Predicting drug approvals: The Novartis data science and artificial intelligence challenge
| Autor: | Yang Zhong, Bin Zhou, Kien Wei Siah, Steffen Ballerstedt, Björn Holzhauer, Simon Wandel, Shifeng Pan, Andrew W. Lo, Tianmeng Lyu, Nicholas Kelley, Yingyao Zhou, Sophie Sun, David Mettler |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
clinical trials
Point (typography) business.industry Computer science media_common.quotation_subject General Decision Sciences probability of success Descriptor prediction artificial intelligence Data science Clinical trial Probability of success Subject-matter expert Drug approval Artificial intelligence data science Baseline (configuration management) business Sophistication media_common XGBoost |
| Zdroj: | Patterns |
| ISSN: | 2666-3899 |
| Popis: | Summary We describe a novel collaboration between academia and industry, an in-house data science and artificial intelligence challenge held by Novartis to develop machine-learning models for predicting drug-development outcomes, building upon research at MIT using data from Informa as the starting point. With over 50 cross-functional teams from 25 Novartis offices around the world participating in the challenge, the domain expertise of these Novartis researchers was leveraged to create predictive models with greater sophistication. Ultimately, two winning teams developed models that outperformed the baseline MIT model—areas under the curve of 0.88 and 0.84 versus 0.78, respectively—through state-of-the-art machine-learning algorithms and the use of newly incorporated features and data. In addition to validating the variables shown to be associated with drug approval in the earlier MIT study, the challenge also provided new insights into the drivers of drug-development success and failure. Highlights • Data science challenge at Novartis in collaboration with MIT researchers • Dataset derived from 2 decades of drug development and clinical trial data • Winning teams outperformed the baseline MIT model by leveraging domain expertise • Predictive analytics can augment human judgment in drug-development risk management The bigger picture The probability of success is a key parameter that clinical researchers, biopharma executives and investors, and portfolio managers focus on when making important scientific and business decisions about drug development. We describe an in-house data science and artificial intelligence challenge organized by Novartis in collaboration with MIT researchers. Using state-of-the-art machine-learning algorithms and extensive feature engineering augmented by domain expertise in drug development, two winning teams developed models that outperformed the baseline MIT model proposed in a prior study. These new predictive models can be used to augment human judgment to make more informed data-driven decisions in portfolio risk management and capital allocation. These results suggest the possibility of developing even more accurate models using more comprehensive and informative data, and a broader pool of challenge participants. In an in-house data science challenge, Novartis researchers developed machine-learning models for predicting drug-development outcomes, using 2 decades of clinical trial data and building upon previous work at MIT. By leveraging domain expertise, 2 winning teams, out of 50 that participated, developed models that outperformed the baseline MIT model through state-of-the-art algorithms and feature engineering. In addition to providing new insights into drug approvals, the models can augment human judgment to make more informed risk-management decisions. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|