Skip to main content
. 2023 Feb 8;15:18. doi: 10.1186/s13321-023-00686-z

Fig. 5.

Fig. 5

Similarity reproduction abilities. Left: 2D structure of the respective reference compound. Middle: Histogram of similarities (calculated using the exact method) of the 100,000 closest molecules to the reference in latent space (”ranking” task). Right: Reproduction of fairly similar compounds to the reference where a threshold of 0.5 was chosen to distinguish between similar and dissimilar compounds (”hit identification” task). A analysis of the performance using a very large reference compound. B performance with a smaller, cyclized reference compound. C performance using a more linear compound with heterocycles