ChemMORT

ChemMORT (Chemical Molecular Optimization, Representation and Translation) is a platform for de novo molecular optimization by navigating chemical space based on neural machine translation and inverse QSAR methodology. Based on reversible molecular representation and multi-objective particle swarm optimization, ChemMORT could effectively optimize any given molecule toward some predefined ADMET endpoints with the preservation of potency by integrating the pre-trained ADMET prediction models, substructural constraint and similarity constraint.

Calculation

SMILES Encoder

The SMILES Encoder allows users to transform a SMILES string to a 512-dimensional vector through the well-trained encoder network based on the Neural Machine Translation model. Owing to its consecutive, reversible, and informative features, such representation is recommended to define the chemical space of molecules.

Descriptor Decoder

The Descriptor Decoder can back-engineer the 512-dimensional vector to the corresponding uniform canonical SMILES string through the well-trained decoder network based on the Neural Machine Translation model. Such function provides a steered solution for de novo molecular optimization, as the possibility output of the decoder can be sampled.

Molecular Optimizer

The Molecular Optimizer, which combines the neural machine translation model and multi-objective particle swarm optimization, is designed to optimize the ADMET properties of molecules based on credible ADMET prediction models and the customized substructural constraints, thus effectively improving the drug-likeness of leads without the loss of potency.

Highlights