With the U.S. Food and Drug Administration approving an increasing number of targeted therapies and hundreds more in late-stage oncology pipelines close behind, pharmaceutical companies are racing to bring cancer patients drugs tailored to the molecular profile of their tumors, otherwise known as precision medicine. Hitting the right target in precision medicine, however, is fraught with potential pitfalls that could be the difference between success and failure for a drug.
The stakes are high
“The explosion of molecular diagnostic testing over the last 10 years has led to discovery of many new genes that could contribute to cancer, but it’s not always clear which of them will be druggable or which ones are drivers,” explains Sheryl Krevsky Elkin, PhD, Chief Scientific Officer at N-of-One, a molecular decision support company.
In addition, pharma and biotech companies must deal with shrinking patient populations for their drugs. Traditionally, pharma companies have been able to bring drugs to market that have broad indications across large patient populations (often resulting in
significant returns on investment). With precision medicine, however, the cost of drug development is typically the same but the population of treatable patients, by nature, narrows, translating to more risk and less future earning potential for companies. In other words, as Elkin says, “The stakes for identifying the best patients are high.”
As a result, pharma companies both small and large have an escalating need for better information on the molecular profiles of their target patient populations and throughout drug development. This information can facilitate better molecular targeting of genes and indications, design of clinical trial cohorts, development of companion diagnostics, and forecasting of trial accrual. Even the development of drugs that target well-known cancer drivers requires molecular analysis and deep understanding of the biochemical mechanisms behind the function of both the gene and the drug.
Take, for example, the clinical development of a specific kinase inhibitor. A phase I clinical trial typically starts by having the gene sequenced for clinical trial participants, yielding hundreds of different mutations. Sorting out which variants are important is not trivial. Some variants may have been described in a solitary obscure journal article, while others might never have been reported at all. The dilemma companies face is which variants to include in their clinical trial design.
“You don’t want to take a risk on a lot of variants of unknown significance if there’s really no data at all, because your drug will be less likely to work across a significant portion of the trial population,”
Elkin says. Furthermore, the selected patient population should be large enough so that patients who could benefit are included, yet small enough so that patients who are unlikely to respond to the drug are excluded.
Getting it right
While companies may understand the need for accurate variant interpretation and patient population selection, and have a plethora of new molecular targets to choose from thanks to the popularity of molecular genomic testing, they face another challenge: Sifting through a large and diverse body of literature to find all the relevant information.
Curated databases of genetic variants have since emerged as a solution and can provide insights based on the literature, but not all databases offer high-quality, comprehensive variant classification. Certain databases, such as COSMIC, OncoKB, and CIViC, are open-source and publicly accessible. While these databases offer the benefit of affordability, they come with inherent drawbacks. For example, by relying on a crowdsourcing model to curate the database of genetic variants, CIViC is limited in size and scalability and lacks comprehensive molecular information.
By contrast, large commercial databases have the benefit of being curated by dedicated, full-time experts that can provide high-quality variant interpretation backed by a rigorous, thorough review of the literature. In fact, N-of-One has the largest commercial knowledgebase of curated evidence that harbors more than 100,000 unique somatic variants, 900 cancer types, and 1,800 genes across tens of thousands of patient cases.
“Working with the right company to supply the requisite analysis is a small investment when considering the total cost of developing a biomarker-driven drug,” Elkin says.
Variant interpretation is essential not just early on, when a molecular target is being selected, but throughout drug development to ensure success. Without it, many pharma companies hit a bottleneck when selecting the target patient population and may miss out on an opportunity to expand the patient pool. Continuous variant interpretation allows companies to expand the treatable patient population from, say, 10 to upwards of 30 variants, and if the drug involves a tumor suppressor pathway, the gains can be even bigger because tumor suppressors typically incur many different mutations. Identification of the many variants, however, is complicated, with the list of variants always changing. “That’s a case where engaging companies like N-of-One in a more ongoing manner to continually review new variants could be really beneficial,” Elkin says.