Protein Folding Prediction From Amino Acid Sequences Using MNNN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for designing protein sequences to form helical structures are inefficient and lack accuracy, particularly in traditional approaches that rely on classical molecular dynamics simulations, which are time-consuming and limited by sequence length, and there is a need for a more efficient and accurate tool to predict dihedral angles for protein folding.
Innovation Solution
A deep learning regression model, specifically a Multi-scale Neighborhood-based Neural Network (MNNN), is used to predict dihedral angles directly from amino acid sequences, incorporating both raw and secondary sequence neighborhoods, enabling fast and accurate prediction of protein structures without relying on templates or structural biological knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classical molecular dynamics simulations are used to predict protein folding, then structural accuracy can be achieved, but the computation time increases significantly from seconds to days or weeks
Solution Approach 1:
The patent replaces classical molecular dynamics simulations (mechanical/physical system) with a deep learning neural network model (information processing system). The MNNN model learns protein folding patterns from training data and predicts dihedral angles directly from amino acid sequences, eliminating the need for time-consuming physics-based simulations while maintaining prediction accuracy.
Solution Approach 2:
The patent performs preliminary action by training the neural network model on extensive protein structure data before actual prediction tasks. The model pre-learns the complex relationships between amino acid sequences and three-dimensional structures, enabling fast predictions without requiring repeated simulations during the actual prediction phase.
2Stability of the object's composition
If traditional protein design methods are used, then structural stability can be verified, but the sequence length is limited and scalability is poor
Solution Approach 1:
The patent creates a universal neural network model that can handle protein sequences of any length within computational limits. The MNNN model processes amino acid sequences through embedding layers and attention mechanisms that are not constrained by fixed sequence length, enabling it to predict structures for both short peptides and long protein chains with consistent accuracy.
3Productivity
If template-based methods are used for protein structure prediction, then prediction speed improves, but accuracy decreases for novel protein structures without templates
Solution Approach 1:
The patent implements self-service by enabling the neural network to predict protein structures de novo without relying on external templates or databases. The MNNN model uses only the amino acid sequence as input and internally learns the structural patterns, allowing it to accurately predict both known and novel protein structures without template bias.
Data Source
AI summary
A method, computer system, and a computer program product for designing one or more folded structural proteins from at least one raw amino acid sequence is provided. The present invention may include computing one or more character embeddings based on the at least one raw amino acid sequence by utilizing a multi-scale neighborhood-based neural network (MNNN) model. The present invention may then include refining the computed one or more character embeddings with at least one set of sequence neighborhood information. The present invention may further include predicting one or more dihedral angles based on the refined one or more character embeddings.