Query-Based Molecule Optimization with Latent-Space Guided Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current molecule optimization methods face challenges in designing drugs for new emerging viruses due to insufficient training data, multiple constraints, and the difficulty of optimizing discrete molecular sequences without changing chemical properties, especially when using black-box models.
Innovation Solution
A query-based molecular optimization (QMO) system that decouples representation learning and guided search, using zeroth-order optimization to efficiently optimize molecule sequences with continuous latent representations, incorporating black-box models and chemical property predictors to satisfy multiple constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If translation-based approaches are used for molecule optimization, then optimization can be performed with paired training sequences, but insufficient training data is available for new emerging viruses
Solution Approach 1:
The patent introduces an intermediary objective function that bridges the gap between sequence optimization and molecular property prediction. This intermediary allows the system to optimize molecular sequences without requiring paired training data, as it uses the predicted molecular properties as the optimization target rather than direct sequence-to-sequence translation
Solution Approach 2:
The system performs preliminary training of molecular property prediction models using available data before performing the actual optimization. This preliminary action allows the optimization process to leverage pre-learned molecular representations and property relationships, reducing the need for task-specific paired training data
2Reliability
If multiple constraints are imposed on molecule optimization, then useful real-world drugs can be designed, but the optimization process becomes more complex
Solution Approach 1:
The patent merges multiple constraint satisfaction requirements into a single composite objective function. This unified objective function simultaneously optimizes for binding affinity, toxicity, and other molecular properties, reducing the complexity of managing multiple separate constraints while maintaining all necessary requirements for drug utility
Solution Approach 2:
The system changes the parameter space by working with continuous latent representations of molecular sequences rather than discrete sequences. This allows for smoother optimization landscapes and more efficient gradient-based optimization, making it easier to satisfy multiple constraints simultaneously
3Ease of operation
If discrete sequence representations are used for molecules, then optimization can be performed on molecular sequences, but direct gradient descent becomes difficult to implement
Solution Approach 1:
The patent transitions from optimizing discrete molecular sequences directly to optimizing continuous latent representations that encode the sequences. This dimensional change allows gradient descent to operate in the continuous latent space, avoiding the difficulty of implementing gradients on discrete sequence spaces while still producing optimized molecular sequences through decoding
Data Source
AI summary
A query-based generic end-to-end molecular optimization (“QMO”) system framework, method and computer program product for optimizing molecules, such as for accelerating drug discovery. The QMO framework decouples representation learning and guided search and applies to any plug-in encoder-decoder with continuous latent representations. QMO framework directly incorporates evaluations based on chemical modeling, analysis packages, and pre-trained machine-learned prediction models for efficient molecule optimization using a query-based guided search method based on zeroth order optimization. The QMO features efficient guided search with molecular property evaluations and constraints obtained using the predictive models and chemical modeling and analysis packages. QMO tasks include optimizing drug-likeness and penalized log P scores with similarity constraints and improving the target binding affinity of existing drugs to pathogens such as the SARS-CoV-2 main protease protein while preserving the desired drug properties. QMO tasks further improves optimizing antimicrobial peptides toward lower toxicity.