Journal of Cheminformatics

Table 4 Summary of REINVENT 4 priors. Mol2Mol comes with six different priors with pairs trained on different types of similarity

From: Reinvent 4: Modern AI–driven generative molecule design

Generator	Dataset	Notes
Reinvent	ChEMBL 25	Published in Ref. [19, 20]
Libinvent	ChEMBL 27	Published in Ref. [46]
Linkinvent	ChEMBL 27	Published in Ref. [47]
Mol2Mol	ChEMBL 28	Published in Ref. [22]
		Similarity^a
		Medium similarity^b
		High similarity^c
		Scaffold^d
		Generic scaffold^e
		Matched molecular pairs^f
Mol2Mol	Pubchem^g	Published in Ref. [54]
		Similarity^g

^aTanimoto similarity \(\ge\) 0.5.
^b.5 \(\le\) Tanimoto similarity < 0.7.
^cTanimoto similarity \(\ge\) 0.7.
^dMolecules sharing the same Murcko scaffold (RDKit).
^eMolecules sharing the same unlabelled Murcko scaffold.
^fMatched molecular pairs have been extracted with mmpdb [77].
^gPubchem was collected in December 2021.
^hTanimoto similarity \(\ge\) 0.5 on ECFP4 fingerprints with counts

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com