Each of the proposed solutions will be assessed in a fully objective, automated manner. First, we will evaluate the submission for the basic properties:

  • Format  to ensure solution is valid (SMILES <-> SDF conversion, fixing common, fixable errors, checking for chemical sanity, you would be surprised how many pentavalent carbons lurk in cheminformatics libraries. Pentavalent: having five bonds. Carbons have four bonds. At most.)
  • Novelty testing new drugs is resource-consuming, we will ensure that submitted solution is novel (not present in the databases and does not replicate a former solution)
  • Safety: We need drugs that can be readily deployed. Is the drug approved? Has it been tested in humans? Does it look like it is toxic? Does it violate known rules for drug-like compounds (e.g. Lipinski’s Rule of 5, see FAQs )? Compounds that do not satisfy these criteria will still be tested, but may be ranked lower. 

We will evaluate every submission that is properly formatted, novel and sane. The evaluation will be based on macromolecular docking:

  • Evaluation by docking will use AutoDock Vina
    • It is free and fast! 
    • You can check your solution yourself!
  • We will first do a restrained docking into a binding pocket.
    • Does the drug bind, if we tell it where we expect it to bind?
  • We will follow it with unrestrained docking.
    • Does the drug bind, where we expect it to? 
    • Can it find the binding pocket? 
    • Enrichment in binding pose. Does the drug bind consistently in the same pose? 
  • We will compare the new drug with other solutions (experimental and theoretical)
    • Is this solution consistent with what we know about this protein and drugs in general?
    • Is it a better binder?

The proposed molecules, which perform best in the automated screening, will be tested in physics-based simulations (molecular dynamics), to assess the probability of the new molecule being functional against its planned target. We will also assess the successful proposals in terms of their off-target effects