On the bumpy road towards ‘personalized medicine’

My uncle who practiced general medicine in a German village all his life wouldn't understand what I'm talking about: the care of his patients was always highly personalized. A new vision of ‘personalized medicine’ arose in contrast to current medical practice that relies on standards of care derived from epidemiologic studies of large cohorts. New molecular methods increasingly will allow us to peak into individual genomes and encourage a design of individualized prevention and treatment plans that may well deviate from the standard of practice now tailored to the ‘average’ person.

»The good news: causative genes for rare Mendelian disorders have been discovered at a stunning rate.«

In 2007, the Junior Senator from Illinois, Barack Obama, introduced Senate Bill S. 976 entitled ‘Genomics and Personalized Medicine Act of 2007’. This bill, which died in Congress in 2008, defined personalized medicine in the modern sense as ‘the application of genomic and molecular data to better target the delivery of health care, facilitate the discovery and clinical testing of new products, and help determine a person's predisposition to a particular disease or condition’. The bill would have provided a framework for support and oversight.

Where are we stuck on the road to using genomics to tailor medical care to individuals based on their genetic makeup, and how can EMBO Molecular Medicine be of help? The human genome project, the international effort that led to the publication of a draft human DNA sequence in 2001 and a ‘finished’ version in 2003, may have oversold its promise to revolutionize medicine and, thus, generated unrealistic expectations of a quick return. The good news: causative genes for rare Mendelian disorders have been discovered at a stunning rate (<200 in 1990; ∼1500 in 2000 and >2600 now) (http://www.ncbi.nlm.nih.gov/omim/). Yet this is not recognized by the public eye that is fixed on the slow progress in understanding the genetic contributions to common disorders.

In recent years, however, as genome-wide association studies (GWAS) became large enough (>10,000 cases and controls) and the set of single nucleotide polymorphisms (SNP) interrogated on microarrays dense enough (500 k to 1 M), the common disease–common variant paradigm has generated some highly replicated SNP-disease associations http://www.genome.gov/GWAStudies/. At present, these results have little clinical utility because the relative risk changes are small (usually < 2.0) and the major causes of the heritability of common disorders are still obscure.

Meanwhile, the increasing awareness of the role of genetic variation in common diseases has renewed the emphasis on family health history with the development of online tools for individuals to collect health information on their families, in conjunction with a move towards electronic medical records (https://familyhistory.hhs.gov/fhh-web/home.action).

Common disease cohorts include sub-groups of individuals who have inherited a rare Mendelian mutation that puts them at much higher risk, as for hypertension, Parkinson disease, breast cancer, glaucoma and others. Thus, population-based risk data may not be applicable to all individuals. Between the endpoints (i) common disease–common SNP allele (>5%) conferring a small change in risk, and (ii) rare disease–rare allele (<0.1%) conferring a large change in risk (essentially representing Mendelian disorders), there is a continuum, not a dichotomy. Therefore, the next goal is to test genetic variants of intermediate frequency (0.1–5%) for disease associations; and to catalogue all SNP, copy-number variants (CNV) and small insertions/deletions (indel) by sequencing samples from many different populations (http://www.1000genomes.org). It is important to study global ethnic populations as allele frequencies may vary greatly and many reported associations are not validated in Non-European populations (Silander et al, 2009).

»SNP associations are only way posts towards identification of the true causative genomic differences«

People who are disappointed by the small effects on risk of the current GWAS data should realize that the recently reported SNP associations are only way posts towards identification of the true causative genomic differences that confer risk for—or protection from—certain diseases. While GWAS data pinpoint the location of predisposing alleles on the genomic map, only in rare instances do they identify the causative gene, let alone the disease-causing mechanism. For most associations it still needs to be determined which gene(s) in the vicinity of the associated SNPs are responsible and how they increase or decrease risk.

To cite an example, overwhelming evidence points to SNPs within a linkage disequilibrium (LD) block at 9p21.3 as associated with increased risk for coronary heart disease, and possibly abdominal aortic aneurysm, stroke and periodontitis (Schaefer et al, 2009), yet no pathogenic mutations were found in the two protein-coding genes in the region, CDLN2B and CDKN2A. The SNPs with the highest-scored associations are ‘intergenic’, but located in a region that gives rise to a large antisense non-coding RNA (ANRIL) with numerous alternative splice products suggesting that complex regulatory mechanisms at this locus are yet to be uncovered. Progress will require collaboration between clinical and basic scientists.

Studies of common diseases may identify risk alleles and protective alleles. Yet in other instances the genome may contribute an allele that provides resistance to diseases that are entirely environmentally caused, such as the 127V variant in the prion gene that appears to protect against the acquired spongiform encephalopathy kuru (Mead et al, 2009) and the well-known deletion mutation in the chemokine receptor CCR5 that confers resistance to HIV.

»the current bumps in the road to personalized predictive and preventive medicine can be overcome.«

Unsolved issues include the proper modelling for multiple SNP associations considering gene–gene interactions, as one cannot assume that odds ratios can simply be added. Also, effects of genomic variants, especially CNV, on tissue-specific expression patterns, as well as alternative splicing and epigenetic modifications need to be considered. Furthermore, gene–environment interactions need to be understood in detail, as such understanding opens the possibility for targeted environmental interventions. To address all these issues, new statistical and computational tools need to be developed.

I view EMBO Molecular Medicine as the ideal place to publish and debate forthcoming studies on the identification of novel associations, causative variants and disease mechanisms, in conjunction with other genotypic and phenotypic characteristics of the individuals and populations. This research has the potential to discover novel pathways, not previously known to play a role in the disorders studied, and ultimately lead to novel treatments. With multidisciplinary effort and patient pursuit, the current bumps in the road to personalized predictive and preventive medicine can be overcome. A major benefit of the rapidly expanding genomic information lies is the vastly extended possibilities of genetic testing, from SNP genotyping to whole genome sequencing. The major roadblocks for the move of personal genomic information into clinical practice will be the topic of a future Editorial.