Search

Most approaches to machine learning from electronic health data can only predict a single endpoint. The ability to simultaneously simulate dozens of patient characteristics is a crucial step towards personalized medicine for Alzheimer's Disease. Here, we use an unsupervised machine learning model called a Conditional Restricted Boltzmann Machine (CRBM) to simulate detailed patient trajectories. We use data comprising 18-month trajectories of 44 clinical variables from 1909 patients with Mild Cognitive Impairment or Alzheimer's Disease to train a model for personalized forecasting of disease progression. We simulate synthetic patient data including the evolution of each sub-component of cognitive exams, laboratory tests, and their associations with baseline clinical characteristics. Synthetic patient data generated by the CRBM accurately reflect the means, standard deviations, and correlations of each variable over time to the extent that synthetic data cannot be distinguished from actual data by a logistic regression. Moreover, our unsupervised model predicts changes in total ADAS-Cog scores with the same accuracy as specifically trained supervised models, additionally capturing the correlation structure in the components of ADAS-Cog, and identifies sub-components associated with word recall as predictive of progression.

The Supreme Court’s recent decision in MedImmune v. Genentech left patent owners who license out their patents in exchange for royalty streams in a bad spot. It is an especially dire spot for patent owners such as universities who incur substantial opportunity costs when they grant an exclusive license to a licensee who pays relatively little up front, in exchange for paying a potentially sizable royalty stream if or when products based on the patents are successfully commercialized and sold in the marketplace. Because the MedImmune decision allows such licensees to bring declaratory judgment actions to invalidate the patent or establish the licensee’s products as non-infringing with no particular trigger and without repudiating the licensee (and thus avoiding the possibility of a patent owner suit for infringement), it is expected that licensees will take a license just to buy time to develop a product with relative freedom to operate and then attempt to invalidate the patent once a product looks like it will become profitable and royalties would have to be paid. While some have suggested that the solution to this is simply for patent owners to demand full payment of the net present value of the royalty stream up front, this will simply not be possible for many start up and mid sized companies who license technologies from universities and other routine licensors. The author instead proposes a method for licensors to take some combination of stock and stock options in the licensee – that essentially the licensee can “afford” to pay in the near term even while short on cash and revenues – while still allowing the licensor to participate in the potential upside of a successful commercialization effort of the licensor’s patents by the licensee, but in a manner that is fully accrued to the licensor upon the execution of the license.