Guest post by Christian Hall, Product Marketing Manager, Stardog
Breaking from the Crowd in Pharmaceutical R&D
The pharmaceutical industry is in the midst of unprecedented changes. In the past decade, it has become increasingly difficult to find viable drug candidates capable of making it through clinical trials. In essence, all the low-hanging fruit has been picked. At the same time, the cost of drug development has skyrocketed, with the average therapy now costing $1 billion to bring to market.
These dual but related forces are decreasing ROI, particularly in research and development (R&D), which accounts for the lion’s share of pharmaceutical companies’ costs. And with each passing day a drug therapy is not brought to market, a company stands to lose $15 million in potential revenue.
The only way for pharmaceutical companies to counteract these strong headwinds is to extract better insight from their existing data and to develop systems that allow for more rapid decision making. Making better use of data up front will help these companies access data from across silos and then more quickly decide which therapies to pursue in the drug development process.
The Pharmaceutical Data Landscape
Pharmaceutical R&D is a highly data-intensive undertaking. Pharmaceutical companies use demographic data, patient clinical data, genomic data, bioinformatics data, and trial data, amongst other sources, to identify possible compounds and combinations of compounds to investigate for drug therapies. This data is:
· Highly dispersed, with datasets sourced across departments and geographies
· Complex, with metadata schema and structures varying dramatically between labs
· Governed by a variety of data standards (CDC, WHO, CDISC)
· Structured, unstructured, and semi-structured and from proprietary, public, and third-party sources
Pharmaceutical companies have had a difficult time integrating the knowledge contained across different labs. Typically, an individual disease category maintains its own database and is disconnected from databases in other disease categories or in other stages of the drug development process. So, even when research is complementary, it is kept apart because of data silos. In the worst case, this can result in researchers spending millions of dollars to replicate research conducted by another lab.
The Problem: Siloed Data
One pharmaceutical company was experiencing just this problem, and they turned to Stardog for a solution. Stardog is an Enterprise Knowledge Graph platform that offers a flexible, reusable data layer to answer complex queries across data silos. At this pharmaceutical company, data could be integrated as needed for individual projects, but only in a very time- and labor-intensive manner. And even then, researchers and business analysts could not be sure they had incorporated all the data relevant to their particular research area. Rather than standing on the shoulders of fellow scientists, they oftentimes were duplicating work.
Replicated many times over across the organization, this conversion, pooling, integration, and mining of data did not result in clear, unified, actionable data, but instead clustered data silos. These systems were adept at producing a single result for a particular research area but were not helpful in cross-functional use cases.
Researchers needed a Google search-like environment where they could explore the entirety of R&D data across the organization so they could more quickly make decisions regarding the particular project they were working on.
Building a Decision-Intelligence Dashboard
This pharmaceutical company created a dashboard that allows researchers to explore data across various data types. Now, when searching for a particular gene, the researcher is served results related to that gene, as well as all synonyms of that gene. So, even if a separate lab uses a different nomenclature, scientists will be able to see and make use of this data.
Moreover, the company created a knowledge panel that proactively presents relevant research content to analysts. With this tool, scientists can explore the connections between different data points and refine their queries accordingly as they learn more about the assumptions that underlie past research. The tool will be used by 1,000s of researchers — and is complemented by machine learning — to identify potential drug targets. All the while, the application adheres to FAIR principles, a collection of guidelines to improve the findability, accessibility, interoperability, and reusability of biomedical digital assets.
Powered by Knowledge Graph
This pharmaceutical customer’s solution is made possible by Stardog’s Enterprise Knowledge Graph platform, a data solution that effectively turns data into knowledge by marrying data with its real-world meaning. Knowledge Graphs deliver insight through their unique ability to find connections in data across data silos in the enterprise. They are also adaptive to new information — accepting and linking new facts into the network seamlessly.
The “graph” in Knowledge Graph refers to a way of organizing data that highlights relationships between data points. Graph data is like a network of interconnected points, in contrast to databases like Oracle or MySQL — relational systems — where data is stored in tables.
Stardog’s Enterprise Knowledge Graph works, first, by taking critical data stored in different places and unifying it, offering flexibility between ETL or virtualization. Virtualization ensures that all relevant data is incorporated and enables tasks like tracing data lineage, detecting correlation and causation and performing root-cause or impact analysis. Once data is in the Knowledge Graph, Stardog adds real-world context to it and begins to augment it via our Inference Engine, which intelligently derives new knowledge from the data and the business logic. Placed in context and enhanced with new information, the data is now ready to be analyzed and put to use within the organization.
Accelerating Drug Discovery in Pharma R&D
The pharmaceutical data landscape is complex, with highly distributed and heterogeneous datasets. In their current state, they slow down the drug discovery process and put tens of millions of dollars of revenue at risk every day.
By leveraging Stardog’s Enterprise Knowledge Graph platform, pharmaceutical companies can unify data across silos to let researchers more easily see the connections in the data. The Knowledge Graph grows as you add more data to it and intelligently derives new connections amongst the data over time.
And accelerating drug discovery is just the beginning – there are myriad other solutions Knowledge Graphs can support in pharma R&D, including:
· Scanning and analyzing scientific texts
· Selecting the best labs, sites or researchers when planning research operations
· Mapping and coding relationships to improve scientific search and knowledge dissemination across labs
· Repurposing compounds based on clinical outcomes
You can learn more about Stardog’s AWS-powered solutions for pharma here.