Deep Learning Model for Protein Complex Inter-Chain Distance Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current deep learning methods struggle to accurately predict inter-chain distances and quaternary structures of protein complexes due to a lack of advanced techniques specifically designed for quaternary structure applications, resulting in low prediction accuracy.

Innovation Solution

A deep learning system and method that uses a 2D attention-powered residual network to predict inter-chain distance maps for protein-related complexes, incorporating tertiary structural features and multiple sequence alignment-derived features, and generates 3D structures using gradient descent optimization or deep reinforcement learning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If current deep learning methods are used to predict inter-chain distances and quaternary structures, then the prediction task can be performed, but the prediction accuracy is low

Engineering Contradiction:

Improveprediction accuracyVSAvoidavailability of advanced techniques for quaternary structure

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The prediction task is divided into two separate deep learning models: one model predicts intra-chain distance maps for individual monomers, and another model predicts inter-chain distance maps for protein complexes. This segmentation allows each model to specialize in its specific task, improving overall prediction accuracy for both tertiary and quaternary structures.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system combines multiple types of input features including tertiary structural features (intra-chain distance maps), multiple sequence alignment-derived features (co-evolutionary scores, PSSM), and predicted contact maps from other methods. By merging these diverse features, the model achieves higher prediction accuracy that leverages complementary information from different sources.

Inventive Principle:

Principle #5Merging (Combining)

2Measurement precision

If advanced deep learning techniques are developed specifically for quaternary structure applications, then prediction accuracy improves, but the device complexity increases

Engineering Contradiction:

Improveinter-chain contact precisionVSAvoidmodel architecture complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The deep learning models are designed with universal architectures that can handle both intra-chain and inter-chain distance predictions using similar computational approaches. The same model framework processes different input feature types (structural, sequence, evolutionary) and produces distance predictions, making the system multi-functional while maintaining manageable complexity through code reusability and consistent design patterns.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20230154561A1Deep learning systems and methods for predicting structural aspects of protein-related complexes

Publication Date: 2023.05.18 THE CURATORS OF THE UNIVERSITY OF MISSOURI

AI summary

Deep learning systems and methods for predicting structural aspects of protein-related complexes are described herein. An example method for predicting inter-chain distances of protein-related complexes includes receiving data associated with a protein-related complex, where the data associated with the protein-related complex includes at least one of a tertiary structural feature and a multiple sequence alignment (MSA)-derived feature. The method also includes inputting the data associated with the protein-related complex into a deep learning model. The method further includes predicting, using the deep learning model, an inter-chain distance map for the protein-related complex.