Real output
Three kinase targets. Every number below came out of the engine.
For each target: a change found in that library, shown as two compounds that were actually made and measured — then molecules built from the same change, with the numbers the engine returned for them.
A change found in the library
Two compounds your chemists already made and measured.
2,4-difluorophenyl replaced by 2,3-dichlorophenyl on the same scaffold
Both compounds are real rows of the ABL1 library, and both numbers are their own measurements. Nothing here is predicted.
Molecules built from it
The same change, applied where nobody has made it yet.
Each pair is a compound from the library and the molecule built from it. Measured is the starting compound’s own value. Precedent is what compounds that already made this change measured. Predicted is the model’s value for the new molecule.
A change found in the library
Two compounds your chemists already made and measured.
a phenyl ring replaced by a thiophen-2-yl ring (phenyl / thiophene bioisostere), one ring per molecule
Both compounds are real rows of the AURKA library, and both numbers are their own measurements. Nothing here is predicted.
Molecules built from it
The same change, applied where nobody has made it yet.
Each pair is a compound from the library and the molecule built from it. Measured is the starting compound’s own value. Precedent is what compounds that already made this change measured. Predicted is the model’s value for the new molecule.
A change found in the library
Two compounds your chemists already made and measured.
The terminal pyridine is oxidised to its N-oxide. Nothing else changes: same anilinoquinoline core, same two chlorines and two methoxy groups, same nitrile, same vinyl linker.
Both compounds are real rows of the SRC library, and both numbers are their own measurements. Nothing here is predicted.
Molecules built from it
The same change, applied where nobody has made it yet.
Each pair is a compound from the library and the molecule built from it. Measured is the starting compound’s own value. Precedent is what compounds that already made this change measured. Predicted is the model’s value for the new molecule.
These are three public targets. The interesting run is the one on your own compounds.
The engine reads the library you point it at. Different compounds, different changes, different molecules — which is the demo worth having.