Bringing a drug from discovery to market can be a costly gamble in the range of billion of dollars on average, potentially taking a decade or more. What makes matters worse is that only a tiny fraction of potential drugs that are pursued end up making it to market. Drugs make it or break it during clinical trials, a critical step in the drug development process where the drugs are given to patients to test if they are safe, and to test if they will actually have the desired outcome (known as efficacy). Clinical trials are sometimes referred to as the "valley of death," because so many promising drugs fail to demonstrate their safety or efficacy in humans that they are tested on, and thus are barred by the FDA from being sold. One way to improve the success rates of clinical trials is find a way to predict, ahead of time, which patients will respond favorably to a drug. A biomarker is simply a measureable characteristic of an individual human that can accurately predict a property, such as drug response. Hence, when recruiting patients for clinical trials, one can test for biomarkers. If an experimental drug has a predictive biomarker, pharmaceutical companies can try to select only those patients who have this biomarker, increasing the chances that they can prove their drug works (has high efficacy) during clinical trials, and ultimately increase the chances that their drug will pass FDA approval and escape the "valley of death." This is important for two main stakeholders. First, patients need more drugs to reach the market so that they can be cured, as well as more biomarkers to help physicians prescribe drugs that will work on them personally. Second, pharmaceutical companies want more drugs to reach the market so that they can increase returns on research and development, reduce drug-production costs, and increase their revenue.
Predicting if a drug will work or not in a patient before the patient has ever even taken the drug is still in its infancy, and has been most adopted in oncology. Current approaches to discovering predictive biomarkers revolve around using statistics to correlate mutations in patients' genes to drug response. Yet, additional information, such as the fact that genes are made up of different units or that genes are blueprints for proteins that interact with each other, is completely ignored. Including this information, this team has discovered new approaches to finding biomarkers even though they are looking at the same data. This team's current progress has been almost exclusively limited to publicly available datasets (such as The Cancer Genome Atlas, The Cancer Cell Line Encyclopedia, etc.), but with this data alone algorithms have been created to elucidate biomarkers and patented 171 previously unknown biomarkers. This yteam has also created a non-commercial tool www.cancer3d.org to allow scientists to access its analysis for their research purposes. The team's goal is to gain access to pharmaceutical companies proprietary data sets which they have generated as they prepare for their clinical trials, analyze their data (or provide tools for them where they can analyze it themselves), find new biomarkers for their experimental drug, and license these biomarkers to the company. A strong predictive biomarker for a cancer drug can easily help get a drug to market, help cancer patients everywhere, and create significant value; biomarkers have been sold for tens of millions of dollars. Through the NSF I-Corps program, this team hopes to learn and improve on customer identification and discovery, customer interactions, how to move from market research to customer acquisition, discover if the team is offering a product that people will want, and ultimately improve and learn on every aspect of commercializing a research project.
