Microsoft word - swan-aaai-final.doc

DIYgenomics crowdsourced health research studies: personal wellness
and preventive medicine through collective intelligence
Melanie Swana,b
P.O. Box 61258, Palo Alto, CA, 94306, USA Abstract
humanity in ways that may have been unthinkable The current era of internet-facilitated bigger data, better previously. Millions of people creating and using content tools, and collective intelligence community computing is on the internet has generated large data sets and tools for accelerating advances in many areas ranging from artificial interpreting and manipulating these data that are intelligence to knowledge generation to public health. In the themselves recursive enablers of subsequent advance. health sector, data volumes are growing with genomic, phenotypic, microbiomic, metabolomic, self-tracking, and Artificial intelligence is one area that has realized other data streams. Simultaneously, tools are proliferating to substantial benefits from the new era of bigger data and allow individuals and groups to make sense of these data in better tools, particularly regarding machine learning, a participatory manner through personal health tracking anomaly detection (e.g.; fraud and spam), and natural devices, mobile health applications, and personal electronic language processing (e.g.; question answering systems like medical records. Health community computing models are emerging to support individual activity and mass IBM's Watson). A quintessential example is Google's success in spelling correction and language translation: crowdsourced health research studies. Participatory health progress in natural language machine learning allowed efforts portend important benefits based on both size and statistical methods to be applied to the large datasets that speed. Studies can be carried out in cohorts of thousands have arisen on the web (Halevy 2009). Simple data instead of hundreds, and it could be possible to apply findings from newly-published studies with near-immediate analysis techniques were successful after decades of speed. One operator of interventional crowdsourced health specially-designed algorithms showed little progress; a key research studies, DIYgenomics, has several crowdsourced point was finally having a very large corpus of data. health research studies in open enrollment as of January One effect of bigger data and better tools is that humans 2012 in the areas of vitamin deficiency, aging, mental themselves start to serve as a computing community, both performance, and epistemology. The farther future of intelligent health community computing could include as individual nodes and through mass collaboration. personal health dashboards, continuous personal health Individuals take in data, process it, and turn it back into the information climates, personal virtual coaches (e.g.; Siri network in new forms with added value. One public good 2.0), and an efficient health frontier of dynamic that has arisen through community computing and personalized health recommendations and action-taking. collective intelligence is the wikipedia (Fallis 2008). The annotation of street map data, creating crowdsourced Introduction
wikipedias of the Earth, is another example (Kamel Boulos 2011). Community computing models are arising in the Bigger Data, Better Tools, and Collective
health sector as well. All of these examples display the predicted Intelligence Computing Communities
communities, escalating in three stages from information- The internet has facilitated the advent of new models of sharing to cooperating to participating in collaborative collective intelligence that are shaping and advancing action (Shirky 2008). Another benefit of internet-based models is that a wide range of diverse participants can be engaged. Crowdsourcing can be used to obtain needed Copyright 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. participants, services, ideas, or content at low cost by soliciting contributions from a large group of people via organized studies have been interventional studies the internet. Better quality outcomes may be obtained designed and operated by citizen scientists, for example through wisdom-of-the-crowds benefits that accrue as those conducted by PatientsLikeMe patients, DIYgenomics people with diverse backgrounds analyze data and propose scientists, and Quantified Self individual novel interpretations. On average, the wisdom of crowds experimenters (Swan forthcoming 2012). The studies arrives at a better answer than any individual can provide, organized by DIYgenomics serve as a particular example including outperforming small groups of experts in making of the emerging collective intelligence health model that is decisions and predictions (Surowiecki 2004). becoming a complement to traditional clinical trials and defining a new ecosystem of preventive medicine (Swan Participatory Health: Bigger Data, Better Tools,
and Collective Intelligence Models
Health may be generating big data faster than any other DIYgenomics Health Research Studies
sector: the number of people participating is growing, the amount of data per person is growing, and the demand for DIYgenomics is a non-profit research organization easy accessibility and robust searchability of these data is established in 2010 for the purpose of organizing group growing. In 2009, it was estimated that all human-created collaboration health research studies with the goal of content to date comprised one zettabyte of information, but that within five years, it would be routine to be generating democratization of health experimentation, studies seek to one zettabyte of information in much shorter time frames, apply the wisdom of crowds to personal health particularly in the field of medicine, due to imaging and management. The generalized hypothesis for studies is that personalized profiling (Enriquez 2010) where billions of one or more genetic polymorphisms (e.g.; mutations) may data points per individual could become the norm (Hood lead to out-of-bounds phenotypic measures (for example, 2011). Better tools are also proliferating in participatory deficient vitamin B levels) that may be ameliorated with medicine and health 2.0, for example with self-tracking personalized intervention. Part of realizing preventive devices, mobile health applications, personal electronic medicine is establishing individualized baseline markers of medical records, health social networks, and crowdsourced wellness, for example a normal level of total cholesterol research studies. A next generation of tools will be for one person might be 130 mg/dL whereas for another required to integrate burgeoning health data streams 180 mg/dL would be normal. It is important for individuals (genomic, phenotypic, microbiomic, metabolomic, self- to have a sense of their own normal levels, and effective measurement tools for learning if there is deviation from these norms. At present, the focus of DIYgenomics studies Health community computing models may offer some is linking genetic mutation with phenotypic evidence and help, for example health social networks and crowdsourced personalized intervention. Additional health data streams health research studies are emerging to support both such as microbiome profiling and whole human genome individual activity and mass collaboration. Health social sequencing will be integrated as they become feasibly networks are online communities for individuals to discuss available to consumers. Seven studies have been launched and inform themselves about conditions, symptoms and in the areas of vitamin deficiency, aging, mental treatments, provide and receive support, track disease performance, and epistemology, and are available for progression, and potentially engage in health studies. At ongoing open enrollment. These studies are operated on present, there are dozens of health social networks, for the Genomera personal health collaboration and genome example, general communities like PatientsLikeMe which sharing platform, and links to the studies are available from has over 125,000 patients in 1,000+ conditions as of January 2012, and condition-specific communities like SugarStats for diabetes. Vitamin Deficiency Studies
Crowdsourced health research studies are investigative DIYgenomics has two vitamin deficiency studies projects conducted by individuals or groups for the purpose underway, investigating the possibility that one or more of understanding and/or improving a health-related issue. genetic polymorphisms (e.g.; mutations) may lead to Studies may be researcher-organized or participant- current blood marker levels that are already out-of-bounds organized. So far, researcher-organized studies have been per recommended levels, and that simple vitamin non-interventional studies organized by professional supplementation may be able to restore blood markers to researchers using crowdsourced cohorts or crowdsourced recommended ranges. The flagship study "Vitamin B-9 data as the input or research focus, for example studies and MTHFR variants" examines the potential role of organized by PatientsLikeMe and 23andMe. Participant- MTHFR genetic polymorphisms in vitamin B deficiency and homocysteine levels, and attempts to determine which results, and review their genetic data for the main vitamin supplement solutions are best from an individual D receptor gene polymorphism, rs10735810. Participants self-experiment regarding optimal supplement dosage, In the MTHFR gene (methylenetetrahydrofolate using Vitamin D Council guidelines recommending 1,000 reductase), two small variations in DNA (SNPs IU per 25 pounds of body weight as a starting dosage rs1801133/C677T and rs1801131/A1298C) may prevent (Cannell 2011). A person who weighs 150 pounds, for vitamin B9 (or folic acid) from being metabolized into its instance, would start taking 6,000 IU per day. A blood test active form (folate). Without this form of vitamin B, would test results after at least eight weeks, and then homocysteine can accumulate which may lead to supplement dosage can be modified. Each 1,000 IU nutritional deficiencies and symptoms associated with increase in vitamin D is estimated to produce an cardiovascular disease, diabetes, vascular damage, nerve approximate 10 ng/ml increase in vitamin D blood level. damage, and blood clots and pregnancy loss. Over 50% of Participant exclusions are those that have underlying the population may have some form of MTHFR mutation. kidney disease or a history or genetic risk of vitamin D The Vitamin B study protocol is to find individuals with problems. The study has 15 participants as of January MTHFR polymorphisms by collecting genotype data from volunteers who have used genetic testing services like 23andMe, try simple interventions like vitamin B Aging Studies
supplements available over-the-counter, and see if they Aging is an important systems biology and preventive work by asking participants to share results from blood medicine challenge where crowdsourced cohorts can make tests performed at commercial labs. Drug companies will a contribution with detailed self-tracking, data reporting, not do this type of study as there is little money to be made and longitudinal analysis. DIYgenomics has three aging- in over-the-counter treatments, but citizen science cohorts related studies. The first is "Aging: telomere length and can, and the results could be extremely useful to telomerase activation therapy" which investigates telomere individuals. The tools to do this kind of experiment, length (shorter telomeres are thought to cause earlier-onset looking at genomic information, measuring treatment disease and aging (Willeit 2011)) the efficacy of a natural results, and analyzing the data, are now cheap or free. product based remedy, TA-65, and potential linkage with The participant tasks are to review a list of Frequently telomerase-related TERT and TERC gene polymorphisms. Asked Questions (FAQs), submit data regarding MTHFR With aging, telomeres shrink by about 100 base pairs per genetic variants, and participate in at least 3 different 2 year. Research from 2009 Nobel Prize winner Elizabeth week trials involving taking over-the-counter vitamins and Blackburn and former Geron Chief Science Officer Cal measuring homocysteine levels with a blood draw each Harley has been used to develop a potential remedy in the time (blood tests are about $70 each). Participant form of the TA-65 telomerase activation therapy (Harley exclusions (self-selected) are those with known vitamin B 2011). More than one thousand individuals are currently deficiencies, those that might not be able to follow the taking TA-65. This study seeks to establish quantitative outlined vitamin protocol, or those who might have health and qualitative measures of the efficacy of TA-65 or problems related to high homocysteine levels. Two astragalus supplements, and whether personal genome MTHFR gene variants are reviewed, rs1801131 and profiles make a difference, specifically whether individuals rs1801133. Blood test and genomic profile data results are with TERT and TERC polymorphisms have shorter available to study organizers and all participants. telomeres to start with and therefore may be more likely to The methodology and pilot study results were published benefit from telomerase activation therapies. in December 2010 (Swan JOPM 2010). The study The participant tasks are to have telomere length continues in ongoing open enrollment and has 25 measured (from commercial vendors such as SpectraCell A second vitamin deficiency study, the "Vitamin D counter astragalus supplement or TA-65 (available from Study," examines vitamin D serum levels and tests how TA Sciences or Recharge Biomedical), journal product different supplementation doses and vitamin D receptor reactions and take a photo on a weekly basis, and re- gene polymorphisms may interrelate in attaining optimal measure telomere length at 6 months or 1 year. TERT and vitamin D levels. Low levels of Vitamin D have been linked to cancer risk (Laino 2011), and since some rs12696304, rs16847897, rs2293607, and rs610160. The individuals are unable to raise their blood levels despite study has 20 participants as of January 2012 and a 250- supplementation, genetic factors may be involved which member randomized cancer study is currently in design could be worthwhile to study (Jacobs 2010). (breast and prostate cancer chemotherapy patients in The participant tasks are to take a vitamin D3 supplement, obtain a 25-hydroxyvitamin D test, record remission will be sought) to examine the potential benefits Mental Performance Study
of TA-65 in rebuilding the immune system. There are many health studies which can be performed A second aging study, "Aging: risk reduction for without the cost and other drawbacks of blood tests, common aging conditions through monitoring and linking genomics with phenotype per online tests. intervention," is a longitudinal study of aging to establish DIYgenomics has such a study, "Processing Reality: personal baseline norms for 50 blood markers, their Impact of Dopamine Modulation on Memory Filtering," potential correspondence to 1,000 gene variants associated examining how genetic variants may be related to with aging, and to experiment with personalized dopamine processing in the brain and how this may impact intervention. The study provides an opportunity to apply the dozens of genome-wide association studies (GWAS) The brain is able to adapt to the unexpected using an which relate to general and specific conditions of aging in inbuilt network that makes predictions about the world and a comprehensive preventive medicine approach. Genomic monitors the results of those predictions. An area at the data is linked with corresponding measures of phenotypic front of the brain, called the orbitofrontal cortex, plays a biomarkers and interventions. The top twenty biological central role and studies have shown that patients with mechanisms of aging in GWAS include: neurodegenerative damage to this area confuse memories with reality and disease, osteoporosis, IGF-1/insulin signaling, lipoprotein continue to anticipate events that are no longer likely to metabolism, inflammation, immune system function, DNA happen (Elsevier 2011, Nahum 2011). This study seeks to damage repair, telomere length, transcription (ex: RNA determine if genetic variants in the dopamine processing editing), catabolism, mitochondrial health, cell cycle/stem pathway impact this process in normal, healthy volunteers. cell health, protein function, and blood operations. The top The study is being conducted in collaboration with the twenty phenotypic biomarkers of aging include: blood Center of Cognitive Neurorehabilitation at the Geneva pressure and hypertension, cholesterol (HDL, LDL, and University Hospital in Switzerland. The participant tasks triglycerides; LDL particle size), BMI, Framingham Risk are to complete a background demographic survey (10 Score, VO2 max, erythrocyte glycoslyation, telomere minutes), and a memory filtering task (30 minutes), which length, lymphocyte growth capability, and granulocyte shows a series of images and asks the viewer whether the strength. The participant tasks are to complete an annual image has been shown previously. Genetic variants in blood test (a comprehensive panel of approximately 50 markers available through DirectLabs ($79) or another (VAL158MET rs4680), DRD2 (rs1076560, rs2283265, source), and if willing, share the data with the cohort, and rs7131056), and SLC6A3 (rs40184, rs27048, rs27072). self-experiment with relevant interventions. 1,000 genetic Participant exclusions include those with psychological or variants are reviewed that have been linked to a variety of neurologic disorders (e.g.; bipolar disorder, schizophrenia, conditions of aging (Swan 2011). The study has 15 epilepsy, Parkinson's disease, prior stroke, traumatic brain injury, or dementia). The study has 27 participants as of A third aging study, "Retin-A: wonder cream for acne and wrinkles?" examines a potential connection between skin-related genetic variants and the widely-experienced Epistemology Study
negative side effects of using Retin-A skin care products. Retinoids (vitamin A compounds), particularly a tretinoin Self-experimentation studies conducted individually and in product Retin-A, are often used to treat acne and wrinkles. groups are emerging as an important complement to Remedies are available by prescription or over-the-counter. traditional clinical trials and other established mechanisms Retin-A peels or thins the outer layer of the epidermis, and of health knowledge generation. To validate crowdsourced thickens the layers below by stimulating collagen health research studies, it is important not only to conduct production. When first using a Retin-A product, some and report on these efforts in a scientifically-acceptable individuals experience a period of irritation with red, flaky, manner, but also to provide a philosophical context for peeling skin. This study investigates whether underlying understanding their role and impact. The epistemology genetic profiles might make a difference and predict project seeks to investigate, characterize, and provide a product response ahead of time. The participant task is to structure and context for knowledge derived through complete a 10 minute online survey of regarding experience with Retin-A products. Gene variants related to Participants are asked to complete an online questionnaire skin allergy and irritation are reviewed: rs1800629, (15-20 min) regarding self-experimentation activities in rs3793784, rs6661961, rs6700998, rs7538876, rs7927894, any area including health, time-management, stress- and rs8011 (Wadyka 2006). The study has 8 participants as reduction, or nutrition, exercise, sleep optimization, etc. Potential Future Studies
are searchable publicly in real-time based on deep DIYgenomics has several other potential future studies, attributes. One operator of interventional crowdsourced and has been designing studies at two levels, one for health research studies, DIYgenomics, has several ongoing on-demand citizen science participation, and one crowdsourced health research studies in open enrollment as for professional studies that have at least 100 participants of January 2012 in the areas of vitamin deficiency, aging, in a randomized, controlled, double blind format (Swan Personalized Medicine 2012). One such study is being designed Limitations of Participatory Health
collaboration with the California Walnut Commission, There are important challenges to the conduct of examining genetic polymorphisms and remedies including participatory health efforts. Perhaps less than 10% of statins, niacin, green tea, and walnuts in lowering LDL and individuals are interested in health, perceiving it as a raising HDL, and in increasing LDL particle size. Another deterministic area where the only incoming information study is in design to investigate calcinosis, which is an will be negative, and that health is a physician's aging disease of the arteries like atherosclerosis, in this responsibility, not one's own. It could take a while for case where calcium builds up unhealthily. In collaboration widespread responsibility-taking for health to arise as few with a Silicon Valley-based biotechnology firm, novel individuals may have the time, interest, or incentive to self- biomarkers for calcinosis will be sought, their potential manage their health. Crowdsourced cohorts may be too link to vitamin K metabolism polymorphisms investigated, slow to help with the more fundamental public health and vitamin K2 tested as a supplement intervention. A system problems of budget shortfalls, rising health care sleep study is contemplated in potential collaboration with costs, expected physician shortages, and the exorbitant cost Zeo to investigate genetic polymorphisms and sleep of bringing new drugs to market (currently estimated at performance. An extension of the Blueberry Study $1.3 billion (Gavura 2011)). Health social network (www.BlueberryStudy.com) is under discussion to participation is growing but slowly. Only some few examine genetic linkage as a follow-on to the 100,000 individuals have subscribed to personal genome demonstration of blueberry consumption leading to services since they launched in 2007, even though costs enhanced mental performance. Another potential study is have dropped to $99 with 23andMe. Consumers are still testing the efficacy of an anti-aging supplement (Juvenon, wondering about the meaning and use of the non- based on acetyl-L-carnitine and alpha lipoic acid) in a deterministic genetic information, and questions about the genetically-stratified cohort. Finally, there is a possible validity and utility of services persist as risk interpretations collaboration to investigate loving style preference and vary across services (e.g.; the risk of heart attack is high genomics in collaboration with the University of Pavia, according to 23andMe, but low according to deCODEme) Italy to extend a 'genetic loading on human loving styles' (Swan Gen Med 2010). Even if the cost of genotyping has study (Emanuele 2007) in a citizen science cohort. experimental measures remain prohibitively expensive. Conclusion
Future Implications
However, in the end, the bigger data, better tools, and intelligent community computing models of participatory The current era of internet-facilitated bigger data, better health might be the right solution at the right time. tools, and collective intelligence community computing is Participatory health efforts portend important benefits in accelerating advances in many areas ranging from artificial both size and speed. Studies can be carried out in cohorts intelligence to knowledge generation to public health. In of thousands instead of hundreds (Do 2011, Dufau 2011), the health sector, data volumes are growing with genomic, and it could be possible to apply findings from newly- phenotypic, microbiomic, metabolomic, self-tracking, and published studies immediately in crowdsourced cohorts. other data streams. Simultaneously, tools are proliferating Participatory health efforts and health social networks to allow individuals and groups to make sense of these data are providing direct value to participants, and have also in a participatory manner through personal health tracking become useful in a broader social context for clinical trial devices, mobile health applications, and personal electronic recruitment. Pharmaceutical companies and researchers medical records. Health community computing models are can recruit crowdsourced cohorts much more quickly and emerging to support individual activity and mass expediently than traditional cohorts, and at lower cost and collaboration through health social networks and with lower study drop-out rates. Self-organized studies by crowdsourced health research studies. Large groups, both health social networks and personal health collaboration patient registries and communities of healthy individuals, communities could help to surface interesting new findings, particularly related to preventive medicine. Not Jacobs, B. 2010. Vitamin D is a really interesting little molecule. everyone needs to be engaged, with the 'Wikipedia effect,' http://azcc.arizona.edu/node/1932. Accessed: January 18, 2012. 1% of individuals actively participating could create a public asset for all. A million social health collaborators Kamel Boulos, M.N., Resch, B., Crowley, D.N., et al. 2011. Crowdsourcing, citizen sensing and Sensor Web technologies for could benefit hundreds of millions of others. public and environmental health surveillance and crisis In the not too far future, it may be possible to see the management: trends, OGC standards and application examples. next levels of intelligent health community computing: personal health dashboards integrating multiple health data Laino, C. October 4, 2011. Low Vitamin D Levels Linked to streams, a continuous personal health information climate Advanced Cancers: Study Shows Three-Fourths of Cancer and body area network that makes unobtrusive behavioral Patients Have Low Levels of Vitamin D. WebMD Health News. Available suggestions, personal virtual coaches in the vein of Siri 2.0, http://www.webmd.com/cancer/news/20111004/low-vitamin-d- and an efficient health frontier of dynamic personalized levels-linked-to-advanced-cancers. Accessed: January 18, 2012. health recommendations and action-taking. Nahum, L., Simon, S.R., Sander, D., Lazeyras, F., Schnider, A. 2011. Neural response to the behaviorally relevant absence of anticipated outcomes and the presentation of potentially harmful Conflicts of Interest
stimuli: A human fMRI study. Cortex 47(2):191-201. The author is the founder of DIYgenomics. Shirky, C. 2008. Here Comes Everybody: The Power of Organizing Without Organizations. 1st edition. New York, NY: Penguin Press. References
Surowiecki, J. 2004. The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Cannell, J. September 28, 2011. Vitamin D supplementation. Business, Economies, Societies and Nations. New York, NY: http://www.vitamindcouncil.org/about-vitamin-d/how-to-get- Swan, M. 2010. Multigenic condition risk assessment in direct-to- your-vitamin-d/vitamin-d-supplementation. Accessed: January consumer genomic services. Genet Med 12(5):279-88. Swan, M., Hathaway, K., Hogg, C., McCauley, R., Vollrath, A. Do, C.B., Tung, J.Y., Dorfman, E., et al. 2011. Web-based 2010. Citizen science genomics as a model for crowdsourced genome-wide association study identifies two novel loci and a preventive medicine research. J Participat Med 2:e20. substantial genetic component for Parkinson's disease. PLoS Swan, M. 2011. DIYgenomics Aging GWAS database. DIYgenomics. Available at: http://bit.ly/rkq4s2. Accessed: Dufau, S., Duñabeitia, J.A., Moret-Tatay, C., et al. 2011. Smart phone, smart science: how the use of smartphones can Swan, M. 2012. Scaling crowdsourced health studies: the revolutionize research in cognitive science. PLoS One emergence of a new form of contract research organization. Personalized Medicine. Forthcoming. Elsevier. January 31, 2011. Snakes and Spiders: Revealing the Swan, M. 2012. Review of crowdsourced health research studies. Wiring That Allows Us to Adapt to the Unexpected. ScienceDaily http://www.sciencedaily.com/releases/2011/01/110131092332.ht Wadyka, S. November 30, 2006. The Thing About Retin-A: It http://www.nytimes.com/2006/11/30/fashion/30skin.html. Emanuele, E., Brondino, N., Pesenti, S., Re, S., Geroldi, D. 2007. Genetic loading on human loving styles. Neuro Endocrinol Lett 28(6):815-21. Willeit, P., Willeit, J., Kloss-Brandstätter, A., Kronenberg, F., Kiechl, S. 2011. Fifteen-year follow-up of association between Enriquez, J. June 2-4, 2010. As the future catches you. Presented telomere length and incident cancer and cancer mortality. JAMA at: 2nd Annual Consumer Genetics Conference Boston, MA. Available at: http://www.consumergeneticsshow.com. Accessed: January 18, 2012. Fallis, D. 2008. Toward an Epistemology of Wikipedia. J Am Soc Inf Sci Tech 59(10):1662-74. Gavura, S. April 14, 2011. What does a new drug cost? Science-based http://www.sciencebasedmedicine.org/index.php/what-does-a-new-drug-cost/. Accessed: January 18, 2012. Halevy, A., Norvig, P., Pereira, F. 2009. The Unreasonable Effectiveness of Data. IEEE Intell Syst 24:8-12. Harley, C.B., Liu, W., Blasco, M., et al. 2011. A natural product telomerase activator as part of a health maintenance program. Rejuvenation Res 14(1):45-56. Hood, L. 2011. Lee Hood. Nat Biotechnol 29(3):191.

Source: http://www.diygenomics.org/files/AAAI_2012_SWAN.pdf