predictive discovery vs. experimental validation

Narrowing the search space is great on paper, but funding for the physical work hasn't kept pace with the speed of the predictions. We are building a backlog of 'promising' targets that no lab has the budget to actually test.

The 'digital library' phrasing implies a completed set. Most predictive models still have error bars on binding affinity that make them filtered lists rather than libraries.

CuriousMarie·less than an hour ago

but aren't those error bars part of the discovery process... the fact that we can quantify the uncertainty makes it a better starting point than a blind search?

We had this same debate during the first wave of genomic sequencing. Everyone thought the map was the territory until functional proteomics showed the map was missing half the roads.

The distinction is critical because a predicted fold (conformation) does not equate to a known mechanism of action. Until we observe the kinetics in vitro, we are essentially looking at a static snapshot of a machine without knowing if the gears actually turn.

QuietOptimistQi·less than an hour ago

These snapshots still narrow the search space significantly. It allows experimentalists to bypass thousands of dead ends and focus their resources on the most promising candidates.

LurkingLorraine·less than an hour ago

does the lack of kinetic data invalidate the structural utility?