Macromolecule Interaction Prediction Negative Sample Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for predicting macromolecule-macromolecule interactions, such as protein-RNA interactions, face challenges due to the difficulty in generating high-quality negative samples, which affects the reliability of computational models used for classification.

Innovation Solution

A method is developed to generate a negative sample set by creating similarity maps for macromolecules of different types, generating vectorized representations of nodes, and determining interaction probabilities to produce a set of negative samples that are distinct from positive samples, thereby improving the accuracy of classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If random sampling is used to generate negative samples, then the generation process is simple and fast, but the quality and representativeness of negative samples deteriorates

Engineering Contradiction:

Improvespeed of negative sample generationVSAvoidquality of negative samples

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent performs preliminary actions by pre-computing similarity matrices and building knowledge graphs before sample generation. The similarity between macromolecules is pre-calculated and stored, allowing rapid retrieval during negative sample generation without performing complex calculations in real-time, thus maintaining both speed and quality.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary mechanism using similarity matrices and knowledge graphs as mediators between the positive sample set and the generated negative samples. These intermediaries capture essential similarity relationships, enabling the system to generate negative samples that are both diverse and representative by filtering through pre-computed similarity information.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If existing prediction models are used, then the overall prediction task can be addressed, but the accuracy deteriorates due to poor negative sample quality

Engineering Contradiction:

Improveapplicability of prediction modelsVSAvoidprediction accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent implements feedback mechanisms where the generated negative samples are used to retrain and refine the prediction models. The model continuously learns from the high-quality negative samples, adjusting its parameters to improve discrimination between positive and negative interactions, thereby enhancing prediction accuracy while maintaining adaptability.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent changes key parameters in the model training process by introducing carefully constructed negative samples with specific similarity characteristics. By adjusting the composition and properties of the training data (positive and negative samples), the model's decision boundaries are refined, leading to improved accuracy without sacrificing versatility.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If comprehensive similarity analysis is performed for all macromolecule pairs, then the quality of negative samples improves, but the computational complexity and time required increases

Engineering Contradiction:

Improvequality of negative samplesVSAvoidcomputational complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the macromolecule set into different groups based on pre-computed similarity metrics. Instead of analyzing all pairs comprehensively, the system divides the search space and applies targeted analysis to generate negative samples from specific similarity clusters, reducing computational complexity while maintaining sample quality.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs comprehensive similarity analysis as a preliminary action, pre-computing similarity matrices and building knowledge graphs before the actual negative sample generation. This upfront computation divides the complex task into manageable stages, where the heavy lifting is done beforehand, and only lighter operations are needed during sample generation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20240273351A1Method of generating negative sample set for predicting macromolecule-macromolecule interaction, method of predicting macromolecule-macromolecule interaction, method of training model, and neural network model for predicting macromolecule-macromolecule interaction

Publication Date: 2024.08.15 BOE TECHNOLOGY GROUP CO LTD

AI summary

A method of generating a negative sample set for predicting macromolecule-macromolecule interaction is provided. The method includes receiving a positive sample set including pairs of macromolecules of a first type and macromolecules of a second type having macromolecule-macromolecule interaction; generating a first similarity map of the macromolecules of the first type; generating a second similarity map of the macromolecules of the second type; generating vectorized representations of nodes in the first similarity map and vectorized representations of nodes in the second similarity map; and generating the negative sample set using the vectorized representations of nodes in the first similarity map and the vectorized representations of nodes in the second similarity map.