Question 19 of 20 Advanced Scenario
How would you decide between a phenotypic screen and a target-based screen?
Short answer
Choose target-based when a credible target exists, can be assayed and has good evidence linking it to the disease, because optimization is then fast and rational. Choose phenotypic when the biology is poorly understood or depends on a system, accepting that you may have to find the target later.
Why
Target-based screening gives you everything optimization needs: purified protein, structure, a clean potency measurement, and selectivity panels. It also fixes the hypothesis at the start, so a wrong target means a well-optimized compound that fails in patients. Phenotypic screening asks the cell or organism directly whether a compound produces the effect you want, so it can find mechanisms nobody proposed and it naturally requires cell permeability and some stability from the start.
Its costs are the slower, noisier assays, lower throughput, harder structure-activity work and the eventual need for target deconvolution. The decision therefore turns on practical questions. How strong is the genetic and human evidence for the target.
Can the target be assayed in a form that resembles its cellular state. Does a cellular model reproduce enough of the disease for a phenotypic readout to mean something. How will patients be selected later.
In practice the two approaches merge. A phenotypic hit becomes a target-based program once the target is found, and a target-based lead must pass cellular and in vivo tests anyway.
A strong answer covers
- Strength of target to disease evidence and assayability
- Phenotypic screens require permeability and some stability by construction
- Target-based gives fast, rational optimization
- Phenotypic can find unanticipated mechanisms
- Model relevance and later patient selection
Common mistakes
- Picking the approach by throughput alone
- Assuming a phenotypic hit will be easy to deconvolute
- Treating a well-validated target as a guarantee of clinical efficacy
The interviewer might ask next
How would you build a phenotypic assay that still gives interpretable structure-activity relationships?