[Software paper] Lacking mention of how features are computed

JustinYKC / FEPOPS

Python implementation of the FEPOPS molecular descriptors

MIT License

5 stars 0 forks source link

[Software paper] Lacking mention of how features are computed #16

Closed exs-cbouy closed 1 year ago

exs-cbouy commented 1 year ago

Refers to the JOSS submission review: https://github.com/openjournals/joss-reviews/issues/5763

The paper is missing one or two sentences on how the features are computed:

partial charges: Gasteiger charges calculated with RDKit
Atomic logP contribution: Crippen method as implemented in RDKit
Hbond donors and acceptors: SMARTS patterns taken from DAYLIGHT

Since the SMARTS pattern are taken from the DAYLIGHT website, they must be added to the references. (and ideally the RDKit as well)

JustinYKC commented 1 year ago

We have expanded the paper (paper, readme and docs - 6280684620464c95da3f80ff766942eb1c86e472) with an explanation of how atomic properties are calculated with RDKit (ref added) – using the Gasteiger algorithm (reference added) for atomic partial charges, Crippen method (reference added) for atomic logP contributions and referenced the paper from which the HBA+HBD acceptor patterns were taken - https://pubs.acs.org/doi/10.1021/ci970431+ . This has been merged into main in 2146bcf0b2737ea671cb4beed6cbfc5632f8baee and is present in OpenFEPOPS v 1.8.0.

exs-cbouy commented 1 year ago

There seem to be an issue with the reference for Gasteiger charges, it should be Gasteiger & Marsili instead of Wildman & Crippen.

stevenshave commented 1 year ago

Thank you, this was certainly an error on my part. The Gasteiger reference has now been properly inserted in 8303a1815f5a176d6c9c6aefcf4d01035d4ee333.