Expression profiling: a cost-effective biomarker discovery tool for the personal genome era

Despite the declining cost of ‘personal genomes’ and the acknowledged usefulness of individual genome sequences to understand human health, for the near future, personal genomes alone are unlikely to become the leading research tool in genomic medicine. I argue that, at present, expression profi ling studies are the most promising and cost­eff ective tool for discovering new disease and drug­response biomarkers. $1,000 personal genomes? Not around the corner Th e proliferation of studies on personal genomes is fueled by reduced sequencing costs and improved bio­ informatics software tools. Substantial advances in understanding human biology in health and disease ­ and improved diagnostic capacities ­ are foreseen with the availability of personal genome sequences [1,2]. Yet, ex­ pec tations from personal genomes should not be over­ stated: individual genome sequencing per se ­ even when combined with comprehensive individual medical records ­ will not suffi ce to decipher the overwhelming complexity of human biology and disease pathology. Our current level of understanding of biology in general, and human biology in particular, is simply too low to interpret the information encoded in individual DNA sequences [3]. Moreover, even with dramatically reduced DNA se­ quen cing costs, the high costs of data interpretation subsequent to the tremendous bioinformatics eff ort of analyzing 6.2 Gb nucleotides of two copies of one individual genome means that personal genome sequen­ cing will not become clinically routine in the near future. Currently, at the 10th anniversary of the Human Genome Project’s conclusion, whole genome sequencing is being put into practice in leading medical academic centers, predominantly for cancer diagnosis and treatment. But it will be many years before it becomes a widespread, routine clinical tool.


$1,000 personal genomes? Not around the corner
Th e proliferation of studies on personal genomes is fueled by reduced sequencing costs and improved bio informatics software tools. Substantial advances in understanding human biology in health and disease and improved diagnostic capacities are foreseen with the availability of personal genome sequences [1,2]. Yet, ex pec tations from personal genomes should not be over stated: individual genome sequencing per se even when combined with comprehensive individual medical records will not suffi ce to decipher the overwhelming complexity of human biology and disease pathology. Our current level of understanding of biology in general, and human biology in particular, is simply too low to interpret the information encoded in individual DNA sequences [3].
Moreover, even with dramatically reduced DNA se quen cing costs, the high costs of data interpretation subsequent to the tremendous bioinformatics eff ort of analyzing 6.2 Gb nucleotides of two copies of one individual genome means that personal genome sequen cing will not become clinically routine in the near future. Currently, at the 10th anniversary of the Human Genome Project's conclusion, whole genome sequencing is being put into practice in leading medical academic centers, predominantly for cancer diagnosis and treatment. But it will be many years before it becomes a widespread, routine clinical tool.

The case for expression profi ling
Regardless of the pace of application of personal genome sequencing in clinical practice, researchers should bear in mind that additional tools are needed to supplement personal genomes so that insights can be made into the molecular biology of complex diseases. Among such tools, genomewide transcriptomic profi ling seems to be a powerful yet straightforward, aff ordable and easily validated molecular biology technology for discovering new diagnostic biomarkers and drug targets [4,5]. For example, expression profi ling currently costs under US$400 per sample using commercial microarrays. Com parisons of healthy and diseased tissues from the same individual, or the same tissue (in particular, white blood cells) from an individual over time, such as before and following drug treatment, may yield knowledge not extractable from personal genome sequences. Th e com plexity of the interrelationship between DNA sequences and cell biology is likely to be far higher than is currently understood. For example, a new level of complexity linking the genome to the proteome has recently been introduced: it is well established that gene expression is regulated by short (22 to 23 nucleotides long) noncoding RNA sequences termed microRNAs. Now, an additional level of complexity has been discovered: circular non coding RNAs, which modulate the action of microRNAs on gene expression [6]. Further surprises are likely to be in store for gene expression regulation by noncoding genome sequences, even though these sequences are part of already published but little understood personal genomes.
Another key advantage of expression profi ling studies is that they also inform about the consequences of epi genomic modifi cations, as transcriptomes refl ect not merely the output of DNA sequences, but also their interplay with nongenetic modifi ers of gene expression. Moreover, transcriptomic studies can inform about alter na tive splicing events in particular when RNA se quencing is applied whereas, at our current level of knowledge, personal genome sequences do not have this capacity.
Expression profi ling data can thus be far more infor mative than personal genomes for deciphering cellular networks and disease biomarkers, and indicating drug targets. Certainly, having both personal genomes and longitudinal gene expression profiles from the same study participants has clear advantages. Indeed, an integrative 'personal'omics profile that combines genomic, transcrip tomic, proteomic, metabolomic and autoantibody profiles from a single individual over 14 months was recently presented [7]. This is undeniably the best way forward for genomic medicine projects when adequate funding is available. However, in the clinical setting, as well as for most academic research groups, costs for such compre hensive projects in particular for large cohorts are prohibitive. In lieu of such funding levels, expression profiling seems to offer the most promising and cost effective approach for genomewide searches for disease and drugresponse biomarkers.

Considerations and limitations
Of course, there are limitations to expression profiling, where each profile represents a single 'snapshot' of a given tissue or cell type at a given time and under distinct physiological conditions. This constraint may be over come (with extra cost) by performing longitudinal gene expression profiles so that the sequential alterations provide information about molecular events during disease progression, tissue remodeling or drug treatment. This disadvantage of expression profiling can thus be turned into an advantage, in particular for searching drugresponse biomarkers, by pinpointing genes or non coding RNA sequences whose expression levels are modi fied by a drug of interest. This in turn can be infor mative for discovery of drugresponse biomarkers that can act as companion diagnostics for new drug targets [5].
Personal genome studies entail privacy risks, not only for study participants, but also for their relatives. Researchers are morally obliged to disclose that anony mity promises cannot be made to individuals who consent to genome sequencing [8] and, indeed, there is conclusive evidence that donors of DNA sequences can be identified [9]. By contrast, lesser privacy risks are posed by microarraybased expression profiling studies. This should be considered in particular when studying vulnerable populations such as children [10], where gene expression studies have the additional advantage of afford ing better privacy protection for research participants.
In conclusion, it seems that considering our limited understanding of biology, and the current research funding situation, expression profiling stands out as the most appropriate and costeffective methodology to gain new insights into complex disorders and to discover disease and drugresponse biomarkers.

Competing interests
The author declares that he has no competing interests.