The pharmaceutical industry has long stayed humble with their rules for successful drug development: "We’re not bigger than the game!" "We will follow rules even if we don’t understand why they work!" Chris Lipinski’s Rule of 5 is a cornerstone, of course, but only kicked things off.
Turns out, the dreaded spectre of attrition was barely slowed down by medicinal chemistry rules. To this day, the challenge of predicting clinical/preclinical success from a compound structure alone has remaineds a pipe dream. The surrender of the pharmaceutical industry in the war against drug-like small molecule drug attrition was really an adaptation to end-arounds: Biologics! Phenotypic screens! Click chemistry! And with these new hopes came new rules, notably Pfizer’s “Rule of 3” for high-throughput phenotypic assays.
These rules can tell you how closely your phenotypic–usually cellular–screen in your dish will mimic the in vivo situation. Because it is phenotypic, it’s “closer” to the in vivo situation than biophysical or biochemical assays, but how close? The three rules tell you. They are:
Rule 1: relevance of the phenotypic assay to the in vivo situation.
Rule 2: relevance of the way that the phenotype is manifested in the assay biological material (termed the “stimulus” by the authors”)
Rule 3: relevance (proximity) of the readout of the assay to the disease
The rules require some further explanation. Rule 1 is the easiest to grasp: If the assay is to inject a mouse with a drug then that is closer to the in vivo situation than putting the drug with some cells in a dish. So, biophysical assay is worse than biochemical, which is worse than cellular, then organoid, then small animal, then primate, then direct human test, more or less.
Rule 2 is a little more subtle. Here, by stimulus the authors mean how “natural” are the cells, animals etc used in the assay. So if the cells are human cells obtained from a patient with the disease and exhibit the disease phenotype “endogenously” because they contain the defective genes naturally, for example, that is better than a cell that was transfected or edited to create the defect.
Rule 3 is about approximations–the fewer the better. So, static protein levels as a readout of the assay is worse than a functional readout related to the disease like blood cell sickling in sickle cell anemia. Generally, here gene expression is worse than protein levels is worse than cellular function is worse than organ or animal function.
By systematizing the thinking on phenotypic screens, the authors have provided a useful starting point to think about the translation of the hits from your phenotype screen into real drugs. But there is much more work to be done.
Now, GeneCentrix is developing a fourth rule based on tissue specificity. These innovative new drug and target tissue profiling tools can be the basis of bringing tissue-expression and drug activity patterns into play in the in vivo prediction game.