Generating enhanced graphical user interfaces for presentation of anti-infective design spaces for selecting drug candidates
Inventors
Lee, Francis • STECKBECK, Jonathan D. • Holste, Hannes
Assignees
Publication Number
US-11848076-B2
Publication Date
2023-12-19
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 is disclosed for presenting, on a computing device, a graphical user interface (GUI) of a therapeutic tool. The method includes presenting, in a first screen of the GUI, a design space for a protein for an application, where the design space includes a set of sequences, where each sequence contains a respective set of activities pertaining to the application. The method also includes receiving, via a graphical element in the first screen, a selection of one or more query parameters of the design space, and presenting, in a second screen of the GUI, a solution space that includes a subset of the set of sequences, where each sequence contains the respective set of activities, where the subset of the set of sequences is selected based on the one or more query parameters.
Core Innovation
The invention addresses limitations in conventional drug discovery methods, which often restrict the search to narrow design spaces, leading to inefficient, incomplete, or inadequate identification of effective therapeutics. Traditional methods struggle with processing vast and complex combinations of protein sequences and their biomedical or biochemical activities, making it challenging to discover optimal drug candidates efficiently.
This patent discloses a method for presenting, via a computing device, a graphical user interface (GUI) for a therapeutic tool that enables users to explore an enhanced design space for proteins. The design space includes a plurality of protein sequences generated by a machine learning model trained with multiple encodings and causal inference techniques, allowing consideration of alternative scenarios and sophisticated filtering. Each sequence in the design space is associated with a respective set of biomedical and biochemical activities relevant to a specific application, such as anti-infective or anticancer drug development.
The GUI comprises several interactive screens. On the first screen, users are presented with the entire design space. On the second screen, a solution space is shown, consisting of protein sequences filtered from the design space based on user-selected query parameters. Users can interactively select sequences and view detailed information—including sequence structure, correlation heatmaps, experimental data, probabilistic inference scores, and external data—on a candidate dashboard in a third screen. The method leverages advanced AI techniques, such as encodings and causal inference, to efficiently generate, display, and analyze large, multi-dimensional protein design spaces and their solution subsets for therapeutic applications.
Claims Coverage
The claims present five primary inventive features relating to the presentation of protein design spaces, user interaction through GUIs, the generation and filtering of protein sequences using machine learning models and causal inference, and the detailed candidate analysis dashboard.
Presenting interactive protein design and solution spaces in a GUI for therapeutic applications
A method is provided for presenting, on a computing device, a graphical user interface (GUI) that includes: - A first screen displaying a design space for a protein application, with the design space comprising a plurality of protein sequences, where each sequence is associated with a respective plurality of biomedical and/or biochemical activities. - The design space is generated from a machine learning model trained with multiple encodings and utilizing causal inference to evaluate alternative scenarios and filter the protein sequence superset. - A second screen is provided showing a solution space, which contains a subset of the protein sequences filtered according to one or more user-selected query parameters.
Machine learning generation of protein sequences using encodings and causal inference
The plurality of protein sequences in the design space are generated by a machine learning model that: - Is trained using a plurality of encodings reflecting different sequence properties and activities. - Uses causal inference to execute one or more alternative scenarios for filtering the superset and generating design space content. This enables efficient handling of large and complex protein design spaces with nuanced understanding of sequence-activity relationships.
User selection of sequences and display of detailed candidate dashboards
Allows users, through a graphical element on the second screen (solution space), to select a sequence from the subset of protein sequences. Upon selection, a third screen—a candidate dashboard—is presented, which provides comprehensive information related to the chosen sequence, including: - The sequence’s structural data - Correlation heatmaps - Experimental data - Probabilistic scores from inference models - External related data
Cluster visualization and energy correlation in solution space
In embodiments, the second screen includes: - A first portion showing one or more color-coded clusters representing the subset of protein sequences. - A second portion presenting data on these clusters, describing corresponding objects such as candidate drug compounds, activities, interactions, drugs, genes, pathways, physical descriptors, and more. The color-coded clusters represent, using energy correlation, each sequence in the subset, explaining how positions in the sequences correlate with each other, both within and across sequences.
Topographical mapping of solution spaces and incorporation of diverse user-defined parameters
The solution space can be presented as a topographical map within the GUI, where each indication on the map corresponds to the level of activity for a sequence at a specific point. The method also supports receiving user-defined query parameters—including biomedical and/or non-biomedical ontology terms—to dynamically filter and interpret the design space and solution space in multiple dimensions, encompassing two-dimensional, three-dimensional, or n-dimensional mathematical representations.
Collectively, these inventive features define an interactive, AI-powered GUI system for exploring, filtering, and analyzing complex multi-activity protein design spaces in therapeutic applications, leveraging machine learning encodings, causal inference, and advanced user visualization.
Stated Advantages
Enables efficient exploration of large, multi-dimensional protein design spaces by using machine learning models with advanced encodings and causal inference.
Facilitates identification and presentation of optimized protein sequences with desired biomedical or biochemical activities.
Provides interactive graphical user interfaces that allow for dynamic filtering, visualization, and selection of sequences based on user-specified query parameters and ontology terms.
Delivers comprehensive candidate analysis, including structure, experimental data, correlation heatmaps, and inference model scores, in a user-friendly dashboard.
Saves computing resources and time by distilling complex biological data into user-comprehensible, interactive formats for drug candidate selection.
Documented Applications
Drug discovery and development for anti-infective, anti-cancer, anti-viral, anti-fungal, anti-inflammatory, anti-cholinergic, anti-dopaminergic, anti-serotonergic, anti-noradrenergic, and anti-prionic applications.
Development of peptide therapeutics and biologics for treating human diseases and medical conditions.
Applications in animal health/veterinary industry, such as treatments for animal diseases (e.g., bovine mastitis).
Industrial applications, including anti-biofouling and generation of optimized control action sequences for machinery.
Design and optimization of structural and functional biomaterials, such as adhesives, sealants, biopolymers, encapsulation films, flocculants, and desiccants.
Business intelligence for matching protein sequences to target product profiles in pharmaceutical or biotechnology settings.
Interested in licensing this patent?