Throughout my two-decade career in the biopharmaceutical industry, my goal has been to make drug development more efficient while reducing costs, and to help the companies I have worked for increase revenue without compromising quality. In pursuing that goal, I have strived to embrace advanced technologies and to optimize organizational and operational structures. Now that I have joined Phesi, I am very excited about the potential of artificial intelligence (AI)-driven drug development to truly deliver novel therapies to patients faster than ever before.
Technology has come a long way since my days at NYU School of Medicine, where I used Excel as my main analytics tool for managing a new lymphocyte proliferation assay. Upon transitioning into pharma I continued to use Excel as a core tool for most projects, but soon recognized that as projects became more complex, it was necessary to design more sophisticated, customized solutions to address clinical trial inefficiencies. My colleagues and I used various modeling and simulation technologies, including Monte Carlo Simulation (MCS), to virtually test multiple scenarios before making data-driven decisions on the most appropriate operational planning parameters to select for each clinical trial.
MCS is based on probabilities and requires some tolerance for risk. This reality became apparent as I transitioned to the CRO side of the industry. I found that pharmaceutical sponsors were more
risk-tolerant and willing to plan for and pursue shorter enrollment cycle times, even when the projected probability of achieving that goal was relatively low. Understandably, CROs need to give their clients more confidence that they can deliver on time. In fact, I remember one of my CRO colleagues being shocked to discover we were planning with anything less than 100% probability of success. This type of planning implies zero risk and absolute certainty that the project will be delivered on time. In addition to being too conservative, this approach is expensive, as it is a plan for perfection, which is never truly achievable. Finding that balance between perceived perfection and getting tasks accomplished is worth striving for, but one must consciously guard against the waste associated with seeking perfection.
The last several years have seen significant advances in technology to keep up with the exponential increase in the amount of generated data each day. In 2013, for example, the world was generating 4.4 zettabytes of data per day, an amount that is expected to increase to 44 zettabytes by 2020 (1 zettabyte = 44 trillion gigabytes)1. We therefore need better methods of analyzing data to make sense of it. It is no longer adequate to write simple Excel formulas or to design simple, static algorithms we can easily understand and manipulate. It is also not sufficient to rely solely on structured data with a neat column or field for everything. Life does not happen in a straight line or a nicely structured way. There are bumps along the way, in the form of disappointments and new information, yet resilient people still manage to achieve their goals.
We need to view the future of data analytics in this way. As the field continues to evolve, we will need powerful, adaptable analytical tools to keep up. This is why I have aligned with Phesi, an organization leading the way in bringing AI technologies such as machine learning and natural language processing to drug development. After all, we need to speed up the clinical study decision-making process, so we can get medical innovations to patients to improve and save lives.