Download big data: current status and future directions agenda

Baltimore Convention Center
Baltimore, MD
Saturday, May 6, 2017
8:30am-4:30pm
BIG DATA: CURRENT STATUS AND FUTURE DIRECTIONS AGENDA
Organizers: Michael F. Chiang, MD, Anne L. Coleman, MD, PhD, FARVO and Seth Blackshaw, PhD
Big Data is one of the most frequently used terms in the press. This course will inform participants about what Big
Data actually is, what are some of the Big Data sets available in vision research, some of the analytics used on Big
Data and some of the potential applications of Big Data. Speakers and topics are subject to change.
8:30-9:00am
Introduction and Welcome
Anne L. Coleman, MD, PhD, FARVO, Jules Stein Eye Institute, UCLA
A general overview of Big Data, its definition, uses, benefits and liabilities will be presented.
9:00-9:20am
Tapping into Health Care Claims Databases to Learn About Ocular Diseases
Joshua D. Stein, MS, MD, Kellogg Eye Center, University of Michigan
This talk will describe some of the common large health care claims datasets that researchers can access to study patients
with ocular diseases. We will go over some of the strengths of performing research using these databases over other sources
as well as limitations of working with these resources. Examples of analyses performed using health care claims data will be
described so the audience can get a flavor for the types of studies and research questions that can be answered using these
data sources.
9:20-9:40am
Mining electronic health record (EHR) data
Michael F. Chiang, MD , Casey Eye Institute, Oregon Health & Science University
This presentation will discuss practical applications of mining EHR data. These will include clinical research applications of
large-scale EHR data, as well as operational applications of EHR data for improving real-world clinical workflow.
9:40-10:00am
Clinical Registries and "Big Data"
Paul P. Lee, JD, MD, FARVO, Kellogg Eye Center, University of Michigan
Clinical registries for eye care have existed since the post US Civil War period's trauma registry. In the last 40 years, several
eye registries have been developed, covering not only trauma and ocular medications but also focused clinical issues such as
cataract and cataract surgery. The Swedish National Cataract Registry is one excellent example. The newest registries, such as
the American Academy of Ophthalmology's IRIS and the American Optometric Association's MORE are even more
comprehensive, expanding registry coverage to the broad range of eye care and eye diseases that affect human health. The
advent and growth of these comprehensive registries will enable us to build upon the quality and safety work that has been
done for cataracts and ocular medications to the broad range of eye conditions as well as to begin to explore detailed
relationships among many factors that have heretofore been difficult to understand due to practical limitations such as sample
size and the necessity of chart abstraction. As such, clinical registries are both a powerful tool and exemplar of the promise of
"big data" to improve vision and lives.
Page 1 of 4
10:00-10:30am
Panel discussion
Michael F. Chiang, MD, Paul P. Lee, JD, MD, FARVO and Joshua D. Stein, MS, MD
10:30-10:45am – Break
10:45-11:05am
Visualizing “Big data”
Aaron Y. Lee, MD, University of Washington
One of the great challenges in Big Data is being able to visualize large volumes of data in order to understand new insights and
trends. This presentation will feature several visualizations aiming to help researchers slice through different dimensions of
data. Many of these visualizations are interactive allowing for deeper hypothesis generation by being able to visualize data.
11:05-11:25am
Applications of machine learning and deep learning
Adnan Tufail, FRCOphth, MD, Moorfields Eye Hospital
The current explosion of activity in machine learning is remarkable. A non-technical overview of machine learning aimed at
clinicians will be given and the potential to capturing the full value of big data discussed. We will discuss the importance of
validation of machine learning algorithms. Current and emerging applications of machine learning in ophthalmology, such as
the grading of diabetic retinopathy and diagnosis of AMD on retinal images, will be given.
11:25-11:45am
Big data and ophthalmic genetics
Michael B. Gorin, MD, PhD, FARVO, Jules Stein Eye Institute, UCLA
There is considerable interest in multiple areas of medicine to use big data from electronic health records and genetic data.
There are a number of opportunities and challenges that confront the use of this approach to address key issues in ophthalmic
genetics and foster the application of molecular genetics in ophthalmic care. A number of projects have successfully used
selective data from large data sets to test for associations with genetic variants/markers. We will review recent efforts in
medicine and ophthalmology to employ institutional and consortia biobanks and EMR data in conjunction with genome wide,
next generation sequencing and/or genotyping to elucidate genetic components of disease. This talk will highlight the
challenges of defining phenotypes for disease using ICD-9, ICD-10 codes and natural language processing, harmonizing
ocular-related measures that are not standardized and determined by multiple methods and devices, dealing with the laterality
of diagnoses and treatments, missing and incomplete data especially with records that rely on varying inclusion of information
from different medical specialties, and the lack of adequate family historical data. Realization of the potential of big data for
ophthalmic genetics will ultimately require a concerted effort to address the consistent acquisition of data for clinical records.
11:45am-12:15pm
Panel discussion
Michael B. Gorin, MD, PhD, FARVO, Aaron Y. Lee, MD, Adnan Tufail, FRCOphth, MD
12:15-1:15pm – Lunch
1:15-1:35pm Epigenetic changes associated with age-related macular degeneration
Jiang Qian, PhD, Wilmer Eye Institute, Johns Hopkins School of Medicine
Age-related macular degeneration (AMD) is the leading cause of severe vision loss in people over 60. Genetic association
studies have successfully identified genetic variants associated with the disease. However, the genetic variants only explain
40-70% of disease variability, suggesting that factors other than the genetics might contribute to the disease process. Here
we attempted to identify the epigenetic factors associated with AMD. We performed epigenetic profiling on retina and RPE
Page 2 of 4
layers from postmortem eyes. By comparing the profiles between AMD and normal eyes, we discovered massive and
consistent changes in chromatin structures in both retina and RPE. Some AMD susceptibility genes were found to have
significant chromatin changes. The epigenetic changes were related to the changes in transcription factor binding in the
genome and the alterations in gene expression levels. Our study is the first systematic assessment of epigenetic changes
associated with AMD and suggests that epigenetics might play a critical role in AMD.
1:35-1:55pm
Transcriptomics
Salil Anil Lachke, PhD, University of Delaware
The identification of genes linked to eye development and its associated defects presents a formidable challenge. In the recent
past however, genome-level transcript profiling (transcriptomics) has become increasingly feasible due to technologies such as
microarrays and RNA sequencing. Their application to interrogate specific eye tissues and cell types holds high promise to
impact ocular gene discovery. However, global gene expression profiling has brought with it new challenges, such as parsing
through the large amounts of data to prioritize select candidates for further analysis. Toward this goal, the development of
strategies such as “in silico subtraction” has led to the construction of a systems-based web-resource tool called iSyTE
integrated Systems Tool for Eye gene discovery). In this presentation, I will discuss the impact of iSyTE and eye-tissue
transcriptomics on identification of genes linked to lens biology and cataract, and outline the ongoing efforts for extending this
approach to other ocular tissues such as the retina and the cornea.
1:55-2:15pm
A standardized toolbox of highly specific, immunoprecipitation-grade antibodies against human
transcription factors and associated proteins
Heng Zhu, PhD, Johns Hopkins School of Medicine
To better understand what makes a cell function normally and what may go awry in disease, we need better tools and
resources, such as renewable protein capture reagents. Monoclonal antibodies (mAbs) are indefinitely renewable and possess
consistent epitope recognition properties, making them more desirable for most basic and clinical applications. To extend
these efforts, the NIH Common Fund in 2010 initiated the Protein Capture Reagent Program (PCRP), an effort that was aimed
at generating highly-specific, extensively characterized and readily available renewable affinity reagents that target a broad
range of human transcription factors and associated proteins. Our JHU-CDI team has developed an integrated production and
validation pipeline for high-throughput and low-cost generation of mouse monoclonal antibodies (mAbs) that recognize their
targets with both high specificity and affinity. By using human proteome microarray validation as a first screening step, we
show that we can rapidly enrich for ultraspecific, high-affinity mAbs that perform well in downstream applications -- particularly
immunoprecipitation and ChIP-Seq. We demonstrated the flexibility of protein microarray-based validation of antibody
specificity, and show that it can be used to identify mAbs that selectively recognize their intended target in native and/or
denatured conformation. We also demonstrate that this approach can be used to rapidly and efficiently identify proteins that
cross-react with commercially available mAbs, and can thus be used as a standardized platform for analysis of antibody
specificity. We have used this pipeline to generate a total of 1,361 mAbs that target a total of 731 unique human
transcription factors and associated proteins, and which worked for one or more common research applications, including
immunoprecipitation, immunoblotting and ChIP-Seq. We have made these mAbs readily available to the research community
through both the Developmental Studies Hybridoma Bank (DSHB) and various commercial suppliers. These reagents serve a
standardized toolbox for biochemical analysis of transcriptional regulatory mechanisms in human cells
2:15-2:30pm – Break
Page 3 of 4
2:30-2:50pm
Parsing retinal development and disease at the single cell level
David Cobrinik, MD, PhD, Keck School of Medicine, USC
Studies of retinal developmental and disease often involve large scale genomic, transcriptomic, and proteomic analyses.
However, interpretations of such studies can be compromised by the admixture of cells of different types and in different
developmental or pathogenic stages. For example, RNA-Seq analysis of whole retina gives a vastly different transcriptional
profile from that of a single purified cell type, and RNA-Seq of a specific cell type in a developing retina may differ from that of
the different cell sub-populations in a different developmental stages. Similarly, a pathogenic process may be obscured if
affected cells are mixed together with the bulk unaffected population. Because of these considerations, understanding of
normal and pathogenic developmental processes within heterogeneous tissues can be improved through analyses of single
cells, followed by grouping of cells with similar properties into individual states, and pseudotemporally ordering the different
cellular states into a developmental progression.
While single cell ‘omics’ analyses provide unprecedented resolution they also generate far larger data sets than standard
omics analyses, as well as challenges for ‘big data’ interpretation and integration. In this presentation I will share our
experience in modeling and comparing retinal development in human, mouse, and embryonic stem cell (hESC) -derived retina,
as well as the developmental aberrations that occur in response to a disease perturbation. The studies will focus on RNA-Seq
of developing cone photoreceptor precursors, which appear to have substantially different features in human, mouse and
hESC retinal tissue. They will also describe changes in the cone precursor developmental trajectory elicited by depletion of the
RB1 gene, which is thought to reflect early steps of retinoblastoma tumorigenesis. These studies will illustrate how the
application of big data approaches to single cell biology can provide novel insights into retinal development and disease.
2:50-3:10pm
Single-cell RNA-Seq: a window into retinal cell specification and diversification
Seth Blackshaw, PhD, Johns Hopkins School of Medicine
Cell type specification in the mammalian retina occurs through a process of temporal patterning, in which individual progenitor
cells successively acquire and lose the competence to generate specific postmitotic cell types. The genes that underlie these
changes in competence, and extent to which retinal progenitors are heterogeneous in the developmental potential, are still
unclear. To address these questions, we have performed large-scale profiling of gene expression in individual cells of the
developing retina using RNA-Seq. These data identify canididate regulators of developmental competence in non-neurogenic
progenitors, early markers of lineage commitment in neurogenic progenitors, and temporally dynamic patterns of gene
expression in postmitotic precursor cells. Potential regulatory relationships among these genes will be discussed.
3:10-3:40pm
Panel discussion
Seth Blackshaw, PhD, David Cobrinik, MD, PhD, Salil Anil Lachke, PhD, Jiang Qian, PhD, Heng Zhu, PhD
3:40-4:15pm
Panel discussion - Brainstorming on next steps and global collaborations
All faculty
4:15-4:30pm
Closing remarks
Michael F. Chiang, MD
This presentation will summarize discussion from this session.
Page 4 of 4