0

Why the GANs (Generative adversarial Network) trained on QM9 dataset (contains 134K molecules but none of them is complete to be eligible for drug-like molecule) produces drug-like molecules.

sine GANs trained on cats images will generate fake images of cats only then why in-complete drug-like sample of QM9 fed into GANs can produce complete drug-like molecules?

Vinay Sharma
  • 101
  • I dispute the premise of this question, have you seen: https://chemistry.stackexchange.com/q/135122/73620 https://paperswithcode.com/sota/drug-discovery-on-qm9 https://www.researchgate.net/publication/342093683_GEOM_Energy-annotated_molecular_conformations_for_property_prediction_and_molecular_generation https://academic.oup.com/bib/article/21/3/919/5498046 - the QM9 Dataset contains molecules which are sufficiently complete. – Rob Sep 15 '22 at 14:16
  • @Rob https://chemistry.stackexchange.com/questions/167632/are-the-133-885-molecules-in-the-qm9-database-drug-like – Vinay Sharma Sep 16 '22 at 03:36
  • Vinay, Karsten's answer: "... No, **probably** not a single one because they are too small. ...". --- Geoff's answer: "... **Almost** ***none*** of these are "drug-like" by the usual rules (e.g., Lipinski Rule-of-Five).". --- See also: https://www.researchgate.net/publication/274256239_Quantum_chemistry_structures_and_properties_of_134_kilo_molecules and section 4.2.4 of https://www.frontiersin.org/articles/10.3389/fphar.2021.827606/full --- PS: You should include the link to your question on Chem.SE in the above question, so people know that you have asked on more than one site. – Rob Sep 16 '22 at 07:56

0 Answers0