A Genetic Algorithm-based Pharmaceutical Portfolio Selection and Scheduling Framework

One of the greatest challenges facing life sciences companies is the ability to discover and develop new products, which will sustain their long-term economic growth. The problem is difficult because of intense competition, the availability of a large number of new product ideas and limited human and capital resources. The planning problem is exacerbated by the presence of significant uncertainties in development times, development costs, resource requirements and anticipated product sales. In addition, the presence of dependencies between products, both in the market place as well as in their development adds further complexity. In this paper, a discrete event simulation for the drug development process is combined with a Genetic Algorithm (GA) to select the best sequence of projects in the presence of uncertainties and dependencies. A graphical tool ,the risk-reward bubble plot is used to arrive at heuristics that prune the GA search. The GA then captures the optimal structure of the highest expected reward/risk sequences, despite an enormously large combinatorial search space. A risk-reward frontier, which illustrates the trade-offs between risks and expected rewards is presented as an output of the optimization exercise. For a nine drug example case study, the optimal portfolio is expected to give a 50 % higher return than the portfolio suggested by the bubble diagram at the same level of risk. A sensitivity analysis demonstrates the robustness of the sequence to changes in the levels of uncertainties and dependency conditions.