2001 Oxford University Press Human Molecular Genetics, 2001, Vol. 10, No.

20 22612267
Pharmacogenetics
Allen D. Roses*
GlaxoSmithKline, Five Moore Drive 5616, Research Triangle Park, NC 27709, USA
Received July 11, 2001; Accepted July 26, 2001
Pharmacogenetics is the variability of drug response due to inherited characteristics in individuals. Drug
metabolizing enzymes have been studied for decades, first as chemical reactions and, more recently, as
specific polymorphisms of known molecules. With the availability of whole-genome single-nucleotide poly-
morphism (SNP) maps, it will soon be possible to create an SNP profile for patients who experience adverse
events (AEs) or who respond clinically to the medicine (efficacy). Proof-of-principle experiments have demon-
strated that high density SNP maps in chromosomal regions of genetic linkage facilitate the identification of
susceptibility disease genes. Whole-genome SNP mapping analyses aimed at determining linkage disequilibrium
(LD) profiles along an ordered human genome backbone are in progress. SNP fingerprints or SNP PRINTs
sm
will
be used to identify patients at greater risk of an AE, or those patients with a greater chance of responding to a
medicine. As LD maps for various ethnic populations are constructed, the number of SNPs necessary to measure
for an individual will decrease. Standardized pharmacogenetic maps for drug registration and post-marketing
surveillance will result in safer, more effective and more cost-efficient medicines. The timing of these pharmaco-
genetic applications will occur over the next 5 years. In contrast, the benefits of pharmacogenomic applications
such as the identification of new tractable targets will not be visible as new medicines for 712 years, due to
the lengthy drug development and registration processes.
Pharmacogenetics and pharmacogenomics are frequently
interchanged terms and can therefore be confused. For the
purpose of clarity in the use of the terms in this review,
pharmacogenetics is defined as the study of variability in drug
responses attributed to hereditary factors in different populations.
Pharmacogenomics is the determination and analysis of the
genome (DNA) and its products (RNA and proteins) as they
relate to drug response. For example, gene expression profiling
using various microarray technologies has enabled the demon-
stration of distinct sub-sets of genes that may be expressed
differentially in disease and healthy tissues. These genomic
techniques can be useful for differential diagnosis of patients,
particularly for heterogeneous diseases that present with
similar clinical phenotypes but differ in molecular expression.
Response to treatment can sometimes be recognized at the
genomic level by tissue gene expression profiles. Expression
profiles, however, differ from the approach of using inherited
differences in our genetic information to predict responses to
medicines, pharmacogenetics.
METABOLIZING ENZYMES
The term pharmacogenetics was introduced into the literature
well before the genomic and genetic advances of the past
decade. Differences between the rates of drug metabolism
among people, associated with particular polymorphic forms
of enzymes involved in drug catabolism, have been established
for several decades (1). Biochemical and, later, genetic variants in
specific isoenzymes such as cytochrome P450 polymorphisms,
were studied to explain the differences in the rates of drug
metabolism. New molecular methods now measure the same
data more rapidly. This science has been extensively reviewed,
and spawned numerous specialty journals devoted to absorption,
distribution, metabolism and excretion (ADME) (2,3). This
expertise is critical and well integrated into the development of
medicines, but is not the subject of this review.
Rather, the focus of this review is the ongoing construction
of standardized methods of rapidly distinguishing individuals
in the population who, based on their genetic make-up,
respond in a particular way to a specific environmental factor
such as a medicine. Metabolizing enzymes represent distinct
candidate genes with recognized functions that can be
hypothesized to influence response to a medicine and therefore
warrant investigation. This candidate approach has a number of
limitations, such as pursuing an attractive, yet incorrect hypo-
thesis. However, profiling of individuals using standardized tools
that resemble whole-genome fingerprint analyses will be
possible in the near future and will begin to obviate the need
for hypothesis generation around favored genes by determining
genetically associated targets (4,5).
FINGERPRINTS: NOT SIMPLE ASSOCIATIONS OR
HAPLOTYPES
Associations between inherited variants and clinical disease
are at the heart of modern medical genetics. When a change in
coding sequence always results in a particular disease expression
over an observable time frame, the effect of the variant is said
to be highly penetrant. If that variant is rare (<1% allele
frequency), it is referred to as a mutation. During the past
century many diseases have been associated with relatively
rare mutations, when either one (dominant) or two variant
*Tel: +1 919 483 7418; Fax: +1 919 315 6013; Email: adr69412@gsk.com
2262 Human Molecular Genetics, 2001, Vol. 10, No. 20
(recessive) copies are inherited. Science is beginning to recog-
nize less rare gene variations that are associated with more
common diseases in a less than cause-and-effect manner and
therefore are called susceptibility genes. An example of a rare
mutation is the triple repeat variant always associated with
Huntingtons disease, although the age of onset varies with the
length of the repeat elements (6,7). The association of the
APOE4 variant with the age of onset distribution of Alzheimers
disease, as well as the protective effect of the APOE2 variant, are
still the most documented and confirmed examples of suscep-
tibility polymorphisms (8,9). The associations of specific APOE
polymorphisms with common forms of Alzheimers disease were
identified within a chromosome 19 region that had been
genetically linked to the disease (10). Combinations of two or
more variants (haplotypes) can also be associated with the
expression of diseases. Sometimes the variants are within the
same gene, sometimes within the same group of genes on a
single chromosome, and sometimes on different chromo-
somes. These variants can be subject to independent selection
during meiosis if they are on separate chromosomes, or if
recombination events occur between markers on the same
chromosome. Thus, the mathematics for identifying disease
associations with multiple haplotype markers from separate
chromosomes requires very large numbers of patients and
controls as well as statistical corrections (e.g. Bonferroni
corrections) to obtain statistically significant results.
In contrast to measuring disconnected variants across the
genome, such as single-nucleotide polymorphisms (SNPs) that
are nominated from a collection of candidate gene variants,
thousands of ordered variants along the genome can be
measured rapidly as a standardized fingerprint or SNP
Print
sm
. The location and order of the variants will always be
the same but the marker polymorphisms will vary in different
individuals. This phenomenon allows patterns of disease or
phenotype-specific associations at each point to be identified at
very specific locations along a common, ordered linear array.
Multiple closely ordered polymorphisms, especially those within
small DNA linkage disequilibrium (LD) regions (50150 kb) that
are inherited together over many generations, can distinguish a
region of LD from other combinations. In essence, they date
the time in evolution when a recombination event occurred in
the neighborhood of a disease susceptibility polymorphism.
Therefore, different combinations or haplotypes of stable
inherited sequence markers within defined LD regions may
mark the DNA strands in the populations that are associated
with disease or phenotype susceptibility. This concept has
already been demonstrated experimentally with the region of
DNA surrounding the APOE4 polymorphism, the well known
susceptibility polymorphism associated with a younger age of
onset of Alzheimers disease. APOE4 was re-identified using a
high-density single-nucleotide polymorphism mapping analysis
across a four million base region that contained the APOE gene
(1113). A short series of SNP markers that are highly associ-
ated with AD (and in LD with APOE4) was demonstrated to
define a DNA region of approximately 45 000 bases (Fig. 1).
Identification of this small specific region against the common
background of no association narrowed the focus to only two
genes coded within that relatively small LD region, APOC-1
and APOE. Only the APOE4 variant of APOE at codon 112 (of
299) was associated with the younger age of onset distribution
of AD (14). This finding has been verified by other studies of
epidemiology in multiple populations with differences in
control allele frequencies, immunocytopathology of human
brain, analyses of neuronal APOE RNA and protein expression
in transgenic mice, among others (1517). In fact, the APOE2
variant at codon 158 was not related to the expression of AD.
Inheritance of each APOE2 allele was associated with an older age
of onset distribution as a protective allele. Yet within the larger
region of LD several other SNPs not within the APOE gene
demonstrated association with AD, but with no biological
relevance. These SNPs simply marked the APOE4 ancestral LD
region compared with the later appearance of APOE3 and its
associated LD region markers that occurred on the APOE3
background DNA strand with evolution. SNP mapping at other
known loci support the detection of regions of extensive
linkage disquilibrium (18,19).
Figure 1. High-density SNP map around APOE4. LD region of 40 kb around the APOE4 polymorphism that is highly associated with Alzheimers disease. Three
additional SNPs demonstrated highly significant association comparing the allele frequencies in Alzheimers disease and controls. Those three SNPs were not part
of the APOE coding region and defined a small region of LD that contained only two genes, APOE and APOC1, within a 4 Mb mapped region of chromosome 19
(Fig. 4). Had this technology been available in 1988 when linkage was described originally (10), the identification of the susceptibility polymorphisms would have
been reduced from years to weeks. [Figure by A.D.Roses (5); published with permission from Nature.]
Human Molecular Genetics, 2001, Vol. 10, No. 20 2263
Subsequently, high-density SNP mapping was used to
identify other previously unknown disease-associated poly-
morphisms within previously unknown susceptibility genes.
Thus, LD association mapping, defined in the context of
disease versus control associations for each point along a high-
density linear SNP map, allows a small region of DNA that
contains very few genes to be identified rapidly. This has been
demonstrated for both psoriasis and migraine, where an 80 kb
region and a 120 kb region were identified, respectively.
Several polymorphisms of a single gene within each small
region of LD are associated with the disease. For psoriasis, this
is a previously novel gene, designated PAT1 or psoriasis
associated transporter 1 (20). It belongs to a class of genes that
were previously unrecognized as playing a role in psoriasis
pathogenesis (Fig. 2A). Several polymorphisms within the
previously well known insulin receptor gene, INSR, on
chromosome 19 are also highly associated with migraine (21)
(Fig. 2B). These disease-specific associations provide the
impetus and direction for further validation related to disease
pathogenesis and to pharmaceutical targeting. The LD association
methodology can also be modeled to localize more traditional
mutations in more traditional highly penetrant diseases in
families.
ANALOGY TO MASS SPECTROSCOPY ANALYSES
Mass spectroscopy is emerging as an effective and accurate
platform for industrial scale SNP analyses (22,23). As a tool
for molecular differentiation, matrix-assisted laser desorption/
ionization-time of flight (MALDI-TOF) mass spectroscopy
has considerable advantages because of its accuracy and ability
to multiplex many samples (24). The read-out from mass
spectroscopy plots the intensity (relative quantity) of the peak
on the y-axis, against the mass in Da on the x-axis (Fig. 3).
PCR fragments that differ in a single base can easily be
differentiated. There are several recent articles describing the
emergence of this method for high-throughput SNP analyses
(2224).
The SNP map of the human genome can also be represented
on the x-axis as a linear order (analogous to increasing mass)
so that if 200 000 consecutively ordered SNPs are to be
analyzed, they can be numbered 1200 000 sequentially. (The
optimum number of SNPs would be the minimum number of
SNPs that could define the LD haplotype map for all common
ethnic and racial groups.) Similarly, if the association at each
SNP for the disease or the adverse event (AE) in patients
compared with controls is represented along the y-axis, then
the locations of LD peaks could be precisely localized and
compared on the map. If association peaks are found in the
disease or AE samples, the implication is that genes within that
small region are involved in the etiology of the phenotype
(Fig. 1). The comparison of SNP maps to determine whether
identical peaks of LD along the genome can be identified and
associated with a defined phenotype can be considered as an
SNP fingerprint or an SNP Print
sm
. Various means of pattern
analyses can be applied for rapid identification of disease or
AE peaks of LD (Figs 3 and 4).
ENVIRONMENT AND GENETICS:
AE-PHARMACOGENETICS
The initial commercial impact of AE-pharmacogenetics will be
on the safety of marketed medicines. There are few social or
ethical controversies about AEs. Everyone dislikes AEs:
patients, physicians, pharmaceutical companies and regulators.
They can all agree that AEs are bad, potentially lethal,
expensive and should be eliminated. AEs represent the specific
interaction of environment and genetics. A drug is introduced
into the patients environment. In some cases there is an
idiosyncratic reaction, an AE that has a stereotypic phenotype
Figure 2. (A) PsoriasisPAT1. LD region of 80 kb found within a larger mapped area of chromosome 3 comparing DNA from psoriasis patients versus controls.
There were four genes coded in this region of LD, only one of which contained several highly significant differences in allele frequency within a single gene, PAT1.
(B) MigraineINSR. LD region of 120 kb found within a larger mapped area of chromosome 19 comparing DNA from migraine patients versus controls. A single
gene was located within this region, the insulin receptor gene, which contained single highly significant SNPs. [Figure by A.D.Roses (5); published with
permission from Nature.]
2264 Human Molecular Genetics, 2001, Vol. 10, No. 20
and is directly related to the timing of the treatment or environ-
mental insult. The objective of AE-pharmacogenetics is to
rapidly identify a genetic profile that characterizes patients
who are more likely to suffer the AE, compared with other
patients who are likely to respond to the drug safely. The goal
is to identify several specific and reproducible regions of LD
associations that are characteristic of patients with AEs and are
not shared by patients in whom the drug was well tolerated.
The panel of widely separated small groups of high-density
SNP mapping peaks across the genome can then be abstracted
to a much simpler set of SNPs within and adjacent to (as
internal controls) those LD regions. This markedly smaller
panel of SNPs (e.g. SNP Print
sm
) can then be used to screen
individuals rapidly, in a convenient overnight test before their
prescriptions are filled. Thus, with drugs that are highly
effective but not used extensively because of the risk of AEs,
the safety profile of the medication can be improved. Because
the environmental factor, the medicine, and the host genetics
are both predictable, the avoidance of AEs can be based on
SNP Prints
sm
or predictable patterns of LD (Fig. 5).
It will also be possible to limit the SNP Print
sm
to only poly-
morphisms that are not knowingly uniquely diagnostic of any
specific disease, i.e. disease mutations or strong susceptibility
polymorphisms. A standardized SNP Print
sm
without diagnostic
constituents would decrease the dissemination of negative
collated information to relatives of the patient.

In addition, it
Figure 3. An illustration of the form that mass spectroscopy data takes in the analysis of mutations (polymorphisms) of the hemoglobin gene. SNP Print
sm
will have
multiple LD peaks specifically and accurately located along the genome map. The shape and recognition of differences along a linear array has great similarities
to the display of data from SNP association studies along an ordered SNP map array (Fig. 4). [Figure provided by Charles Cantor, Sequenom, Inc.]
Figure 4. SNP mapping of the APOE region. The LD data from Figure 1 is illustrated along a much longer 2 Mb scale to demonstrate the ease of recognition.
Multiple peaks of LD regions along the whole genome would appear as sharp peaks, similar to the appearance along a linear map as that of mass spectroscopy data.
Each peak would define a LD region within which there are association differences.
Human Molecular Genetics, 2001, Vol. 10, No. 20 2265
will be possible to determine which variants within specific genes
that are localized within the regions of LD may contribute to the
AE. In fact, in cases where the same idiosyncratic clinical pheno-
type is caused by multiple drugs, it may be possible to discover
the panel of susceptibility polymorphisms and target new
drugs that do not entail the AE risk for multiple diseases in
people who carry a common genetically susceptible SNP
Print
sm
.
THE NUTS AND BOLTS OF
AE-PHARMACOGENETICS
AE-pharmacogenetics is not about revolutionary scientific
insight into disease mechanisms or cellular interactions.
Rather, it is a pattern-recognition scoring system based on
genetic variations and LD for rapid selection of individuals at
risk of serious side effects upon encountering an environmental
factor called a medicine. It is happening now. It is not a wild pipe
dream. It is expensive, and therefore the initial proof-of-principle
experiments are designed to have some commercial value and
are not the type of research undertaken in academic settings.
What is required for demonstrating proof of principle for
AE-pharmacogenetics? There are essentially two elements, the
patients and the tools. While the protocols for obtaining DNA
samples from both AE-patients and patients who have taken
the drug with no AEs are not trivial, the use of the tools and
bioinformatic applications are more germane to the readership
of Human Molecular Genetics. First, a dense, ordered map of
easily measured SNP variants defining regions of LD across
the genome is needed. Sufficient SNPs were placed into the
public domain by The SNP Consortium, and this resource
continues to expand (25). Secondly, there should be rapid,
accurate, automated, high-throughput and inexpensive
($0.001/SNP) genotyping assays available to measure approxi-
mately 100 000200 000 SNPs from each patient and control, and
tens of millions of SNPs from each experiment. Competition is
very active in the biotechnology sector to develop industrial
genotyping capacity, and the retail price per SNP has come
down from $1 to $0.10 per SNP over the past year. The tech-
nology will continue to develop and the cost will continue to
decrease to less than $0.01 per SNP, just as it has with DNA
sequencing.
Also critical is the need for high-capacity bioinformatic
analyses of the huge databases containing clinical phenotypes
as well as SNP data. Databases with the capacity to handle and
analyze large amounts of data will also require investment in
high capacity computing structure. Parallel to industrialization
of SNP mapping, bioinformatic platforms for handling the data
are also being developed. The construction of a readable
interpretation of the data (e.g. an SNP Print) that can be used
for registration purposes with regulatory authorities and as a
pre-prescription test will result in a commercial product that
can be used optimally.
PROOF-OF-PRINCIPLE EXPERIMENTS
It has taken only 4 years to go from the first high-density SNP
maps that demonstrated rapid localization of disease susceptibility
genes to the completion of an ordered whole-genome SNP
map. The next phase, which is currently in progress at Glaxo-
SmithKline, is the selection of SNPs to be used for whole-
genome mapping, and the SNP measurement techniques and
bioinformatic capability to analyze the data. In parallel to the
building of a technical base, patient DNA has been collected to
enable proof-of-principle studies to be conducted.
Abacavir is a very useful and effective medicine used for
HIV treatment. A recognizable hypersensitivity syndrome
occurs as an AE in 5% of patients who use the drug. The HIV
community is well aware of the hypersensitivity AE, resulting
in limitation of the use of this very effective medicine. During
the last 2 years, DNA samples have been collected using
appropriate informed consent and IRB approvals. A large
series of patients with hypersensitivity to abacavir, and control
patients who have been dose-matched but who did not develop
hypersensitivity are ready to be tested. Following the construction
of the SNP map and PCR primer set, the whole-genome
mapping experiment will be executed in 2002 with anticipated
data available for publication early in 2003. The time frame for
the prosecution of the first proof-of-principle experiments for
AE-pharmacogenetics is less protracted that most commentators
suggest (26).
It is, of course, far more likely that AE-pharmacogenetics
will be developed in industry. The investment associated with
clinical ascertainment, data and tissue collection and the
capability to SNP map is more than academic institutions
would invest. The return on investment for a pharmaceutical
company is the increased safety profile of the marketed
medicine and the competitive advantage that safety provides.
In the case of abacavir, the medicine is a component of a triple
therapy combination that allows a patient to decrease the cost
and increase the convenience of being treated. Rather than
ingesting tens of pills at various times of the day, some with
meals and some without, Trizivir

can be taken twice a day and

reduces the heavy pill burden, if the patient can tolerate
abacavir. Thus, the patients benefit, the costs of drugs for
treatment are reduced, and the company that made the invest-
ment in AE-pharmacogenetics derives a return by marketing a
safer and more competitive product.
Figure 5. SNP Print
sm
or SNP LD profiles. The difference of association
between patients with a defined AE and those patients with no AE. Multiple
peaks similar to Figure 4 can be identified on screening, with a finer map of the
LD based on a smaller fragment size illustrated as in Figure 1. For an SNP test,
only the SNPs defining the difference and some technical control SNPs would
be necessary to analyze before a new patient began the medication. [Figure by
A.D.Roses (5); published with permission from Nature.]
2266 Human Molecular Genetics, 2001, Vol. 10, No. 20
EFFICACY PHARMACOGENETICS
The drivers for efficacy pharmacogenetics differ in several
major respects from those for AE-pharmacogenetics. First,
unlike AEs there is currently no social consensus to support
efficacy-based SNP Prints
sm
. While patients may want to be
prescribed and take medicines that work, regulators are more
concerned with limiting AEs than enhancing efficacy and cost
effectiveness. In addition, the potential medical and marketing
impact of efficacy pharmacogenetics warrants careful consid-
eration on a case-by-case basis. If phase III trials are limited to
those patients with a drug-responsive SNP Print
sm
defined in
phase II, then clinical trials could be performed faster, with
fewer patients and less expense. This would, of course,
segment the patient group for which the drug is indicated.
Physicians would also prefer to prescribe efficacious drugs
defined by an evidence-based SNP Print
sm
, but pharmaceutical
companies may worry about limiting their markets. Efficacy
pharmacogenetics can be cost-effective and profit making,
while eventually serving the whole population. Molecules for
non-responders in a phase II trial could be developed in real
time, when non-responders are identified, rather than the
current system of trial and error in the market place for many
years. In the business world, this is referred to as mass custom-
ization, but it is usually based on what automobile or computer
is preferred, not whether the purchased product is optimized
for your health.
The phenotype of efficacy may be more uncertain clinically
than the occurrence of AEs. The drug discovery process will
still need to progress to phase II studies before a well-defined
efficacy profile in a sufficient number of patients could be
tested. Those molecules with obvious efficacy in a large
proportion of patients would not necessarily benefit from SNP
profiling, but those with clear efficacy in limited sub-groups
may benefit, especially if the sub-group would be considered
too small to develop a commercially viable product. By
including only those patients with the apparent efficacy SNP
Print
sm
, the size of larger, double-blind, clinical development
studies would be reduced. Development of new medicines,
including high efficacy potential blockbusters could be accelerated
by smaller, faster clinical trials for response-defined patient
groups.
Another confounding variable is the placebo effect. For
genetic analyses, a strong placebo effect would be indistin-
guishable from a genetic susceptibility effect, thereby making
the determination of an efficacy SNP Print
sm
more difficult.
For genetic analyses, the placebo effect is the equivalent of
pseudo-genes. Modeling studies are currently underway to
examine these variables: disease by disease, molecule by
molecule. The commercial advantage is actually quite simple:
clinically non-responsive patients could be identified early in
the clinical studies, so that additional lead molecules may be
targeted to them (early drug discovery customization).
Traditionally, this market is generally not identified until years
after a medicine is used extensively. One can envision a series
of medicines for each disease that are based on SNP Prints
sm
.
Fewer trial and error prescriptions would be better for the
practice of medicine as well as for those companies that market
safe and effective drugs. While pharmaceutical companies
could create value through the mass customization of
medicines, the total drug bill is anticipated to decrease and the
rate of effective treatment would increase. This results from
mass segmentation with more patients being prescribed the
correct medication for them, and fewer individual patients
being prescribed medicines which are poorly tolerated or non-
effective. Changes in the drug discovery process would occur
to meet the need of multiple customized products, with several
lead molecules and several analogues carried through drug
discovery.
The first examples of efficacy pharmacogenetics are in
clinical use today. Perhaps the best known example is the use
of the DAKO Hercep Test, which is a semi-quantitative
assay for testing breast tumor tissue for over-expression of the
HER-2/neu protein. Herceptin

is a biologic (antibody)
approved by regulators for therapeutic use only in patients
whose tumors test positive for the HER-2/neu protein. Thus,
breast tumor patients were sub-divided into predictable
responders and non-responders for clinical trials and subse-
quent marketing. The clinical effect of the product was demon-
strated through the testing of breast cancer patients
(approximately one-third test positive), whereas the clinical
effect may have seemed marginal if all patients were analyzed
as a whole (27,28).
ETHICS AND PRIVACY
Many of the technologies and strategies described above will
require an in-depth examination of ethical and data privacy
issues. For example, false positives for AE-pharmacogenetics
(i.e. the patient is incorrectly predicted to be a poor candidate
for the medicine) performed in patients with a disease prior to
receiving a drug adds no diagnostic information. In fact, in
such an example, a false positive would only indicate that a
person should not take a medicine. As long as the SNP Print
sm
contained no SNP that would provide primary diagnostic
information for any disease, the risk of accidental discovery of
unwanted medical information would be minimal (29). Of
greater significance, concerns about insurance companies
taking advantage of the test for undisclosed diagnostic
information would be largely minimized. A separate set of
SNPs and other polymorphisms could be used for disease
diagnostic purposes, the use of which would remain to be
debated publicly with respect to ethics and data privacy. As an
aside, the largest variable in that debate is whether the pertinent
population of patients has medical coverage guaranteed for all or
whether the risk for disease diagnostic capability is only a
problem for those who must qualify for medical insurance.
Another issue concerns the commercial use of research data
and informed consent considerations. AE-pharmacogenetics
will undoubtedly involve patients who are prescribed a
medication and have provided informed consent regarding the
use of their information for the purpose of analyzing AEs,
should they occur. Patients participating in disease-specific
genetic studies that are designed to research common diseases
will continue to be consented as they have for decades. Well
phenotyped patient groups will be in demand, as more drug
discovery and development pipelines will be based on human
genetic targets. The immediate danger is quite obvious. Most
studies of patient populations with specific diseases are
performed in academic medical centers, often without specific
informed consent for commercial uses. Some time later, when
a project has been successful, the data derived from studies of
Human Molecular Genetics, 2001, Vol. 10, No. 20 2267
those patients may be commercialized in a biotechnical
company or used by a large pharmaceutical company. Care
must be taken to inform patients and controls of these commercial
implications in a clear manner that does not threaten their
medical care.
There are many other points worthy of discussion regarding
privacy, trusted third party DNA banks, informed consent
issues, medical data access and the differences between
disease-specific genetic studies and AE and efficacy pharmaco-
genetic studies. Additionally, the role of regulators will need to
be modified as systematic methods for identifying patients at
greater risk of AEs will provide evidence-based data that will
need to be incorporated into surveillance and case reporting
systems. The practice of medicine will continue to evolve based
on better diagnostics, safer and more effective medicines, reduc-
tions in unmet medical needs, use of appropriate technologies
and increasing sensitivity to individuals interests and data
privacy. Publicity tends to focus on the negative and shocking
scenarios but, as the science is explained and incorporated into
health care, education and the anticipation of positive scenarios
will defeat major medical and iatrogenic problems.
REFERENCES
1. Poolsup, N., Li Wan Po, A. and Knight, T.L. (2000) Pharmacogenetics
and psychopharmacotherapy. J. Clin. Pharm. Ther., 25, 197220.
2. Caldwell, J., Gardner, I. and Swales, N. (1995) An introduction to drug
disposition: the basic principles of absorption, distribution, metabolism,
and excretion. Toxicol. Pathol., 23, 102114.
3. Vesell, E.S. (2000) Advances in pharmacogenetics and pharmacogenomics.
J. Clin. Pharmacol., 40, 930938.
4. Roses, A.D. (2000) Pharmacogenetics and future drug development and
delivery. Lancet, 355, 13581361.
5. Roses, A.D. (2000) Pharmacogenetics and the practice of medicine.
Nature, 405, 857865.
6. The Huntingtons Disease Collaborative Research Group (1993) A novel
gene containing a trinucleotide repeat that is expanded and unstable on
Huntingtons disease chromosomes. Cell, 72, 971983.
7. Gusella, J.F. and MacDonald, M.E. (1994) Huntingtons disease and
repeating trinucleotides. New Engl. J. Med., 330, 14501451.
8. Saunders, A.M., Strittmatter, W.J., Schmechel, D., George-Hyslop, P.H.,
Pericak-Vance, M.A., Joo, S.H., Rosi, B.L., Gusella, J.F.,
Crapper-MacLachlan, D.R., Alberts, M.J. et al. (1993) Association of
apolipoprotein E allele epsilon 4 with late-onset familial and sporadic
Alzheimers disease. Neurology, 43, 14671472.
9. Corder, E.H., Saunders, A.M., Risch, N.J., Strittmatter, W.J., Schmechel, D.E.,
Gaskell, P.C.,Jr, Rimmler, J.B., Locke, P.A., Conneally, P.M., Schmader, K.E.
et al. (1994) Protective effect of apolipoprotein E type 2 allele for late
onset Alzheimer disease. Nat. Genet., 7, 180184.
10. Pericak-Vance, M.A., Bebout, J.L., Gaskell, P.C.,Jr, Yamaoka, L.H.,
Hung, W.-Y., Alberts, M.J., Walker, A.P., Bartlett, R.J., Haynes, C.A.,
Welsh, K.A. et al. (1991) Linkage studies in familial Alzheimers disease:
evidence on chromosome 19 linkage. Am. J. Hum. Genet., 48, 10341050.
11. Lai, E., Riley, J., Purvis, I. and Roses, A.D. (1998) A 4-Mb high-density
single nucleotide polymorphism-based map around human APOE.
Genomics, 54, 3138.
12. Martin, E.R., Lai, E.H., Gilbert, J.R., Rogala, A.R., Afshari, A.J., Riley, J.,
Finch, K.L., Stevens, J.F., Livak, K.J., Slotterbeck, B.D. et al. (2000)
SNPing away at complex diseases: analysis of single nucleotide
polymorphisms around APOE in Alzheimer disease. Am. J. Hum. Genet.,
67, 383394.
13. Roses, A.D. (2000) Pharmacogenetics and the practice of medicine.
Nature, 405, 857865.
14. Roses, A.D. (1998) Genetic associations. Lancet, 351, 916.
15. Relkin, N.R., Tanzi, R., Breitner, J., Farrer, L., Gandy, S., Haines, J.,
Hyman, B., Mullan, M., Poirer, J., Strittmatter, W. et al. (1996)
Apolipoprotein E genotyping in Alzheimers disease (consensus
statement of the National Institute of Aging/Alzheimers Association
Working Group). Lancet, 347, 10911095.
16. Roses, A.D. (1997) Apolipoprotein E, a complex gene with biological
interactions in the aging brain. Neurobiol. Dis., 4, 170185.
17. Xu, P.-T., Schmechel, D., Rothrock-Christian, T., Burkhart, D.S., Qiu, H.-L.,
Popko, B., Sullivan, P., Maeda, N., Saunders, A.M., Roses, A.D. and
Gilbert, J.R. (1996) Human apolipoprotein E2, E3 and E4 isoform specific
transgenic mice: human-like pattern of neuronal immunoreactivity in
central nervous system not observed in wild type mice. Neurobiol. Dis., 3,
229245.
18. Bonnen, P.E., Story, M.D., Ashorn, C.L., Buchholz, T.A., Weil, M.M. and
Nelson, D.L. (2000) Haplotypes at ATM identify coding-sequence
variation and indicate a region of extensive linkage disequilibrium.
Am. J. Hum. Genet., 67, 14371451.
19. Koch, H.G., McClay, J., Loh, E.W., Higuchi, S., Zhao, J.H., Sham, P.,
Ball, D. and Craig, I.W. (2000) Allele association studies with SSR and
SNP markers at known physical distances within a 1 Mb region embracing
the ALDH2 locus in the Japanese, demonstrates linkage disequilibrium.
Hum. Mol. Genet., 9, 29932999.
20. Hewett, D., Samuelsson, L., Polding, J., Enlund, F., Cantone, K., See, C.G.,
Smart, D., Chadha, S., Inerot, A., Enerback, C. et al. (2001) Identification
of a psoriasis susceptibility candidate gene by linkage disequilibrium
mapping with a localised single nucleotide polymorphism map.
Genomics, in press.
21. McCarthy, L.C., Hosford, D.A., Riley, J.H., Bird, M.I., White, N.J.,
Hewett, D.R., Peroutka, S.J., Griffiths, L.R., Boyd, P.R., Lea, R.A. et al.
(2001) Single nucleotide polymorphism (SNP) alleles in the Insulin
Receptor (INSR) gene are associated with typical migraine. Genomics,
in press.
22. Jarman, K.H., Daly, D.S., Petersen, C.E., Saenz, A.J., Valentine, N.B. and
Wahl, K.L. (1999) Extracting and visualizing matrix-assisted laser
desorption/ionization time of flight mass spectral fingerprints.
Rapid Commun. Mass Spectr., 13, 15861594.
23. Bakhtiar, R. and Nelson, R.W. (2000) Electrospray ionization and
matrix-assisted laser desorpiton ionization mass spectrometry. Emerging
technologies in biomedical sciences. Biochem. Pharmacol., 59, 891905.
24. Leuschner, J. (2000) MALDI TOF mass spectroscopy: an emerging
platform for genomics and diagnostics. Expert. Rev. Mol. Diagn., 1, 1118
25. The International SNP Map Working Group (2001) A map of human
genome sequence variation containing 1.42 million single nucleotide
polymorphisms. Nature, 409, 928933.
26. Post, G. (2000) Molecular diagnostics: a powerful new component of the
healthcare value chain. Expert. Rev. Mol. Diagn., 1, 15.
27. Wang, S., Saboorian, M.H., Frenkel, E., Hynan, L., Gokaslan, S.T. and
Ashfaq, R. (2000) Laboratory assessment of the status of Her-2/neu
protein and oncogene in breast cancer specimens: comparison of
immunohistochemistry assay with fluorescence in situ hybridisation
assays. J. Clin. Pathol., 53, 374381.
28. Baselga, J. (2000) Current and planned clinical trails with trastuzumab.
Semin. Oncol., 27, 2732.
29. Roses, A.D. (2001) How pharmacogenetics will affect the practice of
neurology. Arch. Neurology, in press.
2268 Human Molecular Genetics, 2001, Vol. 10, No. 20