − 4 to 5.5 ×10 − 6. Specifically, the
total r2 was 0.38 for novelty seeking, 0.32 for harm
avoidance, 0.41 for reward dependence, and 0.32
for persistence. The average total r2 for the four
temperament traits was 0.36. The total r2 was 0.39
for cooperativeness, 0.33 for self-determination,
and 0.39 for self-transcendence. The average for
the three character traits was 0.37.
Genotype variation
The final important characterization of polygenic
disorders was the variation in the genotypes uti-
lized for different traits (Fig. 2A,B). As an exam-
Multiple candidate genes
ple, the TPH gene was scored 201 in novelty
seeking and harm avoidance, but scored 012 in
persistence. The DAT1 gene was scored 012 in
self-transcendence, but scored 210 in self-directed-
ness. Many other examples of the same genes using
different genotypes for different traits can be seen.
Discussion
Complex polygenic disorders have been particu-
larly recalcitrant to analysis using the family-based
techniques commonly employed for single gene
disorders. For example, in a recent study of autism
using a large number of sib pairs, Risch et al. (52)
concluded that ‘alternative methods of analysis are
required’. Since polygenic disorders are due to the
effect of multiple genes, we propose that the most
parsimonious approach is to examine the effect of
multiple candidate genes using association studies.
Since each gene contributes to a small percentage
of the total variance, and since genetic heterogene-
ity is so great, it may also be more productive to
examine the relative role of functional groups of
genes than to hold individual genes to a standard
of replication they can never attain. All the advan-
tages of identifying individual genes – understand-
ing the cause, pathophysiology, and most
important, treatment – can be attained using the
MAA technique. For example, if a given behav-
ioral disorder is found to be predominately due to
a combination of serotonin and NE genes, treat-
ment can focus on drugs that modify both of these
neurotransmitter systems.
Test of the Cloninger hypothesis
The present results partially support the un-
modified Cloninger hypothesis in that over 25% of
the total explained variance of reward dependence
was due to norepinephrine genes, and this was the
most prominent of the nine groups of genes exam-
ined (Fig. 3). The unmodified Cloninger hypothesis
was not supported for novelty seeking or for harm
avoidance. For novelty seeking, the serotonin
genes contributed to a greater percentage of the
variance (22%) than the dopamine genes (12.5%),
and for harm avoidance, all three groups of genes
were about equally represented. The modified
Cloninger hypothesis was partially met in that the
norepinephrine genes accounted for the largest per-
centage of the variance for reward dependence,
and when restricted to only examining novelty
seeking, harm avoidance, and reward dependence,
the ratio for the dopamine genes to other groups of
genes was highest for novelty seeking (Fig. 3).
When other sets of genes were included, the
381
Comings et al.
modified Cloninger hypothesis was not supported
for novelty seeking, since the other hormone genes
accounted for the highest percentage of the vari-
ance, and the serotonin genes accounted for a
greater percentage of the variance (22%) than the
dopamine genes (12.5%). The modified Cloninger
hypothesis was also not supported for harm avoid-
ance, since the total variance for the serotonin
genes was less than for the norepinephrine and
other hormone genes.
The third hypothesis, that each of the nine
groups of genes all play a role in most of the
personality traits and that some genotypes may be
positively correlated with one trait, but negatively
correlated with a different trait, came closest to
explaining the data. Thus, with the exception of an
absence of a contribution of the opioid genes in
harm avoidance, persistence, and self-transcen-
dence, and an absence of a contribution of other
NT genes in self-transcendence, each of the func-
tional groups of genes played a role in each of the
seven traits. We anticipate that the failure of the
opioid and other NT genes to contribute to some
of the traits may disappear as more genes in these
groups are studied. In addition, as shown in Fig.
2A,B, the gene coding for specific genes varied
among the different traits, indicating this was al-
most as important as which genes were involved.
Other personality traits
While we have emphasized novelty seeking, harm
avoidance, and reward dependence in testing of the
Cloninger hypothesis, the results for the other per-
sonality traits were of interest. We were intrigued
by the considerable role of dopamine genes, and
the DRD4 gene in particular, in spiritual transcen-
dence. Dopamine receptors, including the D4 re-
ceptor (53), play an important role in the function
of the prefrontal cortex. Spirituality may especially
utilize the prefrontal cortex and thus predomi-
nantly utilize dopaminergic systems.
It is of note that the cooperativeness score tends
to be much higher in women than in men (27). In
this regard, the finding that three different sex
hormone genes were significant and that the total
variance attributed to the sex hormone genes was
higher for cooperativeness than any other of the
seven traits is of interest. Also, Cloninger (28)
suggested that serotonin genes would be involved
in persistence. We found that this was the case.
Three serotonin genes, the serotonin transporter,
HTR2A, and TPH were all significant (Fig. 2A).
For persistence, the total r2 was highest for sero-
tonin genes (Fig. 3), and the ratio of other groups
382
of genes to serotonin genes was highest for sero-
tonin genes (Fig. 4).
Relative genetic influence on temperament versus
character traits
Cloninger (28) has suggested that the character
traits may have more of an environmental influ-
ence and less of a genetic influence than the tem-
perament traits. However, on the basis of twin
studies, Kirk et al. (54) have estimated that the
heritability of self-transcendence is 0.37 in males.
While the above results for the total r2 values
suggest a comparable genetic contribution for all
of the TCI traits, studies with additional genes are
necessary before finalizing such conclusions.
Future directions
The present study is designed to illustrate some of
the potential advantages of simultaneously exam-
ining the effect of multiple genes on a given pheno-
type. We view this as only an initial attempt to
examine multiple genes simultaneously. The fol-
lowing are a number of potential future directions
and issues that require further study.
More cases. When small effects are involved, as is
the case for each gene in polygenic inheritance, an
increase in the number of cases studied helps to
decrease errors due to random effects. Thus, the
single most desirable future direction using the
MAA technique for behavioral or other variables
would be to increase the N or merge results from
different investigators using the same instrument
and the same polymorphisms. Alternatively, the
results from different studies could be compared.
A priori considerations. The power of a result is
usually considered to be greater if there is an a
priori reason to believe a specific gene may be
involved for a specific phenotype. However, given
the enormous complexity in neuronal interactions
in the central nervous system, virtually any gene
affecting the function of any neurotransmitter,
neuropeptide, hormone, or secondary messenger
system could be involved. This is the value of a
technique that screens the relative effectiveness of a
number of candidate genes.
Examination of more genes. The addition of more
genes has the advantage of identifying genes or
groups of genes that previously had not been con-
sidered as playing a role in a given phenotype.
Thus, the role of hormone genes and other neuro-
transmitter genes in several of the traits, as shown

in Fig. 2A,B, probably would have been missed
using a purely a priori approach. The choice of the
genes we used was partially dependent upon which
polymorphisms were available for the respective
candidate genes. In this study, many important
candidate genes were missing. However, in a rela-
tively short period of time the number of candidate
gene polymorphisms should at least double. With
the identification of new SNPs, a much larger
number of genes can be included in future studies.
While the inclusion of an additional 60 – 100 candi-
date genes could account for a significant propor-
tion of the genetic variance of these and other
behavioral traits, adding more genes also increases
the need for adding more cases and for replication
studies to separate those genes or groups of genes
that are real from those that are random effects.
Examination of more polymorphisms per gene. The
rapid increase in the identification of SNPs will also
allow the use of more than one polymorphism per
candidate gene. This would allow a maximization of
the true role of each gene in different phenotypes.
To further maximize the role of each gene, future
studies could examine approaches to testing the role
of individual alleles in the short tandem repeat
polymorphisms, rather than grouping them into
two allele groups by size, as was done here.
Gene coding. As discussed above, performing the
gene coding on the same set of subjects as the
analysis allows the relative effect of each gene or
functional group of genes to be examined, but
inflates the total variance for all the genes and
increases the chances of random errors. The best
way to avoid this would be to use a separate portion
of the sample for coding the genes. This will require
future studies with a much larger sample size.
Although the use of multivariate regression analysis
eliminates the need for a Bonferroni correction for
the number of genes examined, this would be more
rigorous if the gene coding was performed on a
separate sample. In a previous study, we utilized the
MAA technique to examine the role of 20 do-
pamine, serotonin, and norepinephrine genes in
ADHD and used gene coding based both on a
different sample and the same sample (3). While the
results were generally the same for both, more genes
were included in the regression analysis with same-
sample scoring compared to different-sample scor-
ing. This may be due either to the great genetic
heterogeneity of polygenic disorders or to random
effects, or both.
Replication. Because of the genetic heterogeneity of
polygenic disorders, it is anticipated that attempts
Multiple candidate genes
at replication will show more variation for individ-
ual genes than for the ratios of the total variance for
functionally related groups of genes. This is based
on the assumption that to a certain degree, different
genes within a functional group may be substituted
for each other. We suspect that the small percentage
of the variance attributable to each gene and genetic
heterogeneity, rather than hidden ethnic stratifica-
tion, is the major reason why studies of the role of
single genes in polygenic behavioral disorders are so
variable and difficult to replicate.
Implications for treatment. This approach can not
only provide important insight into the genetic
etiology of behavioral disorders, but can also
provide clues to treatment. For example, Cloninger
et al. (27) have shown that self-directedness and
cooperativeness are more strongly negatively corre-
lated with a range of personality disorders than any
of the other TCI traits. The results shown in Fig. 2B
show that both of these traits utilized ‘other neuro-
transmitters’ more than any of the other traits.
Personality traits are classically difficult to alter
using standard psychotropic medications. These
results suggest that medications that target one or
more of these other neurotransmitters might
provide effective treatment for one or more person-
ality disorders.
In summary, we believe that the MAA technique
may provide a useful new approach for studying the
genetics of complex disorders. It meets the dictum
that the best way to study disorders that are due to
the combined effect of multiple genes is to examine
the combined effect of multiple candidate genes.
Acknowledgements
Supported in part by the National Institutes of Drug
Abuse grant RO1-DA08417, the Tobacco Related Re-
search Disease Program grant 4RT-0110, and the Gaming
Entertainment Research and Education Foundation & Na-
tional Center for Responsible Gaming.
References
1. Risch N, Merikangas K. The future of genetic studies of
complex human diseases. Science 1996: 273: 1516 – 1517.
2. Risch NJ. Searching for genetic determinants in the new
millennium. Nature 2000: 405 (6788): 847 – 856.
3. Comings DE, Gade-Andavolu R, Gonzalez N et al. Com-
parison of the role of dopamine, serotonin, and noradren-
ergic genes in ADHD, ODD and conduct disorder.
Multivariate regression analysis of 20 genes. Clin Genet
2000: 57: 178 — 196.
4. Comings DD, Gade-Andavolu R, Gonzalez N et al. Multi-
variate analysis of associations of 42 genes in ADHD,
ODD and conduct disorder. Clin Genet 58: 31 – 34.
5. Kruglyak L. Prospects for whole-genome linkage disequi-
librium mapping of common disease genes. Nat Genet
1999: 22: 139 – 144.
383
Comings et al.
6. Benjamin J, Li L, Patterson C et al. Population and
familial association between the D4 dopamine receptor
gene and measures of novelty seeking. Nat Genet 1996: 12:
81–84.
7. Epstein RP, Novick O, Umansky R et al. Dopamine D4
receptor (D4DR) exon III polymorphism associated with
the human personality trait of novelty seeking. Nat Genet
1996: 12: 78–80.
8. Ebstein RP, Belmaker RH. Saga of an adventure gene:
novelty seeking, substance abuse and the dopamine D4
receptor (DRD4 ) exon III repeat polymorphism. Mol Psy-
chiatry 1997: 2: 381–384.
9. Noble EP, Ozkaragoz TZ, Ritchie TL, Zhang X, Belin TR,
Sparkes RS. D2 and D4 dopamine receptor polymor-
phisms and personality. Am J Med Genet 1998: 81: 257 –
267.
10. Ono Y, Manki H, Yoshimura K et al. Association between
dopamine D4 receptor (D4DR) exon polymorphism and
novelty seeking in Japanese subjects. Am J Med Gen
(Neuropsych Genet) 1997: 74: 501–503.
11. Blum K, Braverman ER, Cull JG et al. Association of
polymorphisms of dopamine D2 receptor (DRD2) and
dopamine transporter (DAT1) genes with schizoid/
avoidant behaviors (SAB) and pathological violence (PV).
Biol Psychiatry 1997: 42: 296S.
12. Epstein RP, Segman R, Benjamin J, Osher Y, Nemanov L,
Belmaker RH. 5-HT2C (HTR2C) serotonin receptor gene
polymorphism associated with the human personality trait
of reward dependence: interaction with dopamine D4 re-
ceptor (D4DR) and dopamine D3 receptor (D3DR) poly-
morphisms. Am J Med Gen (Neuropsych Genet) 1997: 74:
65–72.
13. Manuck SB, Flory JD, Ferrell RE, Dent KM, Mann JJ,
Muldoon M. Aggression and anger-related traits associ-
ated with a polymorphism of the tryptophan hydroxylase
gene. Biol Psychiatry 1999: 45: 603–614.
14. Mazzanti C, Lappalainen J, Naukkarinen H et al. Sib-pair
linkage and association analysis of the serotonin trans-
porter promoter polymorphism and TPQ personality
traits. Am J Med Gen (Neuropsych Genet) 1998: 81:
487–488.
15. Ebstein RP, Gritsenko I, Nemarov L, Frisch A, Oshler Y,
Belmaker RH. No association between the serotonin trans-
porter gene regulatory region polymorphism and the tridi-
mensional personality questionnaire (TPQ) temperament
of harm avoidance. Mol Psychiatry 1997: 2: 224–226.
16. Jönsson EG, Nöthen MM, Gustavsson JP et al. Lack of
evidence for alleleic association between personality traits
and the dopamine D4 receptor gene polymorphisms. Am J
Psychiatry 1997: 154: 697–699.
17. Gebhart C, Fiireder T, Fuchs K et al. No association of
novelty seeking with DRD3 and DRD2 receptor gene
polymorphisms. Am J Med Gen (Neuropsych Genet) 1998:
81: 497–498.
18. Pogue-Geile M, Ferrell R, Deka R, Debski T, Manuck S.
Human novelty seeking personality traits and dopamine
D4 receptor polymorphisms: a twin and genetic association
study. Am J Med Gen (Neuropsych Genet) 1998: 81:
44–48.
19. Sullivan PF, Fifield WJ, Kennedy MA, Mulder RT, Sell-
man JD, Joyce PR. Novelty seeking and a dopamine
transporter gene polymorphism (DAT1). Biol Psychiatry
1997: 42: 1070–1072.
20. Comings DE. Why different rules are required for poly-
genic inheritance: lessons from studies of the DRD2 gene.
Alcohol 1998: 16: 61–70.
21. Comings DE. Polygenic inheritance of psychiatric disor-
384
ders. In: Blum K, Noble EP, eds, Sparkes RS, Chen THJ,
Cull JG, associate eds. Handbook of Psychiatric Genetics.
CRC Press: Boca Raton, 1997: 235 – 268.
22. Cloninger CR. Neurogenetic adaptive mechanisms in alco-
holism. Science 1987: 236: 410 – 416.
23. Bouchard TJ, Jr. Gene, environment, and personality.
Science 1994: 264: 1700 – 1701.
24. Goldsmith HH, Losoya SH, Bradshaw DL, Compos JJ.
Genetic of personality: a twin study of the five factor
model and parental – offspring analyses. In: Martinl R,
Kohnstamm G, eds. The Developing Structure of Temper-
ament and Personality from Infancy to Adulthood. Hills-
dale, NJ: Erlbaum, 1994: 241 – 265.
25. Loehlin JC, Nichols RC. Heredity, Environment and Per-
sonality: A Study of 850 Sets of Twins. Austin: University
of Texas Press, 1976.
26. Tellegen A, Lykken DT, Bouchard TJ, Jr, Wilcox KJ,
Segal NL, Rich S. Personality similarity in twins reared
apart and together. J Pers Soc Psychol 1988: 54: 1031–
1039.
27. Cloninger CR, Svrakic DM, Przybeck TR. A psychobio-
logical model of temperament and character. Arch Gen
Psychiatry 1993: 50: 975 – 990.
28. Cloninger CR. The genetics and psychobiology of the
seven factor model of personality. In: Sild KR, ed. Biology
of Personality Disorders. American Psychiatric Press:
Washington, DC, 1998: 63 – 92.
29. Comings DE. SNPs and polygenic disorders: a less gloomy
view. Mol Psychiatry 1999: 4: 314 – 316.
30. Arinami T, Itokawa M, Enguchi H et al. Association of
dopamine D2 receptor molecular variant with schizophre-
nia. Lancet 1994: 343: 703 – 704.
31. Ogilvie AD, Battersby S, Bubb VJ et al. Polymorphism in
the serotonin transporter gene associated with susceptabil-
ity to major depression. Lancet 1996: 347: 731 – 733.
32. Lappalainen J, Virkkunen M, Dean M, Ozaki N, Linoila
M, Goldman D. Identification of a cys-ser substitution in
the 5-HT2C (htr2C) receptor gene and allelic association
to violent behavior and alcoholism. Am J Hum Genet
1994: 55: A156.
33. Comings DE, Muhleman D, Dietz G, Sherman M, Forest
G. Sequence of human tryptophan 2,3-dioxygenase: pres-
ence of a glucocorticoid response-like element composed of
a GTT repeat and an intronic CCCCT repeat. Genomics
1995: 29: 390 – 396.
34. Maqbool A, Hall AS, Ball SG, Balmforth AJ. Common
polymorphisms of b1-adrenoceptor: identification and
rapid screening assay. Lancet 1999: 353: 897.
35. Hinds HL, Hendricks RW, Craig IW, Chen ZY. Charac-
terization of a highly polymorphic region near the first
exon of the human MAOA gene containing a GT dinucle-
otide and a novel VNTR motif. Genomics 1992: 13: 896–
897.
36. Hicks AA, Kamphuis W, Barnard EA, Darlison MG. The
possible role of GABAA receptors in neurological and
psychiatric disorders. Psychiat Genet 1991: 2: 78 – 79.
37. Hicks AA, Johnson KJ, Barnard EA, Darlison MG. Dinu-
cleotide repeat polymorphism in human X-linked GABA-
A receptor alpha3-subunit gene. Nucleic Acids Res 1992:
19: 4016.
38. Glatt KA, Sinett D, Lalande M. Dinucleotide repeat poly-
morphism located in the Angelman syndrome region of
chromosome 15 at the GABAA receptor a5 locus. Am J
Hum Genet 1992: 51: A363.
39. Hicks AA. Further evidence for clustering of human
GABAA receptor subunit genes: localization of alpha-6-
subunit gene (GABRA6) to distal chromosome 5q linkage
analysis. Genomics 1994: 20: 285 – 286.

40. White JH, Wise A, Main MJ et al. Heterodimerization for
the formatin of a functional GABAB receptor. Nature
1998: 396: 679–682.
41. Wapelhorst B, Bell GI, Risch N, Spielman RS, Concannon
P. Linkage and association studies in insulin-dependent
diabetes with a new dinucleotide repeat polymorphism at
the GAD65 locus. Autoimmunity 1995: 21: 127–130.
42. Schmidt LG, Sander T, Gscheidel N et al. Possible associ-
ation of the common Asn40Asp substitution polymor-
phism of the human u-opioid receptor gene with novelty
seeking in alcohol-dependent subjects. Am J Med Gen
(Neuropsych Genet) 1998: 81: 507.
43. Weber JL, May PE. Dinucleotide repeat polymorphism at
the PENK locus. Nucleic Acids Res 1990: 18: 2000.
44. Hixson JE, Vernier DT. Restriction isotyping of human
APOE by gene amplification and clevage with Hha I. J
Lipid Res 1990: 31: 545–548.
45. Edwards MC, Clemens PR, Tristan M, Pizzulti A, Gibbs
RA. Pentanucleotide repeat length polymorphism at the
human CD4 locus. Nucleic Acids Res 1991: 19: 4791.
46. Sleddens HF, Oostra BA, Brinkman AO, Trapman J.
Trinucleotide (GGN) repeat polymorphism in the human
androgen receptor (AR) gene. Hum Molec Genet 1993: 2:
493.
47. Sleddens HF, Oostra BA, Brinkman AO, Trapman J.
Trinucleotide repeat polymorphism in the androgen recep-
tor gene (AR). Nucleic Acids Res 1992: 20: 1427.
Multiple candidate genes
48. Green ED, Maffei M, Braden VV et al. The human obese
(OB) gene: RNA expression pattern and mapping on the
physical, cytogenetic, and genetic maps of chromosome 7.
Genome Res 1995: 5: 5 – 12.
49. Yamada Y, Xiang K-S, Bell GI, Seino S, Nishi M. Dinu-
cleotide repeat polymorphism in a gene on chromosome 20
encoding a G-protein coupled receptor (D20S32e). Nucleic
Acids Res 1991: 19: 2519.
50. Sausville E, Camey D, Battey J. The human vasopressin
gene is linked to the oxytocin gene and is selectively
expressed in a cultured lungs cancer cell line. J Biol Chem
1985: 260: 10236 – 10241.
51. Repaske DR, Phillips JA, Kirby LT, Tze WJ, D’Ercole AJ,
Battey J. Molecular analysis of autosomal dominant neu-
rohypophyseal diabetes insipidis. J Clin Endocrinol Metab
1990: 70: 752 – 757.
52. Risch N, Spiker D, Lotspeich L et al. A genomic screen of
autism: evidence for a multilocus etiology. 1999: 65: 495–
507.
53. Ariano MA, Wang J, Noblett KL, Larsen ER, Sibley DR.
Cellular distribution of the D4 dopamine receptor protein
in the CNS using anti-receptor sera. Brain Res 1997: 752:
26 – 34.
54. Kirk KM, Eaves LJ, Martin NG. Self-transcendence as a
measure of spirituality in a sample of older Australian
twins. Twin Res 1999: 2: 81 – 87.
385