We consider a pharmaceutical Research & Development (R & D) pipeline management problem under two significant uncertainties: the outcomes of clinical trials and their durations. We present an Approximate Dynamic Programming (ADP) approach to solve the problem efficiently. Given an initial list of potential drug candidates, ADP derives a policy that suggests the trials to be performed at each decision point and state. For the classical R&D pipeline planning problem with deterministic trial durations, we compare our ADP approach with other methods from the literature, and find that it can find better solutions more quickly in particular for larger problem instances. For the case with stochastic trial durations, we compare the ADP algorithm with a myopic approach and show that the expected net profit obtained by the derived ADP policy is higher (almost 20% for a 10-drug portfolio).

我们考虑了两个重大不确定因素下的药物研发管道管理问题：临床试验的结果及其持续时间。我们提出一种近似动态编程（ADP）方法来有效解决问题。在给出潜在候选药物的初始列表的情况下，ADP会得出一项政策，建议在每个决策点和每个州进行试验。对于具有确定性试用期的经典R＆D管道规划问题，我们将ADP方法与文献中的其他方法进行了比较，发现它可以更快地找到更好的解决方案，特别是对于较大的问题实例。对于随机试用期的情况，