The Applicability of Big Data in Personalized Medicine

Sequencing data, imaging data, clinical data, and data from cancer registries presents several challenges in terms of heterogeneity and a lack of precision. Big analytics serve several purposes including identifying complex data representations, optimization of algorithms, traceability, and predictability. Big data analytics in healthcare context solve several problems such as interpreting findings in personalized medicine and identifying care patterns.

Geography: Global; Focus Area: Healthcare data analysis for precision medicine

Biomedical research that is driven by omics approaches is the future of personalized medicine. Cirillo and Valencia (2019) discuss how cloud infrastructures will serve as foundations of data analysis for personalized medicine through the application of deep learning and machine learning techniques. 

The availability of large-scale clinical and molecular data is a significant challenge for data analysis and interpretation (Cirillo & Valencia, 2019). The success of personalized medicine will depend on the efficiency of data-driven systems to generate mechanistic models for the design of clinical procedures.

International Data Sources for Personalized Medicine

The availability of data has been driven by several factors (Cirillo & Valencia, 2019):

• High throughput genome sequencing at low costs and increased access

• Availability of information extracted from imaging data and clinical records

• Massive efforts in community-based data collection through global initiatives such as Global Alliance for Genomics and Health (GA4GH), big data to knowledge (BD2K) initiative, and ELIXIR research infrastructure

• Domain-specific initiatives such as International Rare Disease Research Consortium (IRDiRC) and Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE)

• Genomic and epigenetics data from Human Genome Diversity Project (HGDP), Encyclopedia of DNA Elements (ENCODE), Global Network of Personal Genome Projects (PGP), International Human Epigenome Consortium (IHEC), and NIH Roadmap Epigenomics Mapping Consortium (Roadmap)

• Cancer registries including International Cancer Genome Consortium (ICGC) and the Cancer Genome Atlas (TCGA)

Promotion: Genomics and Personalized Medicine: What Everyone Needs to Know® Paperback – Illustrated, 24 March 2016 by Michael Snyder (Author)

   

Challenges Related to Big Data Analytics

Big data is multi-spectral, heterogeneous, incomplete, and imprecise. Analytics for big data require capabilities to interpret complex data representations, algorithm optimization, modeling, and increased computational power (Cirillo & Valencia, 2019). Analytics systems for unstructured data need to support prediction, traceability, and decision support, and be able to identify patterns of care. Functional characterization of medical data often proceeds at a slower pace when compared to data gathering, and increased challenges are encountered when multi-omics data has to be combined with phenotypic patient data to arrive at interpretations in personalized medicine. 

Multiple Data Types Available from Big Data

High volume, sensitive, and complex patient data, such as neuroimaging and genomics data is growing by petabytes annually. This data is accompanied by metadata and quantitative measurements. Heterogeneous data requires scalable and integrative platforms. Common data types available to explore frontiers in personalized include (Cirillo & Valencia, 2019):

• Symptom descriptions (unstructured data)

• ICD codes (structured data)

• Whole genome sequencing data (WGS data)

• Whole exome sequencing data (WEX)

• Transcriptome sequencing data

• Proteome and interactome profiling data

• EHR patient data

• Imaging data

• Multi-omics data

• Data from implants and wearable devices

• Patient generated health data (PGHD)

• Real-time data from sensors and biometric devices

• Mobile-tracked data such as treatment history and lifestyle choices

Promotion: Visualizing Health and Healthcare Data: Creating Clear and Compelling Visualizations to "See how You're Doing" Paperback – Import, 17 December 2020 by K Rowell (Author)

   

Heterogeneity in Big Data Applications

Big data has been applied to multiple disciplines in medicine including cancer and rare disease research, biomarker and drug development, diabetes and cardiovascular disease research, and neurodegeneration studies. Deciphering distributed data collaborative effort (Cirillo & Valencia, 2019). Some examples of personalized medicine projects are personalized brain models of the Human Brain project undertaken by the European Commission. Likewise, the Human Genome project forms the basis of large-scale biomedical projects, and the International Personalized Medicine consortium (PerMed) facilitates the study of mechanisms of hematopoiesis and epigenetics. Data handling is regulated by ethical framework for anonymous processing, such as General Data Protection Regulation (EU). The effectiveness of biomedical data rests on certain characteristics beyond security, such as accessibility, ease of finding data, interoperability, and reusability (FAIR).

Complex Data Analysis

Data analysis requires high performance computing (HPC) for the extraction of knowledge from big data. Computing capabilities such as streaming data analysis, data-intensive simulations, machine learning, deep learning, and neural networks for high dimensional data, and the investigation of multi-view data through data-driven integrated workflows promote complex data analysis (Cirillo & Valencia, 2019). 

Deep learning methods (convolutional / neural networks and recursive neural networks) have interesting applications in prediction and classification such as assessing disease risk, hospital outcome prediction, medical image analysis, prediction of promoter-enhancer interaction and transcription factor binding site, metagenome classification, drug design, and epileptic seizure prediction (Cirillo & Valencia, 2019). 

Therefore, big data can be transformational in healthcare and biomedical research. Big data types and analytics are highly challenging and require advanced machine learning methods to arrive at effective inferences. Continued research and development in these areas will lead to innovative solutions in the field of personalized medicine. 

Promotion: Big Data in Healthcare: Statistical Analysis of the Electronic Health Record Hardcover – 28 February 2020 by Farrokh Alemi (Author)

   

Keywords

healthcare data, data analysis, big data analytics, heterogeneity, complex data patterns, precision medicine, big data, big data frameworks, personalized medicine

References

Cirillo, D., & Valencia, A. (2019). Big data analytics for personalized medicine. Current Opinion in Biotechnology, 58, 161–167. https://doi.org/10.1016/j.copbio.2019.03.004

Big Data Analytics for Building Healthcare Frameworks

Structured and unstructured healthcare data is important for monitoring disease and the health of populations, and facilitating decision-making in precision medicine. Researchers define specific characteristics of healthcare big data to effectively utilize it for multiple purposes. In this connection, big data frameworks allow the extraction of continuous data, data analysis, and personalizing data. As a result, high-performance systems can convert large data volumes into actionable insights for improving the health of populations and improving care quality at reduced costs.

Geography: Global; Focus Area: Healthcare big data for improving care and reducing costs

Structured and Unstructured Healthcare Big Data

Healthcare big data is recorded and stored as structured data (disease terminology, symptom and diagnostic information, patient admission history, laboratory results, services availed, drug and billing information), semi-structured data (data generated from sensors and devices for monitoring patient behavior), and unstructured data (clinical letters, prescriptions, discharge summary, biomedical literature). Healthcare data serves multiple purposes (Palanisamy & Thirunavukarasu, 2019):

• Analysis of healthcare leads to important information for use by stakeholders for efficient clinical decision-making, population health management, and disease surveillance. 

• The development of healthcare frameworks using big data assists in personalized care and precision medicine. 

• Understanding patterns in data leads to novel healthcare solutions such as the patient-centric model and disease-centric model.

Promotion: Data Analytics: The Ultimate Beginner's Guide to Data Analytics Paperback – Import, 23 July 2019 by Edward Mize (Author)

   

Utilizing Healthcare Big Data

To achieve a purpose, healthcare data needs to be curated using domain specific tools and techniques. Complex, interdependent medical data needs to be simplified, interconnections between health features identified, and target attributes need to be selected for healthcare analytics (Palanisamy & Thirunavukarasu, 2019). Data streamed from devices must be integrated to feed into predictive models, medical image formats must be constructed and compressed. Big data tools are also useful in data enrichment activities. 

Redefining the characteristics of big data as “silo, security, and variety” in place of “volume, velocity, and variety” helps understand the context of big data in healthcare (Palanisamy & Thirunavukarasu, 2019). “Silos” refer to legacy databases containing public healthcare information in hospitals. “Security” refers to the care required to maintain healthcare data. “Variety” refers to the multiple formats of healthcare data including structured, unstructured, and semi-structured data.

Big Data Frameworks for Healthcare Data

Several big data frameworks have been proposed in research studies (Palanisamy & Thirunavukarasu, 2019):

• A framework consisting of data source layer, transformation layer, big data platform layer, and analytical layer takes internal and external data sources in multiple formats and extracts, transforms, and loads data using staging techniques (middleware and data warehousing). Further specific operations are performed by the Hadoop distributed file system (HDFS) using map-reduce. Finally, “querying, reporting, online analytical processing, and data mining” is performed in the analysis stage.

• Another patient-centric personalized healthcare framework used a collaborative filtering approach to using patient similarities to predict personalized disease risk. The system was designed to handle ICD codes.

• Another framework used vital signs extracted from accelerometers (as continuous time series data) to provide healthcare services. Motion data, respiration data, and ECG signals were used to feed into an interoperable system designed using Hadoop and Map-Reduce algorithm.

Promotion: Actionable Intelligence - A Guide to Delivering Business Results with Big Data Fast! Hardcover – Import, 24 October 2014 by KB Carter (Author)

   

In addition, several disease specific frameworks have been designed such as for AIDS, in policy enforcement, authentication and authorization (secure health systems), decision-support for practitioners, public health services, and drug design (Palanisamy & Thirunavukarasu, 2019).  

In effect big data frameworks are designed to meet specific healthcare objectives through the use of standard guidelines. Given the continuous data growth and evolution, integration is a primary means of aggregating data from multiple sources for transforming data, tracking procedures and metrics, and using business rules to build prescriptive and predictive models. Powerful, high-performance systems can query large volumes of data and transform it into “actionable knowledge” and identifying trends and deviations to improve public health and clinical health delivery. Insights gained from the data analysis decrease costs and improve care delivery. 

Keywords

healthcare data, security, population health, cost cutting, patient-centric, healthcare costing, improving care quality, variety, big data, big data frameworks, silo, personalized medicine

References

Palanisamy, V., & Thirunavukarasu, R. (2019). Implications of big data analytics in developing healthcare frameworks – A review. Journal of King Saud University - Computer and Information Sciences, 31(4), 415–425. https://doi.org/10.1016/j.jksuci.2017.12.007