Receptor-Ligand Learning Model Using 3D Protein Structure Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for determining receptor-ligand combinations focus on atomic-level chemical structural formulas, lacking optimal granularity and information for accurate estimation, particularly considering higher-order protein structures and coordinate information.
Innovation Solution
A non-transitory computer-readable recording medium stores a learning program that executes a process involving the generation of vector dictionaries from amino acid sequences and three-dimensional coordinate information, using machine learning to improve the estimation of receptor-ligand combinations by incorporating higher-order protein structures and coordinate data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models use atomic-level chemical structural formulas to determine receptor-ligand combinations, then the model can process molecular structure data, but the estimation accuracy is insufficient due to lacking optimal granularity and higher-order protein structure information
Solution Approach 1:
The patent segments the receptor structure into multiple primary structures (individual amino acid sequences) that are then processed separately. Each primary structure is represented as a vector, and these vectors are combined to form a comprehensive representation of the higher-order protein structure. This segmentation allows the model to capture both local atomic-level interactions and global structural context, resolving the contradiction between processing atomic details and incorporating higher-order structure information.
Solution Approach 2:
The patent transforms the representation of protein structures by adding a new dimension - converting primary structure sequences into vector representations that capture three-dimensional conformational information. This dimensional transformation enables the model to simultaneously utilize atomic-level chemical structural formulas and higher-order protein structure data, improving estimation accuracy without losing structural information.
2Device complexity
If machine learning models focus only on chemical structural formulas of receptors and ligands, then the processing is simplified, but the granularity and amount of information are not optimal for accurate estimation
Solution Approach 1:
The patent changes the parameters used to represent molecular structures by introducing vector representations of primary structures. Instead of using only traditional chemical structural formulas, the model transforms amino acid sequences into numerical vectors that encode both sequence information and three-dimensional conformational data. This parameter change enables the model to process more informative data while maintaining computational feasibility through efficient vector operations and machine learning algorithms.
3Measurement precision
If the model incorporates higher-order protein structures and coordinate information, then the estimation accuracy improves, but the data processing and model complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-processing the protein structure data into vector representations before feeding it to the machine learning model. The primary structures are converted into vectors that encode both sequence and conformational information in advance, so that during the actual prediction phase, the model only needs to perform efficient vector operations rather than processing raw structural data from scratch. This preliminary transformation reduces the computational burden during inference while maintaining high estimation accuracy.
Data Source
AI summary
A non-transitory computer-readable recording medium has stored therein a learning program that causes a computer to execute a process including acquiring teacher data associating input data including a plurality of primary structures and structure information of the plurality of primary structures with labels, the plurality of primary structures being included in a higher-order structure of a receptor combined with a ligand, the label indicating whether the receptor and the ligand are combinable with each other and executing machine learning of a machine learning model based on the teacher data.