What Need does RepoMan address?

Repurposing FDA-approved medications for novel indications may offer substantial advantages over drug development based on novel compounds, in terms of both time and cost. Because drug safety and tolerability have already been established, the path to clinical application is far shorter than for novel compound development.

One challenge in repurposing efforts is extrapolating human CNS availability from in vitro models. Chemical properties like molecular weight, models like LogP and traditional PK approaches (i.e., dose and sacrifice rats) have been used for judging CNS availability to date. However, these tools may not be available to clinical/translational researchers with limited resources. This calculator uses an alternate approach, based on the assumption that drugs that enter brain are likely to have adverse effect profiles similar to other brain-penetrant drugs. This is only a heuristic for high speed, low cost screening of compounds for further evaluation, not a perfect solution. Nevertheless, with cost and time near zero, is may be a reasonable first step.

What is RepoMan?

RepoMan, or Drug Repurposing Manager, implements the adverse effect similarity estimates described in McCoy and Perlis, in review. It is a tool for manipulating and drawing inferences based pairwise similarity measures and z-scores originally reported by Tatonetti and colleagues, and accessible via Tatonetti et al, Sci Trans Med 2012. Side effects reflect both package insert and postmarketing data. The additional chemical property data (molecular weight and log-P) are drawn from PubChem, last updated 7/6/2013.

In brief, similarity measures between each individual drug, and a panel of comparator drugs, are calculated as the mean of similarity scores and z-scores. The presented p-values are approximate upper bounds (based on 100 permutations) and should be viewed simply as a rough empirical estimate of statistical significance. The axis of the results plot are as labeled but have little absolute meaning; rather, their utility is limited to within-panel comparisons (that is, values are transformed to make differences between individual compounds as clear as possible).

How to Repo?

Panels of reasonably related drugs (like those in the pre-build panel dropdown) tend to produce expected results with clusters of similar agents identified. You can test the edges of this approach with class-specific or wholly arbitrary panels (for example, all atypical antipsychotics or one agent from multiple unrelated classes respectively).

The implementation is for demonstration rather than production purposes. Please use a modern standards compliant browser (current Chrome or Firefox are your best bets). If you run into problems clear your browser cache and reload the page. If all else fails, practicing mindfulness or watching videos of dancing cats may also be helpful.