Page Comparison

Wrapping up in-progress work

• Coverage tool

• • JW – Would be willing to merge this today

  • AD – Based on my new understanding of OFFTK (that from_file, from_file_obj, and from_object do different things), the previous results may not be what we thought they were. So I worked on this some more, but there’s more work to be done. Also, I’m finding more interesting cases of failed comparisons, which I could write up.

  • (General) – AD will work on polishing this with his remaining time, and documenting how JW can re-run the toolkit molecule comparison to start debugging the different categories of failures.

• Coverage-based feature reduction --> Ready for review?

  • AD – This is ready for review+merge.

  • JW – I’ll take over this branch + PR, and will review + merge it when ready.

  • AD – This also doesn’t process molecules until they’re needed for a test, and handles datasets as dict-like objects containing the raw bytes of each record, instead of loading+processing the all molecules at once.

• Safely handling multicomponent, or otherwise unusual SMILES

  • Not necessary to do everything – We’d be happy to just get multicomponent filtering.

  • AD will not work on this, and instead work on wrapping up the coverage tool + providing instrucitons to rerun his analysis.

• Stereo issues (#1011) discussion

• Multicomponent molecule input

  • (General) – AD will open an issue describing the approach for this.

Safer processing of SMILES files?

• AD – from_smiles and from_file (given a SMILES file) are doing different things. The best solution would be to parse the SMILES file yourself and then feed each entry into from_smiles.

  • (General) – AD will write up an Issue describing the approach he’d take to this.

• AD – Two styles: OE and RDK

  • Variations:

    • Whether there’s a header (then we know what the columns will be)

    • If there’s no header - Each line is SMILES[whitespace]identifier

      • Some complication around whether the identifier can have a space in it

      • In CXSMILES the first whitespace will have a tab, then after that info may be split by spaces