Skip to content

GitLab

Closed
Created by Miquel Duran-Frigola@mduranDeveloper

Raws for the chemical layer

The outmolecules/outmolecules.py script should help you get an idea.

The actual "fingerprinters" (i.e. way to create raw data) are placed at chemutils/raws.py.

Remember:

  1. Do the union set of inchikeys of all the previous steps (tables).
  2. fp2d, fp3d, scaffolds, subskeys and physchem tables are persistent. We don't have to calculate them every times. Therefore, only do the calculations for new molecules. fp3d and physchem are the slowest. Especially in the case of fp3d, calculations may fail due to the complexity of some molecules. In such a case, it could be interesting to store the inchikey in the fp3d raws table, with a NULL value, so that we don't attempt the prediction every time we update the CC.