Neural Network Biopolymer Sequence Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing in silico modeling techniques are cumbersome, slow, and inefficient for generating biopolymer sequences that form complexes through binding, as they rely on physics-based models and search algorithms.
Innovation Solution
The use of a neural network-based system that embeds a graph representation of biopolymer structures, processed by a graph neural network or equivariant neural network, to generate associated biopolymer sequences that conform to a reference structure, employing a conditional generative model and energy landscape for efficient sequence prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based models and search algorithms are used to generate biopolymer sequences, then sequence accuracy can be achieved, but the process becomes cumbersome, slow, and inefficient
Solution Approach 1:
The patent replaces physics-based models and search algorithms (mechanical/computational systems) with a neural network-based system. The neural network is trained on biopolymer sequence-structure pairs and can generate sequences that conform to target backbone structures, achieving both accuracy and speed by learning patterns from data rather than performing exhaustive physics-based searches
Solution Approach 2:
The patent changes the approach from deterministic physics-based calculations to probabilistic neural network predictions. By training the neural network on large datasets of known biopolymer structures and sequences, the system learns to predict sequences that are likely to fold into target structures, transforming the problem from a computational search to a pattern recognition task
2Ease of manufacture
If existing in silico modeling techniques are used, then some sequence generation capability is provided, but the methods are cumbersome and inefficient
Solution Approach 1:
The patent replaces complex physics-based modeling workflows with a trained neural network system. Once trained, the neural network can generate sequences through straightforward forward propagation, eliminating the need for iterative physics-based simulations and manual optimization steps
Solution Approach 2:
The patent performs preliminary training of the neural network on large datasets of biopolymer sequences and structures before actual sequence generation. This pre-training phase captures essential structure-sequence relationships, enabling rapid and simple sequence generation later without requiring complex runtime computations
Data Source
AI summary
In some embodiments, methods and corresponding systems are disclosed for providing associated biopolymer sequence(s) to conform to a reference structure. The reference structure includes a target complex and the one or more associated biopolymer sequences. The biopolymer sequences are obtainable by the method, including embedding a graph representation using a neural network. The graph representation is featurized from the reference structure and includes a topology of the biopolymer with monomers as nodes and interactions between monomers as edges. The methods, in certain embodiments, further include processing the graph representation with a graph neural network or equivariant neural network that iteratively updates node and edge embeddings with a learned parametric function. The methods may further include converting the embedded graph representation to an energy landscape using a decoder. The methods can further include obtaining one or more biopolymer sequences from the energy landscape.