You are on page 1of 3

Views & Comments

Engineering 2015, 1(3): 277–279


DOI 10.15302/J-ENG-2015075

Big Data for Precision Medicine


By Daniel Richard Leff and Guang-Zhong Yang*

ABSTRACT This article focuses on 2  Big data for precision medicine regard, hierarchical learning for EHR
the potential impact of big data analysis data, along with information regarding
to improve health, prevent and detect The electronic patient health record occupational, recreational, and muscu-
disease at an earlier stage, and (EHR) is a source of big data contain- loskeletal data, has been used to classify
personalize interventions. The role ing information regarding socio-demo- treatment options [7].
that big data analytics may have in graphics, medical conditions, genetics, It is expected that big data for health
interrogating the patient electronic health and treatments; yet the human ability can play an important role in pharma-
record toward improved clinical decision to process this data without effective cogenetics and stratified healthcare.
support is discussed. We examine decision support is finite. For health- Patients with a similar cancer subtype
developments in pharmacogenetics care, the goal is to provide a continually often respond differently when chal-
that have increased our appreciation learning infrastructure with real-time lenged with the same chemotherapeu-
of the reasons why patients respond knowledge production and to develop tics. For example, CYP2D6 is a poly-
differently to chemotherapy. We also a system that is preventative, predic- morphic gene associated with response
assess the expansion of online health tive, and participatory [2]. In order to to Tamoxifen [8], BRAF mutations
communications and the way in which achieve this goal, computer models are (Y472C) have been linked to Dasatinib
this data may be capitalized on in order required to help clinicians organize response in non-small cell cancer of the
to detect public health threats and the data, recognize patterns, interpret lung [9], and multiple gene signatures
control or contain epidemics. Finally, results, and set thresholds for actions. have been recently associated with the
we describe how a new generation of Examples of big data analytics for new response of rectal cancer to chemora-
wearable and implantable body sensors knowledge generation, improved clini- diotherapy [10]. Genomic instability
may improve wellbeing, streamline cal care, and streamlined public health is believed to be responsible for the
management of chronic diseases, and surveillance are already apparent. For observed diversity of drug response.
improve the quality of surgical implants. example, the EHR has been successfully Recent efforts have focused on expos-
mined for post-market surveillance of ing the complex interplay of genom-
KEYWORDS big data, biosensors, medications and improved pharmaco- ics and chemotherapeutic sensitivity,
body-sensing networks, implantable vigilance. resistance, and toxicity [11–13]. For
sensors, clinical decision support In the United Kingdom, the National example, the Cancer Genome Atlas re-
systems, pharmacogenetics, mHealth Health Service intends to be paperless search network has launched the Pan-
by 2018. The EHR will provide an inte- Cancer project [11] to analyze multiple
1 Introduction gral resource for future clinical decision tumor types and molecular aberrations
support systems (CDSSs) that may over- in cancer types, and to enable scientists
The complexity, diversity, and rich con- come human limitations in data com- to discover new aberrations. Similarly,
text of data being generated in health- prehension and multitasking. CDSSs several projects such as the Cancer Cell
care are driving the development of big are already being developed to assess Line Encyclopedia [12] and the Genom-
data for health [1]. Volume, velocity, and improve protocol adherence [3], for ics of Drug Sensitivity in Cancer [13]
variety, veracity, variability, and value medication reminders [4], to improve are generating large genomic databases
are the “V’s” of big data, and these are screening [5], and to predict hospital re- to specifically interrogate links between
encapsulated in the inherent challenges admission [6]. Certain conditions have genomic biomarkers and drug sensitiv-
of biomedical and health informatics. multiple treatment options and CDSSs ity in hundreds of cancer cell lines. As
Effective ways of tackling these chal- are being developed to help clinicians evidence of the ability to leverage large
lenges would pave the way for more optimize strategies. For example, prob- pharmacogenetic databases to predict
intelligent healthcare systems focused lems associated with anterior cruciate drug sensitivity, recent data suggests
on prevention, early detection, and per- ligament (ACL) may be treated by phys- that computational algorithms for pre-
sonalized treatments. iotherapy, medicine, or surgery. In this dicting drugs for individual cell lines

The Hamlyn Centre, South Kensington Campus, Imperial College London, London SW7 2AZ, UK
*Correspondence author. E-mail: g.z.yang@imperial.ac.uk

© The Author(s) 2015. Published by Engineering Sciences Press. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)

www.engineering.org.cn Volume 1 · Issue 3 · September 2015 Engineering 277


Views & Comments
can be improved based on genomic Ebola virus detection (EVD) captured enable wound infections to be detected
profiles and drug-response data [14]. progressive increases in the number of and treated at a sub-clinical stage, pre-
Future work will involve testing the re- tweets discussing EVD case identifica- venting fulminant sepsis. White blood
sults of such algorithmic predictions on tion in Nigeria occurring at least three cells and neutrophil counts could be
tumor response and toxicity in patients days prior to the news alert and seven continually monitored in patients un-
undergoing chemotherapy. days before the official Centre for Dis- dergoing chemotherapy cycles such
One of the major contributors of ease Control warnings. Finally, several that the earliest sign of neutropenia
dramatic change in healthcare is the investigators are exploiting the poten- could be coupled with granulocyte
upsurge in communication instigated tial of social media toward behavioral stimulation to mitigate sepsis.
by social media. One recent estimate change and improvements in health-
suggests over one billion healthcare care, including targeted interventions 3 Conclusions
tweets, reflecting the enormity of dia- for developing countries for the pur-
logue between healthcare providers, pose of global health promotion [18]. In conclusion, the sources and com-
patients, organizations, and third par- Another important driver for big putational techniques for big data are
ties. Approximately 20% of patients data is the widespread deployment rapidly increasing and the possible uses
with chronic healthcare conditions of sensing technologies. There has for improving health and well-being
such as diabetes, cardiovascular dis- been increasing interest in wearable are manifold. For healthcare, the goal
ease, and cancer, go online to actively and implantable sensing owing to is to provide a continually learning in-
seek others and share experiences of improvements in technologies, includ- frastructure with real-time knowledge
related conditions; such actions are ing sensors with enhanced wireless production and to develop a system
creating patient communities on so- communications that have increased that is preventative, predictive, and
cial media sites such as Twitter and bandwidth and improved microelec- participatory. Big data analysis clearly
Facebook. Social media is providing tronics. As a result of these develop- has tremendous potential to improve
new avenues for investigators to enroll ments, continuous, multimodal, and healthcare and transform the health
patients in research, and for patients context-aware sensing is now feasible of populations. However, successfully
to engage in sharing their health data. [19]. Simultaneously, advances in sen- exploiting this potential will depend on
For example, in TuAnalyze [15], a joint sor miniaturization, embodiment, bio- solving challenges associated with data
initiative between TuDiabetes and the fouling mitigation, and microelectronic privacy, security, ownership, and gov-
Boston Children’s Hospital, diabetics fabrication schemes have improved the ernance.
can monitor, evaluate, and share their versatility and reliability of implant-
results while actively participating in able biosensors. Episodic monitoring is Compliance with ethics
research on diabetes. Arguably one of being replaced, therefore, with continu- guidelines
the most interesting applications of big ous sensing and parallel improvements
data analytics is in the ability to pre- in integrated care, toward personal- Daniel Richard Leff and Guang-Zhong
dict and track major outbreaks in order ized and stratified healthcare. Patients Yang declare that they have no conflict
to improve public healthcare resources at high risk of critical events, such as of interest or financial conflicts to dis-
and the dissemination of healthcare arrhythmias, following myocardial in- close.
messages to victims using social me- farction may benefit from continuous
dia. Predictions of serious healthcare monitoring of blood pressure, pulse, References
emergencies such as exacerbations of and cardiac rhythm such that arrhyth-
asthma can be better predicted in mod- mias can be detected in near real time 1. J. Andreu-Perez, C. C. Poon, R. D. Merrifield, S.
els that combine social media analysis and signals sent to a smartphone for T. Wong, G. Z. Yang. Big data for health. IEEE J.
with environmental data. Unlike con- ulterior processing [20]. Similarly, pa- Biomed. Health Inform., 2015, 19(4): 1193–1208
ventional models that base predictions tients with chronic health conditions 2. L. Hood, M. Flores. A personal view on systems
with a two-week lag, Ram et al. [16] such as diabetes may benefit from medicine and the emergence of proactive P4
were able to create accurate predic- implantable microsystems for continu- medicine: Predictive, preventive, personalized
tion models of the volume of daily ous monitoring [21], toward improved and participatory. New Biotechnol., 2012, 29(6):
emergency-department visits for acute long-term glucose control. Neuro- 613–624
asthma (volume defined as low, mod- modulation, such as with implantable 3. J. G. Klann, V. Anand, S. M. Downs. Patient-
erate, or high) using Twitter activity, spinal stimulators, is having an impact tailored prioritization for a pediatric care deci-
Google searches, and air-quality data in the management of chronic back sion support system through machine learning.
[16]. Similarly, for major public epi- pain [22]. Implantable sensors that J. Am. Med. Inform. Assoc., 2013, 20(e2): e267–
demics such as Ebola, data can be col- monitor the axial load on an individual e274
lected and analyzed to support early subject’s spine [23] may lead to person- 4. B. G. Nair, S. F. Newman, G. N. Peterson, W.
warning systems for epidemic trends alized orthopedic prostheses. In future, Y. Wu, H. A. Schwid. Feedback mechanisms
and to deliver messages for healthcare it is hoped that sensors applied in close including real-time electronic alerts to achieve
education interventions [17]. For ex- proximity to the site of operative in- near 100% timely prophylactic antibiotic admin-
ample, Odlum et al. [17] demonstrated tervention, which monitor local hemo- istration in surgical cases. Anesth. Analg., 2010,
that analyzing tweet activity around dynamics or tissue ionic content, will 111(5): 1293–1300

278 Engineering Volume 1 · Issue 3 · September 2015 www.engineering.org.cn


Views & Comments
5. K. B. Wagholikar, et al. Clinical decision sup- 11. Cancer Genome Atlas Research Network; J. N. Househ, O. Da’ar, K. Alsurimi. A social media
port with automated text processing for cervical Weinstein, et al. The Cancer Genome Atlas Pan- campaign to promote breastfeeding among
cancer screening. J. Am. Med. Inform. Assoc., Cancer analysis project. Nat. Genet., 2013, 45(10): Saudi women: A web-based survey study. Stud.
2012, 19(5): 833–839 1113–1120 Health Technol. Inform., 2015, 213: 247–250
6. J. Futoma, J. Morris, J. Lucas. A comparison of 12. J. Barretina, et al. The Cancer Cell Line Ency- 19. G. Z. Yang. Body Sensor Networks. 2nd ed. Lon-
models for predicting early hospital readmis- clopedia enables predictive modelling of anti- don: Springer-Verlag, 2014
sions. J. Biomed. Inform., 2015, 56: 229–238 cancer drug sensitivity. Nature, 2012, 483(7391): 20. F. Rincón, P. R. Grassi, N. Khaled, D. Atienza,
7. K. Mei, J. Peng, L. Gao, N. N. Zheng, J. Fan. 603–607 D. Sciuto. Automated real-time atrial fibrillation
Hierarchical classification of large-scale patient 13. M. J. Garnett, et al. Systematic identification of detection on a wearable wireless sensor plat-
records for automatic treatment stratification. genomic markers of drug sensitivity in cancer form. In: Proceedings of the Annual International
IEEE J. Biomed. Health Inform., 2015, 19(4): 1234– cells. Nature, 2012, 483(7391): 570–575 Conference of the IEEE Engineering in Medicine
1245 14. J. Sheng, F. Li, S. T. Wong. Optimal drug predic- and Biology Society. Piscataway, NJ: IEEE Service
8. M. P. Goetz, et al. The impact of cytochrome tion from personal genomics profiles. IEEE J. Center, 2012: 2472–2745
P450 2D6 metabolism in women receiving ad- Biomed. Health Inform., 2015, 19(4): 1264–1270 21. M. M. Ahmadi, G. A. Jullien. A wireless-
juvant tamoxifen. Breast Cancer Res. Tr., 2007, 15. Anon. TuAnalyze is here! 2010-05-19. http:// implantable microsystem for continuous blood
101(1): 113–121 www.tudiabetes.org/forum/topics/tuanalyze- glucose monitoring. IEEE Trans. Biomed. Circuits
9. B. Sen, et al. Kinase-impaired BRAF mutations is-here Syst., 2009, 3(3): 169–180
in lung cancer confer sensitivity to dasatinib. 16. S. Ram, W. Zhang, M. Williams, Y. Pengetnze. 22. J. Padwal, M. M. Georgy, B. A. Georgy. Spinal
Sci. Transl. Med., 2012, 4(136): 136ra70 Predicting asthma-related emergency depart- cord stimulators in an outpatient interventional
10. T. Watanabe, T. Kobunai, T. Akiyoshi, K. Mat- ment visits using big data. IEEE J. Biomed. Health neuroradiology practice. J. Neurointerv. Surg.,
suda, S. Ishihara, K. Nozawa. Prediction of Inform., 2015, 19(4): 1216–1223 2014, 6(9): 708–711
response to preoperative chemoradiotherapy 17. M. Odlum, S. Yoon. What can we learn about 23. M. K. Moore, S. Fulop, M. Tabib-Azar, D. J.
in rectal cancer by using reverse transcriptase the Ebola outbreak from tweets? Am. J. Infect. Hart. Piezoresistive pressure sensors in the
polymerase chain reaction analysis of four Control, 2015, 43(6): 563–571 measurement of intervertebral disc hydrostatic
genes. Dis. Colon Rectum, 2014, 57(1): 23–31 18. S. Bahkali, N. Alkharjy, M. Alowairdy, M. pressure. Spine J., 2009, 9(12): 1030–1034

www.engineering.org.cn Volume 1 · Issue 3 · September 2015 Engineering 279

You might also like