By Aileen Agricola | October 28, 2015
How simple graphing solutions are leading to better treatment and saving lives.Written by Emil Eifrem, CEO Neo4j
The potential of new database technology to solve modern business challenges is widely recognized. Some of the world’s largest organisations have already adopted graph databases for everything from to providing user recommendations for new products to buy (Walmart) and friends to connect with online (Facebook/Twitter). But now the power of graph databases is being harnessed by modern medicine to revolutionize the treatment of diseases such as cancer, ensuring that patients receive the most effective drugs as early as possible to maximize their chances of beating the illness.
The ongoing battle
Cancer is not a single disease, but rather a term for a broad class of over 10,000 diseases that cause the body’s cells to mutate and multiply without control. These mutations can happen anywhere in the body, resulting in different outcomes in each patient.
A huge array of drugs already exist to treat the various forms of cancer, with more brought to market all the time, but part of the problem with current treatment is the inability of many medical professionals to access the information they need to effectively match the best drugs to individual patients. Patient specific factors such as age, gender, cancer type, disease progression and specific gene mutations can all play a significant role in determining how effective one drug is compared to a similar alternative.
While there is an ever-growing body of global research on how these factors can affect the way an individual responds to certain drugs, there is currently no centralized database containing all of this key information. A lot of it remains siloed in research facilities, inaccessible to the individual doctors tasked with making the critical decisions around patient treatment. Of the databases that do currently exist, few are intelligent enough to allow for tailored searches based on critical patient specific factors.
For example, ovarian cancer is one of the most aggressive forms of cancer, with just 5 percent of those diagnosed surviving more than 12 months from the point of diagnosis. As such, getting the right treatment as quickly as possible is critical. At present, there are three recognized chemotherapy treatments for ovarian cancer, but without access to research insights and patient case studies, doctors cannot be sure which treatment is best for any specific patient. Even if they can locate pertinent information, to manually cross reference it all would be an extremely time-intensive exercise. As such, trial and error is generally used to find the most effective treatment. However, chemotherapy is not only expensive (typically $100,000 per round), but it often takes three months to complete a round and realize its effectiveness. As a result, the time wasted trying to find the most effective treatment can often mean the difference between life and death.
How can graphs help?
Graph databases excel in tasks related to search and recommendation because they not only store information about individual things, but also the relationships between those things. This capability allows users to ask questions that were previously not possible with traditional database technologies. The data relationships stored in the graph database can express the nature of each connection (e.g. drug family, type of cancer targeted) and capture any number of qualitative or quantitative facts about that relationship (e.g. optimal dosage level, treatment success rate, effectiveness against mutations, and date brought to market). Once loaded into a graph database, an entirely new set of relationship-based questions can be asked, opening up new possibilities.
When applied to cancer treatment, it means doctors can ask questions like the following, to help tailor treatments to individual patients:
“Find all cancer treatment drugs available since 2012 which have proved most effective in treating stage two ovarian cancer in women under the age of 30”
The result is to assist medical professionals in recommending better treatments, using data relationships to increase precision and accuracy.
Leading the charge
Annai Systems, a genomics and data management company based in Silicon Valley, recently commissioned a research team to create a new system capable of helping doctors make the most informed decisions possible when considering the most effective treatment of patients with ovarian cancer.
The team knew that the key to solving the challenge was finding a way to effectively combine all of the existing (but siloed) research, knowledge and data into a single, standardized database that could quickly analyse all factors and provide recommendations based on this information.
They identified around 20 individual databases containing three key types of information that were critical to the decision making process:
Patient data: Biological information on individual patients and the specific factors of their illness; Reference data: Existing gene and drug information from around the world; and Experience data: Information in the form of research papers and clinical trials into different cancer mutations and the best drugs to treat them.
The team then entered all of the information from the 20 different databases into a single Neo4j graph database, refining/standardizing the information and evolving the relationships between the data sets as they went. After several months, they had built a comprehensive, uniformed database and could connect all of the different treatment options with specific patient biomarkers, producing a ranked list of treatments for each based on its effectiveness (drawn from the empirical data). The result is a fast, efficient way to pinpoint the best course of action for an individual based on their unique patient characteristics, reducing the need for trial and error, and saving precious time in the fight against the disease.
The solution is now being implemented by medical firms specializing in ovarian cancer treatment and databases of this type are expected to play a key role in continuously improving cancer treatment for the next decade.
For now, the ultimate solution for beating cancer remains buried within the human genome and likely won’t be fully unlocked for another 10-15 years. However, by making the best use of what is already known about the disease, medical professionals are able to improve the odds for every patient currently fighting cancer and graph databases can be a key asset in this ongoing battle.
Keywords: emil eifrem