English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep generative models are attracting great attention for molecular designwith desired properties. Most existing models generate molecules bysequentially adding atoms. This often renders generated molecules with lesscorrelation with target properties and low synthetic accessibility. Molecularfragments such as functional groups are more closely related to molecularproperties and synthetic accessibility than atoms. Here, we propose afragment-based molecular generative model which designs new molecules withtarget properties by sequentially adding molecular fragments to any givenstarting molecule. A key feature of our model is a high generalization abilityin terms of property control and fragment types. The former becomes possible bylearning the contribution of individual fragments to the target properties inan auto-regressive manner. For the latter, we used a deep neural network thatpredicts the bonding probability of two molecules from the embedding vectors ofthe two molecules as input. The high synthetic accessibility of the generatedmolecules is implicitly considered while preparing the fragment library withthe BRICS decomposition method. We show that the model can generate moleculeswith the simultaneous control of multiple target properties at a high successrate. It also works equally well with unseen fragments even in the propertyrange where the training data is rare, verifying the high generalizationability. As a practical application, we demonstrated that the model cangenerate potential inhibitors with high binding affinities against the 3CLprotease of SARS-COV-2 in terms of docking score.

Fragment-based molecular generative model with high generalization ability and synthetic accessibility