LOGICS: Learning optimal generative distribution for designing de novo chemical structures

Bae, Bongsung; Bae, Haelee; Nam, Hojung

doi:10.1186/s13321-023-00747-3

Journal of Cheminformatics

Table 2 Comparison of bioactivity optimization performance for KOR and PIK3CA with various GPC methods

From: LOGICS: Learning optimal generative distribution for designing de novo chemical structures

	Prior	VGPC	Segler	REINVENT	DrugEx	AHC	AugMem	LOGICS
KOR
Validity^a,α	0.95 ± 0.00	0.93 ± 0.02	0.96 ± 0.00	0.98 ± 0.00	0.97 ± 0.00	0.97 ± 0.00	0.96 ± 0.00	0.98 ± 0.00
Unique^b,α	0.99 ± 0.00	0.99 ± 0.00	0.94 ± 0.00	0.90 ± 0.00	0.99 ± 0.00	0.87 ± 0.00	0.96 ± 0.00	0.99 ± 0.00
Novelty^c,α	0.94 ± 0.00	0.99 ± 0.00	0.99 ± 0.00	0.94 ± 0.00	0.99 ± 0.00	0.95 ± 0.00	0.97 ± 0.00	0.98 ± 0.00
Diversity^d,α	0.88 ± 0.00	0.79 ± 0.00	0.85 ± 0.00	0.86 ± 0.00	0.83 ± 0.00	0.86 ± 0.00	0.86 ± 0.00	0.85 ± 0.00
PredAct^e,β	5.95 ± 0.00	7.04 ± 0.00	8.30 ± 0.00	7.04 ± 0.01	7.10 ± 0.00	7.16 ± 0.00	7.00 ± 0.00	7.57 ± 0.00
PwSim^f,β	0.11 ± 0.00	0.10 ± 0.00	0.12 ± 0.00	0.12 ± 0.00	0.12 ± 0.00	0.12 ± 0.00	0.12 ± 0.00	0.13 ± 0.00
FCD^g,β	27.2 ± 0.02	38.8 ± 0.06	22.3 ± 0.03	26.0 ± 0.06	30.4 ± 0.06	24.6 ± 0.05	26.0 ± 0.06	22.2 ± 0.01
OTD^h,β	5.37 ± 0.00	5.85 ± 0.00	5.09 ± 0.00	5.11 ± 0.00	5.37 ± 0.00	5.23 ± 0.00	5.27 ± 0.00	4.95 ± 0.00
PIK3CA
Validity^a,α	0.95 ± 0.00	0.85 ± 0.00	0.97 ± 0.00	0.99 ± 0.00	0.98 ± 0.00	0.97 ± 0.00	0.98 ± 0.00	0.99 ± 0.00
Unique^b,α	0.99 ± 0.00	0.99 ± 0.00	0.94 ± 0.00	0.65 ± 0.00	0.99 ± 0.00	0.91 ± 0.00	0.92 ± 0.00	0.71 ± 0.00
Novelty^c,α	0.94 ± 0.00	0.99 ± 0.00	0.99 ± 0.00	0.93 ± 0.00	0.99 ± 0.00	0.96 ± 0.00	0.97 ± 0.00	0.99 ± 0.00
Diversity^d,α	0.88 ± 0.00	0.82 ± 0.00	0.78 ± 0.00	0.79 ± 0.00	0.80 ± 0.00	0.82 ± 0.00	0.81 ± 0.00	0.73 ± 0.00
PredAct^e,β	6.84 ± 0.00	8.05 ± 0.00	8.75 ± 0.00	8.83 ± 0.00	8.39 ± 0.00	8.01 ± 0.00	7.99 ± 0.00	9.54 ± 0.00
PwSim^f,β	0.10 ± 0.00	0.11 ± 0.00	0.11 ± 0.00	0.17 ± 0.00	0.11 ± 0.00	0.10 ± 0.00	0.10 ± 0.00	0.18 ± 0.00
FCD^g,β	41.0 ± 0.08	43.7 ± 0.06	45.7 ± 0.02	32.7 ± 0.08	44.1 ± 0.11	51.0 ± 0.07	50.9 ± 0.03	29.4 ± 0.10
OTD^h,β	5.99 ± 0.01	5.93 ± 0.00	5.78 ± 0.01	4.47 ± 0.02	5.88 ± 0.01	5.94 ± 0.00	5.97 ± 0.00	4.27 ± 0.02

Bold represents the best-performing value among the methods
^aValidity is the ratio of valid generations to 20,000 generations from the model
^bUniqueness is the ratio of unique generations to the valid generations
^cNovelty is the ratio of unique generations that are not found in the pre-training dataset
^dDiversity measures how dissimilar the 1,000 generations are
^ePredAct is the mean of predicted activities of the valid generations
^fPwSim is the mean of pairwise similarities between generations and test set activities
^gFCD is the Fréchet Chemnet Distance between the generations and the test set activities
^hOTD is the optimal transport distance between generations and the test set activities
^αStandard metrics
^βOptimization metrics

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com