To identify promising drugs – and avoid failures – researchers are using machine learning and latest text mining methods
In 2001, when Jamie was diagnosed with chronic myelogenous leukemia, a cancer that starts inside the bone marrow, the disease had few effective cures. Fourteen years later, thanks to advances in cancer treatment, she is able to manage the disease and live a full life. Jamie is profiled in the PhRMA.org series I’m Not Average, which shows how breakthrough medicines are enabling patients to live longer, healthier, happier lives.
Yet many patients and their doctors wait for years before promising treatments become available. All too often, unforeseen side effects send researchers back to the drawing board, just when they thought they were close to bringing a new medication to market.
It takes, on average, at least 10 years for a drug to make the journey from discovery to the marketplace at an average cost of $2.6 billion, according to a new study in the Journal of Health Economics, based on data from the Tufts Center for the Study of Drug Development. The overall cost includes not only the development costs for drugs that successfully made it to market, but also for the drugs that failed along the way. Today, the likelihood that a drug entering clinical testing will eventually be approved is estimated to be less than 12 percent.
Want to publish your own articles on DistilINFO Publications?
Send us an email, we will get in touch with you.
But what researchers are learning is that by using certain technologies early in the drug-development process, they can identify issues that might cause a drug to fail early on, in many cases before the compound even goes into clinical testing. Then they can either modify the compound to address the issues, while maintaining the therapeutic effects, or make an early decision to no longer pursue the drug candidate, thereby averting a more expensive later stage failure.
“Training a machine to see more than we can”
Machine learning — essentially, using computer technology and analytical tools to train a “machine” to see more than we can — is one way that technology can help streamline the process of finding and developing new drugs. For example, Silicon Valley startup twoXAR used a machine learning system to identify promising drugs to combat Parkinson’s disease. In an interview in Datanami, twoXAR co-founder Andrew A. Radin said, “We loaded a bunch of data on Parkinson’s disease into the system, pressed the go button, a few minutes later we had a list of drugs that were listed as highly efficacious.”
When Radin searched further, one of the top hits was a drug that Dr. Tim Collier, Director of the Udall Center of Excellence for Parkinson’s Disease Research at Michigan State University and contributor to various neurology journals published by Elsevier, had identified as being potentially important. The system had managed in minutes to help validate work that had been underway in Collier’s lab for years. The result is an ongoing collaboration between twoXAR and the center that promises to accelerate the discovery and development of candidate anti-Parkinson’s drugs.
Mining big data
“Pharma company researchers are working hard to manage all the big data coming their way,” said Dr. Jaqui Hodgkinson, VP of Product Development Biology and Preclinical Products at Elsevier and a former clinical data scientist for Glaxo Wellcome. “Managing and understanding that data is critical to getting new medicines to market sooner. That’s why we’re continually expanding our text mining systems to handle input from diverse sources.”
To discover and develop a new drug, researchers need to know, at the minimum, what has already been published in peer-reviewed biomedical journals about their compound. But to get the most relevant information — and save time , money and unnecessary experimentation — it helps to use a system that can also process and analyze related input, such as regulatory information, reports of side effects from medications related to the one they’re investigating — and even comments from social media.
Deep text mining and analysis is also key to drug-repurposing — that is, finding new uses, or indications, for existing drugs. This is an important business strategy for pharmaceutical companies because it helps them increase the return on their R&D investment. But it also helps them stay true to their commitment to address areas of unmet medical need, so this strategy helps patients.
At Elsevier, we’re working with the UK nonprofit Findacure to help researchers identify drugs approved for other disorders that could also help combat rare diseases. As part of a collaboration that began in September 2015, Elsevier is providing informatics expertise and advice, as well as access to the published literature, on a drug called sirolimus, which is being repurposed as a treatment for an extremely rare disease: congenital hyperinsulinism.
We’ll also help in a later stage, when sirolimus is ready for testing, with tools such as Pathway Studio, which enables the study and visualization of disease mechanisms, gene expression, and proteomics and metabolomics data, to assess CHI’s biological make-up in depth, and then shortlist additional promising potential treatments that could be repurposed safely and effectively.
Using the same technologies, a similar approach was used to help pharmaceutical companies identify new indications for, among others, the TNF-inhibitor adalimumab, and the anti-cancer drug, imatinib.
We continue to expand our tools to enhance their accuracy and help our business partners, academics and nonprofits such as Findacure. For example, processes such as sentiment analysis — a way to identify words and phrases that indicate opinions, attitudes and lack of certainty (e.g., “suggests,” “seems to indicate”) to biomedical literature searches and relevant input from social media — will be important. Although it’s still early, with challenges such as the need for linguistic resources and the ambiguity of words and their intent (e.g., with irony and sarcasm), Dr. Hodgkinson said sentiment analysis will play an increasingly important role in streamlining data analysis in the near future.
We’re also collaborating in the public and private sectors to help everyone have greater access to data that will make finding new drugs faster, easier, less costly — and less disappointing to patients and doctors who are waiting for more effective treatments. For example, Elsevier recently launched an Open Data pilot, which makes the raw research data submitted with an article accessible online to all users alongside the published article. We also have been helping authors post their data in relevant public data repositories with a linkback to their published article. This helps researchers working on similar projects to access vital information about what other groups are doing and potentially validate that work.
In a project co-funded by a National Science Foundation EAGER grant, Elsevier also is working on a data search pilot with the Carnegie Mellon School of Computer Science to facilitate the querying of tabular content extracted from articles and imported from research databases.
We anticipate that the importance of collaborations will only grow, particularly now that precision medicine — targeting specific treatments to patients most likely to benefit — is taking hold. In the US, the National Institutes of Health’s Precision Medicine Initiative Cohort Program is working to “engage partners across all communities — scientific, medical, health, and societal — and public as well as private sectors,” by inviting participation of “patients and patient advocacy organizations, academic medical centers, clinicians, scientists from multiple disciplines, pharmaceutical companies and medical product developers, scientific societies and research coalitions, privacy experts, and medical ethicists.”
Our technologies are facilitating collaborations across sectors and organizations to help make targeted medicines a reality. We know that all these advancements bode well for patients like Jamie, whose lives stand to be transformed for the better as researchers adopt the most powerful available tools to help them on their quest for new treatments.
Date: November 6, 2018