Only Connect Personal Genomics and the Future of American Medicine Misha Angrist Institute for Genome Sciences & Policy, Duke University, Durham, North Carolina, USA Abstract Access to one���s own complete genome was unheard of just a few years ago. At present we have a smattering of identifiable complete human genomes, but the coming months and years will undoubtedly bring thousands more. What will this mean for the practice of medicine in the US? No one knows, but given the remarkable drop in the cost of DNA sequencing over the last few years, it seems a safe bet that within the next decade, primary care physicians will order patients��� whole genome sequences with no more fanfare than they would a complete blood count. But the challenges of transforming that easily accessible information into cost savings and better health outcomes will be daunting. Obviously, we lack interpretive abilities and phenotypic information commensurate with our skill in amassing DNA sequences. Worse, we have exacerbated these problems by failing to embrace the increasing ubiquity of genomic information, the populace���s interest in it, and its relevance to virtually every medical specialty. The success of personal genomics will require a profound cultural shift by every entity with a stake in human health. ������The future is the worst thing about the present.������ ��� Gustave Flaubert[1] Whatever the ultimate fate of healthcare reform in the US,[2] whatever the limitations of genome-wide association studies,[3] whatever loopholes exist in the Genetic Information Non- discrimination Act of 2008,[4] whatever objections the medical establishment offers to the concept of the masses obtaining their genomic information,[5,6] make no mistake: the genomes are coming. It is inevitable. In the last few years, sequencing costs have plummeted by more than three orders of magni- tude.[7] The recent announcement that the California startup company Complete Genomics Inc. was able to sequence com- plete human genomes for less than $US5000 in reagents sug- gests that the arrival of the mythical $US1000 genome will soon be at hand.[8] The data thus far suggest that people want this information, even with serious and abiding questions surround- ing its predictive value.[9-11] As I write this in early 2010, a handful of identifiable people have had their genomes sequenced at high coverage.[12,13] I am one of them, having had my whole genome sequenced by David Goldstein���s lab in 2009 and my partial exome (the 1% of the genome that codes for protein) sequenced as part of the Personal Genome Project (PGP).[14] I would not presume to extrapolate my experience to a population-wide genome- screening model the infrastructure attached to whole-genome sequencing remains too inchoate and is not yet scalable to whole populations. But my collected experiences of open- source genotype interpretation via SNPedia (www.SNPedia. com), direct-to-consumer genome scanning, participation in the PGP, and obtaining my own genome ��� along with 3 years of close observation of the burgeoning world of personal geno- mics ��� have, I hope, given me some insight into the requisites for personal genomics to be applied, such that it can be more of a boon to human health than a drain on an already frail health- care system. Here I discuss what I perceive to be some of the most sig- nificant challenges to integration of personal genomics into clinical medicine. I suggest some possible ways forward, though I am under no illusions that any path we choose will be easy. 1. Phenotypes at Large In my view, a lack of detailed phenotypic data is the ���ele- phant in the room��� for personal genomics, at least with respect to clinical utility. We have become highly skilled at aggregating genomic data, but our willingness to collect the phenotypic COMMENTARY Mol Diagn Ther 2010 14 (2): 67-72 1177-1062/10/0002-0067/$49.95/0 �� 2010 Adis Data Information BV. All rights reserved.

information we wish to associate with our reams of genomic information remains inadequate. And our lack of will has consequences: incorrect phenotyping means more samples are needed to achieve equivalent statistical power.[15] Thus, a commitment to better phenotyping makes both scientific and financial sense. In part, our reticence is understandable. When we visit our physicians, they examine us and ask us questions. They take measurements, order tests, whack us on the knees, and try to elicit what, if anything, is wrong with us. This is the time- honored art of clinical diagnosis, and we have yet to come up with a less cumbersome substitute for it. 23andMe, Inc., the most popular direct-to-consumer genome-scanning company, is attempting to carry out research via surveys of its custom- ers.[16] Self-reported data are hardly useless,[17,18] but neither are they the same as a visit to a physician.[19,20] Perhaps the best exemplar of the do-it-yourself approach to phenotyping is PatientsLikeMe.com, which has constructed a sophisticated database that tracks members��� medical informa- tion and has actually led to peer-reviewed publications.[21-23] But the PatientsLikeMe model, at least so far, is based on communities of people who share frank clinical diseases and have already been immersed in the healthcare system because of them.[24] PatientsLikeMe is an extraordinary achievement and, in my opinion, something to be supported and emulated. But if personal genomics practitioners are to realize their goal of preventing disease, they must bring the PatientsLikeMe ap- proach to bear on the healthy among us. They must cultivate communities of well people who are willing to invest the time and effort to attend to their own phenotypes with the same exactitude as PatientsLikeMe members coping with chronic diseases. Otherwise, with few exceptions, personal genomics has, in my estimation, dragged its feet with respect to phenotypes. The 1000 Genomes Project is collecting no trait data.[25] Nor, out- side of surveys, are the personal genomics companies (to my knowledge). The DataBase of Genotypes and Phenotypes (dbGAP) includes de-identified phenotypic data submitted by individual investigators de-identification has been criticized because it may impoverish phenotypic data in ways that open- consent platforms such as the PGP and PatientsLikeMe need not confront.[26] As of early 2010, the PGP relied on partici- pants to submit their own medical information.[14] The hope is that more entities will collect such data prospectively, just as the UK Biobank is doing.[27] Obviously, phenotype collection is difficult and expensive. It cannot be automated easily and may require high-level medical expertise and clinical acumen. But it can be done. Paabo and colleagues,[28-30] in their extensive studies of the FOXP2 gene in the mammalian lineage and its contribution to language, col- lected data on hundreds of distinct traits in mice. Could a similar effort be undertaken in humans? I believe it could. Indeed, I would argue that this pressing need presents an opportunity for the medical genetics community to assert itself. The late Victor McKusick called medical geneticists ������the last generalists������ (Rienhoff HY, personal communication). They have an appreciation for nuchal folds, bifid uvulas, and other esoteric signs and symptoms that most everyone else simply doesn���t and can���t share. If there is to be a Human Phenome Project, it will require dysmorphologists to assume center stage. My worry, as I discuss below, is that the cultural chasm between traditional medical genetics and the early-adopter community that has embraced personal genomics is simply too wide to realize the goal of widespread human phenomics. 2. Genomic Interpretation: Man���s Search for Meaning Even with phenotypic data in hand, the extraction of clinically and/or personally relevant meaning from human sequences will be among the most daunting tasks. But interpretation is another obvious necessity. Consider Alzheimer���s disease and the apolipoprotein E (APOE) gene. Two copies of the APOE e4 allele are associated with a 10- to 15-fold greater risk of garden-variety late-onset Alzheimer���s disease.[31,32] Robert Green et al.[33] found that many individuals at risk for Alzheimer���s disease were interested in learning their genotypes for the APOE gene. Participants in Green and colleagues��� REVEAL (Risk Evaluation and Edu- cation for Alzheimer���s Disease) studies[33-36] who learn that they are APOE e4 positive tend to recall this information more frequently, rarely regret their decision, and are more likely to change their health behaviors. But APOE is exceptional in many ways. We have 15 years of population data on APOE. APOE e4 is not deterministic, but its association with Alzheimer���s disease is powerful and robust. We know what it means, at least in probabilistic terms. What happens when individuals are presented with information per- tainingtotraitsgovernedbymanymoreweakallelesinmanymore genes coupled with environmental effects, as most seem to be? In the near term, barring the discovery of more APOE-type loci, it might behoove us to focus initially on carrier status for traits transmitted in a straightforward Mendelian way. Indeed, at least one company has staked its future on offering custo- mers exactly this sort of information.[37] But let���s not kid ourselves: even so-called ���simple��� Mendelian disorders are 68 Angrist �� 2010 Adis Data Information BV. All rights reserved. Mol Diagn Ther 2010 14 (2)