3Z identifies and develops drug candidates for disorders of the central nervous system through a five-stage workflow that starts in zebrafish and ends with predictions from large human datasets. Each stage removes weaker candidates and adds an independent line of evidence.
Stage 1 of 5
For each project we assemble a library of compounds, usually approved or clinically tested drugs with established safety data. Before any compound reaches our zebrafish, we assess each candidate against four questions that decide whether a positive result could realistically become a new medicine. Only compounds that pass this assessment proceed to screening. The same questions are revisited with stronger data as a candidate advances through the later stages.
Stage 2 of 5
More than 80 percent of known human disease genes have a counterpart in zebrafish, and their larvae are small enough to be tested in large numbers. This allows us to measure drug effects on the behavior of a whole, living nervous system at a scale usually limited to cell-based assays.
We first create a zebrafish model of the disorder, typically by using precise genome editing to disrupt the zebrafish counterpart of a human risk gene. We then define its phenotype by identifying the behavioral measures that reliably separate the model from healthy wild-type fish. Where approved treatments exist, we confirm that they correct this phenotype, showing that the model responds to clinically effective drugs.
Next, we screen the compounds that passed the developability assessment. Larvae are dispensed into multi-well plates and exposed to each compound at several doses. Automated tracking records the behavior of every larva, and a machine-learning classifier identifies compounds that shift behavior back toward the healthy profile. Only compounds that restore the healthy control phenotype without signs of toxicity or unwanted effects, such as sedation, advance to the next stage. The figure below illustrates this screening stage.
Stage 3 of 5
Compounds that pass the screen are studied in more depth in zebrafish. Further behavioral tests in larvae and adult fish cover additional features of the disorder. Each candidate is benchmarked against approved treatments and against related compounds from the same drug class. We measure brain penetration, map how each candidate changes activity across the brain and use multi-omics analyses of brain tissue to show which molecular pathways it restores toward the healthy state and whether these overlap with pathways affected by established treatments. Together, these studies clarify how each candidate works and point to biomarkers that can track its effect.
Stage 4 of 5
The strongest candidates are then tested in established rat and mouse models, using behavioral tests that capture core features of the disorder, often over several weeks of repeated dosing. We measure drug levels in plasma and brain to confirm that the compound reaches the brain at relevant concentrations. Because the species, genetic background, route of administration and treatment length all differ from the zebrafish screen, a consistent effect in rodents indicates that the finding is robust and not tied to a single model.
Stage 5 of 5
Finally, we turn to large human genetic datasets and biobanks to predict how a candidate may perform in patients. Drug-target Mendelian randomization uses natural variation in the genes that encode the candidate’s targets to test whether acting on those targets is likely to affect the disorder. For drugs already in clinical use, polygenic risk scores let us ask whether people with high genetic liability for the disorder who take the drug show fewer disease-related traits. Comparison with other drugs prescribed for the same original condition helps rule out indirect effects. Because drug targets with human genetic support are more likely to succeed in clinical trials, this stage helps reduce risk before clinical development begins.
Would you like to learn more about our research or discuss a collaboration? We would be glad to hear from you.
