ABSTRACT
New drug development guarantees a very high return on success, but the success rate is extremely low. Pharmaceutical companies have attempted to use various strategies to increase the success rate of drug development, but this goal has been difficult to achieve. In this study, we developed a model that can guide effective decision-making at the planning stage of new drug development by leveraging machine learning. The Drug Development Recommendation (DDR) model, we present here, is a hybrid model for recommending and/or predicting drug groups suitable for development by individual pharmaceutical companies. It combines association rule learning, collaborative filtering, and content-based filtering approaches for enterprise-customized recommendations. In the case of content-based filtering applying a random forest classification algorithm, the accuracy and area under curve were 78% and 0.74, respectively. In particular, the DDR model was applied to predict the success probability of companies developing Coronavirus disease 2019 (COVID-19) vaccines. It was demonstrated that the higher the predicted score from the DDR model, the more progress in the clinical phase of the COVID-19 vaccine development. Although our approach has limitations that should be improved, it makes scientific as well as industrial contributions in that the DDR model can support rational decision-making prior to initiating drug development by considering not only technical aspects but also company-related variables.
ABSTRACT
BACKGROUND: The recent innovation activities of global top-tier pharmaceutical companies in accordance with global and regional health concerns were investigated in order to identify their innovations contributing to population health. METHODS: "Innovation activity" was defined as the number of drugs for which R&D activities have been reported within the last three years. Such activities were measured by collecting the data on drug developments and classifying them by developer company, phase of development, therapeutic use, and the country in which the development conducted. Subsequently, we examined and compared the correlations between the global innovation activities of the top 20 pharmaceutical companies and the disease burden measured in disability-adjusted life years (DALYs) by income level and region. In addition, this study analyzed the association between country-specific innovations and DALYs in the corresponding countries. RESULTS: At a global level, the innovation activities were not associated with global DALYs. However, when analyzed by income level, the innovation activities were associated with DALYs in high income and upper middle income countries while it was not associated with DALYs in low middle income and low income countries. In terms of region, correlations were found between the innovation activities and DALYs in the European region, the Americas, and the Western Pacific region whereas such correlations were not found in the African, Eastern Mediterranean, and South-East Asian regions. Similar to the analyses by income level and region, correlations between country-specific innovations and DALYs were only found in high income or high GDP countries. In addition, an empirical analysis of several cases including Canada, Germany, South Korea, and the United Kingdom revealed that pharmaceutical innovation is more closely related to market size than disease burden. CONCLUSIONS: This study identified that discrepancies between pharmaceutical innovation and public health needs, i.e., disease burden values, have persisted until recently. To alleviate this imbalance, both public and private sectors should not only fulfill their respective roles and responsibilities regarding these issues, but also make strategic and collaborative efforts such as Product Development Partnerships (PDPs) directed toward public health improvement.