System and method for the latent space optimization of generative machine learning models
A system and method for optimizing the latent space in generative machine learning models, and applications of the optimizations for use in the de novo generation of molecules for both ligand-based and pocket-based generation. The ligand-based optimizations comprise a tunable reward system based on a multi-property model and further define new measurable metrics: molecular novelty and uniqueness. The pocket-based optimizations comprise an initial multi-property optimization followed up by either a seed-based optimization or a relaxed-based optimization.
1. A system for optimizing the latent space in generative machine learning models, comprising:
a computer system comprising a memory and a processor;
a latent space optimizations module, comprising a plurality of programming instructions stored in the memory and operating on the processor, wherein the plurality of programming instructions, when operating on the processor, causes the computer system to:
train a plurality of machine learning models, each machine learning model trained to represent a particular property of a common object;
receive a seed object, wherein the seed object is the same type of object as the common object;
optimize the seed object's properties based on the plurality of machine learning models, wherein the plurality of optimized seed object properties forms a mixed objective; and
maximize the mixed objective by performing gradient descent on a latent vector encoded from the seed object.
2. The system of claim 1 , wherein one model in the plurality of machine learning models is a 3D model trained on ground truths extracted from cheminformatics software to predict synthetic accessibility.
3. The system of claim 1 , wherein one model in the plurality of machine learning models is a 3D model trained on ground truths extracted from cheminformatics software to predict drug-likeness.
4. The system of claim 1 , wherein one model in the plurality of machine learning models is a model trained on poses of ligands docked in protein binding pockets and finetuned on public protein databases to provide bioactivity affinity based on IC50 values.
5. The system of claim 1 , wherein one model in the plurality of machine learning models is a model trained on poses of ligands docked in protein binding pockets to determine whether a conformer's pose relative to a binding site is the correct docked pose.
6. The system of claim 1 , wherein one model in the plurality of machine learning models is a discriminator model which discriminates between generated and real molecules.
7. A method for optimizing the latent space in generative machine learning models, comprising the steps of:
training a plurality of machine learning models, each machine learning model trained to represent a particular property of a common object;
receiving a seed object, wherein the seed object is the same type of object as the common object;
optimizing the seed object's properties based on the plurality of machine learning models, wherein the plurality of optimized seed object properties forms a mixed objective; and
maximizing the mixed objective by performing gradient descent on a latent vector encoded from the seed object.
8. The method of claim 7 , wherein one model in the plurality of machine learning models is a 3D model trained on ground truths extracted from cheminformatics software to predict synthetic accessibility.
9. The method of claim 7 , wherein one model in the plurality of machine learning models is a 3D model trained on ground truths extracted from cheminformatics software to predict drug-likeness.
10. The method of claim 7 , wherein one model in the plurality of machine learning models is a model trained on poses of ligands docked in protein binding pockets and finetuned on public protein databases to provide bioactivity affinity based on IC50 values.
11. The method of claim 7 , wherein one model in the plurality of machine learning models is a model trained on poses of ligands docked in protein binding pockets to determine whether a conformer's pose relative to a binding site is the correct docked pose.
12. The method of claim 7 , wherein one model in the plurality of machine learning models is a discriminator model which discriminates between generated and real molecules.