Generating anti-infective design spaces for selecting drug candidates
Inventors
Lee, Francis • STECKBECK, Jonathan D. • Holste, Hannes
Assignees
Publication Number
US-11424008-B2
Publication Date
2022-08-23
Expiration Date
2041-05-13
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
In one aspect, a method includes generating a design space for a peptide for an application. The generating includes identifying sequences for the peptide, and updating the sequences by determining, for each of the sequences, a respective set of activities pertaining to the application. The updating produces updated sequences each having updated respective activities. The method includes generating, based on the updated sequences, a solution space within the design space. The solution space includes a target subset of the updated sequences. The method includes performing, using a machine learning model to process the solution space, trials to identify a candidate drug compound that represents a sequence having a level of activity that exceeds a threshold level, and transmitting information describing the candidate drug compound to a computing device.
Core Innovation
The invention provides a method for generating a design space for a peptide related to a specific application. This method involves identifying a plurality of peptide sequences and updating these sequences by determining, for each sequence, a respective set of activities pertaining to the intended application. The result is an updated group of sequences, each having updated respective activities, forming a more comprehensive design space.
Based on these updated sequences, a solution space is generated within the design space, consisting of a target subset of sequences selected for their particular activities. The method uses a machine learning model to process this solution space, running trials to identify a candidate drug compound (a sequence) that achieves an activity level exceeding a defined threshold. Information describing the identified candidate drug compound is then transmitted to a computing device for further use.
The problem addressed is that traditional drug discovery techniques are limited in scope, inefficient, or unable to discover therapeutics for certain diseases due to constrained design spaces or overwhelming biological data. By enabling the analysis and search of expanded design spaces using advanced machine learning approaches, including multiple model types and dimensionality reduction, the invention seeks to improve the discovery and selection of drug candidates, especially designed peptides, for varied applications.
Claims Coverage
The claims disclose several inventive features, particularly around generating and processing peptide design spaces with machine learning models, defining and using solution spaces, comparing model outputs, and transmitting candidate compound information.
Generating and updating peptide design spaces based on activity profiles
The method involves generating a design space for a peptide by identifying multiple sequences and updating each sequence by determining a set of biomedical and/or biochemical activities relevant to a particular application. This results in an updated set of sequences, each annotated with their respective activity profiles.
Solution space identification within the design space
A solution space is created within the broader peptide design space. This solution space is a target subset of the updated sequences, each selected for having the updated respective activities that fit application-specific criteria or constraints (such as query parameters or threshold activity levels).
Machine learning model-driven trial identification of candidate drug compounds
A machine learning model processes the solution space to perform one or more trials. These trials are aimed at identifying a candidate drug compound as a sequence demonstrating at least one level of activity that meets or surpasses one or more specified threshold levels.
Comparative selection between multiple machine learning models based on candidate values
After identifying candidate drug compounds using different machine learning models, the system determines values (such as performance metrics or scores) for each candidate. The method compares these values and, based on the comparison, selects one of the machine learning models for continued use in identifying candidate drug compounds.
Transmitting candidate compound information to a computing device
Upon identifying a suitable candidate drug compound via machine learning model evaluation, the system transmits descriptive information about the candidate drug compound to a computing device, thereby integrating the discovery process with external analysis or manufacturing processes.
The inventive features center around an end-to-end approach for expanding, evaluating, and utilizing peptide design spaces with the aid of machine learning, including processing activity-annotated sequences, defining targeted solution spaces, selecting optimal candidates via model-based comparisons, and delivering candidate information for practical implementation.
Stated Advantages
The system enables efficient searching of expanded design spaces by incorporating diverse drug information, reducing computational burden when handling large datasets.
The architecture is capable of generating superior candidate drugs with desirable features not accessible through conventional, smaller design spaces.
Application of machine learning, including causal inference and advanced neural networks, allows identification and optimization of candidate drug compounds with increased scalability, accuracy, and efficiency.
Resource consumption is reduced by narrowing the set of viable candidate drug compounds before classification, thus saving time, processing power, memory, and network bandwidth.
The platform can continuously improve as it learns from updated data, enabling recursive tuning and enhancement of model performance and drug candidate quality over time.
Algorithmically designed drug compounds can be validated in vivo and in vitro for broad-spectrum activity and improved resistance profiles.
Custom tailoring of machine learning model packages for third parties is supported, addressing specific market or data-type needs.
Documented Applications
Discovery and optimization of peptide therapeutics for anti-infective, anti-cancer, antimicrobial, anti-viral, anti-fungal, anti-inflammatory, and anti-prionic applications.
Design and development of functional biomaterials such as adhesives, sealants, binders, chelates, diagnostic reporters, and combinations thereof.
Development of structural biomaterials including biopolymers, encapsulation films, flocculants, and desiccants.
Use in animal health and veterinary applications, such as treatment of bovine mastitis.
Industrial uses such as anti-biofouling or optimization of machinery control sequences.
Markets for new therapeutic indications, including eczema, inflammatory bowel disease, Crohn's Disease, rheumatoid arthritis, asthma, autoimmune diseases, inflammatory disease progressions, and oncology treatments.
Application in video game industry to optimize decision-making sequences for non-player characters using similar machine learning methodologies.
Use in integrated circuit and chip industries for optimization of mask works generation and efficient system-on-chip routing.
Interested in licensing this patent?