Virtual Landscape for Evaluating Chemical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems are unable to generate new chemical compounds in a low-dimensional space and fail to identify absent chemical structures that conform to a reduced dimensional space, particularly in the context of biologic identifiers like nucleic acid or protein sequences.
Innovation Solution
A computer-implemented method and system that constructs a virtual n-dimensional manifold to evaluate and predict new chemical and biologic entities. This involves converting biologic sequences and chemical identifiers into coded forms, generating an n-dimensional map using unsupervised learning, and filtering out distant coded forms to rank them based on similarity and proximity to known entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If dimensionality reduction techniques are used to map high-dimensional data to low-dimensional space for visualization, then the data can be visually represented and analyzed more easily, but the system cannot generate new chemical entities or identify absent structures in the reduced space
Solution Approach 1:
The patent applies dimensionality reduction techniques (PCA, t-SNE, UMAP) to map high-dimensional chemical descriptor space to low-dimensional visual representation (2D/3D), enabling easy visualization while preserving structural relationships. This resolves the contradiction by allowing both visualization and new entity generation through the same transformed coordinate system.
Solution Approach 2:
The patent introduces an intermediary coordinate transformation system that maps between high-dimensional chemical descriptor space and low-dimensional visual space. This intermediary layer enables both visualization and generation of new chemical entities by operating in the transformed coordinate system, bridging the gap between visual analysis and creative generation capabilities.
2Measurement precision
If the system focuses on visualizing existing chemical structures in reduced dimensional space, then visualization is achieved, but the system fails to identify absent chemical structures that conform to the reduced dimensional space
Solution Approach 1:
The patent performs preliminary mapping of the complete reduced-dimensional space using existing chemical structures, identifying gaps and unoccupied regions before generating new entities. This preliminary action enables systematic discovery of absent structures by comparing the mapped landscape against the target chemical space, improving both identification accuracy and discovery productivity.
Solution Approach 2:
The system uses its own visualization and mapping capabilities to automatically identify gaps and generate new chemical entities without external intervention. The reduced-dimensional space mapping serves the dual purpose of visualization and guiding new entity generation, allowing the system to self-improve its chemical discovery capabilities through iterative refinement.
3Quantity of substance
If the system processes large variable datasets with multiple descriptors, then comprehensive chemical analysis is achieved, but the computational complexity increases and requires dimensionality reduction for manageable visualization
Solution Approach 1:
The patent segments the high-dimensional chemical descriptor data into multiple dimensions (0D constitutional descriptors, 1D structural fragments, 2D graph variants, 3D quantum-chemical descriptors, 4D GRID descriptors) and processes each dimension separately before integrating them into a unified reduced-dimensional representation. This segmentation manages computational complexity while preserving comprehensive chemical information.
Solution Approach 2:
The patent transforms high-dimensional chemical descriptor data into low-dimensional visual representations using dimensionality reduction techniques, converting complex multi-variable datasets into manageable 2D/3D visualizations. This dimensionality change reduces computational complexity while maintaining the essential relationships and patterns in the comprehensive chemical data.
Data Source
AI summary
The present invention is directed to generating structured text document from the output of a query for compounds using a n-dimensional map where the n-dimensional map includes an arrangement of biological or chemical entities enumerated within a collection of documents describing a particular biological target of interest.