Improving value assessment of high-risk, high-reward biotechnology research: the role of 'thick tails'.

This paper presents work toward improving the efficacy of financial models that describe the unique nature of biotechnology firms. We show that using a 'thick tailed' power law distribution to describe the behavior of the value of biotechnology R&D used in a Real Options Pricing model is significantly more accurate than the traditionally used Gaussian approach. A study of 287 North-American biotechnology firms gives insights into common problems faced by investors, managers and other stakeholders when using traditional techniques to calculate the commercial value of R&D. This is important because specific quantitative tools to assess the value of high-risk, high-reward R&D do not currently exist. This often leads to an undervaluation of biotechnology R&D and R&D intensive biotechnology firms. For example, the widely used Net Present Value (NPV) method assumes a fixed risk ignoring management flexibility and the changing environment. However, Real Options Pricing models assume that commercial returns from R&D investments are described by a normal random walk. A normal random walk model eliminates the possibility of drastic changes to the marketplace resulting from the introduction of revolutionary products and/or services. It is possible to better understand and manage biotechnology research projects and portfolios using a model that more accurately considers large non-Gaussian price fluctuations with thick tails, which recognize the unusually large risks and opportunities associated with Biotechnology R&D. Our empirical data show that opportunity overcompensates for the downside risk making biotechnology R&D statistically more valuable than other Gaussian options investments, which may otherwise appear to offer a similar combination of risk and return.

Combinatorial design of passive drug delivery platforms.

We introduce a novel computational approach to designing passive drug delivery systems based on porous materials such as hydrogels. Our approach uses three tools: a method to establish the exact release pattern from all possible loading sites inside a given hydrogel; a method to generate a large number of hydrogel structures to be tested numerically, and finally an optimization algorithm which leads to the selection of optimal hydrogel structures. Using this approach, we show that controlled release curves can be obtained by using a genetic algorithm for the optimization step. Strategies to generalize this approach to other systems are also discussed.