3D Binding Site Detection With Rotation-Invariant Molecular Features

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for detecting molecule binding sites, particularly protein binding pockets, suffer from reduced accuracy due to the reliance on voxel features that require manual design and lack of rotation invariance, leading to inconsistent detection results when the molecule is rotated.

Innovation Solution

The method employs point cloud data of the target molecule, extracting rotation-invariant location features using first and second target points to construct a rotation-invariant location feature, which is used as input for a site detection model like GCN to predict binding sites accurately.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If voxel features are used for binding site detection, then the detection process can be automated, but the accuracy decreases due to lack of rotation invariance and manual design requirements

Engineering Contradiction:
Improveautomation of binding site detectionVSAvoiddetection accuracy
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The patent transforms the molecular representation from fixed voxel grids to rotation-invariant point cloud features. By changing the parameter representation from grid-based to coordinate-based with rotation-invariant calculations (using spherical harmonics and invariant distance metrics), the system achieves both automation and high accuracy simultaneously, resolving the contradiction between automation extent and measurement precision

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the manual voxel feature design process with an automated deep learning-based feature extraction system. The neural network automatically learns optimal features from raw molecular coordinates, substituting the mechanical manual design process with an intelligent automated system that achieves superior accuracy without sacrificing automation

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Stability of the object's composition

If rotation-invariant features are implemented, then detection results become consistent under molecular rotation, but the feature extraction complexity increases

Engineering Contradiction:
Improveconsistency of detection resultsVSAvoidfeature extraction complexity
Core Design Contradiction:
Stability of the object's compositionVSDevice complexity

Solution Approach 1:

The patent introduces spherical harmonics and invariant distance metrics as intermediary mathematical tools that bridge the gap between raw molecular coordinates and rotation-invariant features. These intermediaries automatically handle the rotation invariance requirement through their mathematical properties, reducing the apparent complexity of the feature extraction process while ensuring detection result consistency

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the parameter representation from raw 3D coordinates to rotation-invariant parameters (spherical harmonics coefficients, invariant distances). This parameter transformation inherently provides rotation invariance, making the detection results consistent under molecular rotation while the complexity is managed through efficient mathematical formulations

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3920188B1Molecular binding site detection method and apparatus, electronic device and storage medium
Publication Date: 2026.02.11 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3920188B1 patent drawingFigure 1
  • EP3920188B1 patent drawingFigure 2
  • EP3920188B1 patent drawingFigure 3

AI summary

This disclosure discloses a method and apparatus for detecting a molecule binding site, an electronic device, and a storage medium, relating to the field of computer technologies. In this disclosure, three-dimensional (3D) coordinates of each site in a target molecule are obtained, a first target point and a second target point corresponding to the each site are determined, and a rotation-invariant location feature in the 3D coordinates of the each site is further extracted. A site detection model is invoked to perform prediction on the extracted location feature, to obtain a prediction probability of the each site being a binding site, so as to determine a binding site based on the prediction probability. The first target point and the second target point are associated with each site and have spatial representativeness to some extent, which is conducive to constructing a rotation-invariant location feature that can completely reflect the detailed structure of the target molecule, thereby avoiding loss of details caused by designing a voxel feature for the target molecule, and improving the accuracy of a process of detecting a molecule binding site.