The winning scaffold is not necessarily where the most interesting chemistry ends. A better search may start by moving away from it into structurally different, property-matched chemical space. QTX3034 is a good example. Around this KRAS G12D program the public record shows several promising mechanistic and chemical branches. Only one remained visible as the surviving development path.
Where the light is brightest
We often search where the light is brightest.
In drug discovery, that usually means staying close to what is already known: the successful scaffold, the published series, the obvious analogs.
But that creates a bias.
The closer we stay to the known molecule the more likely we are to keep exploring the same chemical logic.
So I started thinking about a different approach:
What if we use a known molecule as a reference point, but deliberately search outside its obvious chemical neighborhood?
The method
The idea is simple.
Take a reference molecule.
Find compounds with low structural similarity to it.
Independently find compounds with a similar physicochemical profile.
Intersect the two sets.
Then remove compounds built around the same core scaffold.
What remains should be an alternative chemical-space set: different chemistry, but still living in a relevant physicochemical space.
The QTX3034 case
This idea came from a recent case I looked at around QTX3034, a KRAS G12D inhibitor.
I started from a LinkedIn post by Chris De Savi about 13 newly disclosed molecules and traced some of the history behind QTX3034.
The final candidate was not the whole story.
Earlier Quanta work emphasized active GTP-bound KRAS. Later, QTX3034 emerged with a different mechanism, binding the inactive GDP-bound state.
There was also QTX3046, another clinical candidate. It was chemically different and showed much tighter reported G12D binding than QTX3034.
It also reached Phase 1.
Then recruitment stopped and QTX3046 disappeared from the public pipeline.
The reason was not disclosed.
The patent history shows the same pattern at the chemistry level: many alternative structures, geometries and stereochemical solutions were explored, while only one branch remained visible as the development candidate.
I wrote more about that case in my first post on the QTX3034 story and in a follow-up on the chemistry behind it.
From one case to a general approach
That is what led me back to the broader question.
Instead of always expanding around the winning scaffold could we deliberately build alternative chemical-space sets around any reference molecule?
Vitas-M Laboratory has a large in-stock collection so technically this can be done for QTX3034 and in principle for almost any molecule of interest.
Would such an approach be useful in your research?
More specifically: would you want to see a structurally different, property-matched set built around one of your own reference molecules?