Peptide Mutation Policy for MHC Binding Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational tools lack the ability to generate new peptide sequences with specified binding properties for MHC proteins, which are crucial for immunotherapy applications such as targeting viruses and tumors.
Innovation Solution
A machine learning model is trained to predict and generate modifications to peptide sequences using a mutation policy, embedding peptide and protein states as vectors, and employing reinforcement learning to increase the presentation score, thereby producing new peptides with enhanced binding affinity to MHC proteins.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If computational tools are used to predict peptide-MHC binding interactions, then binding affinity prediction is achieved, but the ability to generate new peptides with specified binding properties is lacking
Solution Approach 1:
A mutation policy model serves as an intermediary between existing peptide-MHC binding prediction tools and the generation of new peptides with specified properties. The model learns optimal mutation strategies from training data and applies them to generate novel peptides, bridging the gap between prediction capabilities and generative capabilities without requiring complete redesign of computational infrastructure
Solution Approach 2:
The mutation policy model is trained in advance on a dataset of known peptide-MHC bindings to learn effective mutation patterns. This preliminary training phase enables the model to subsequently generate new peptides with desired binding properties by applying learned mutation strategies, rather than requiring complex real-time optimization during peptide generation
Data Source
AI summary
Methods and systems for training a machine learning model include embedding a state, including a peptide sequence and a protein, as a vector. An action, including a modification to an amino acid in the peptide sequence, is predicted using a presentation score of the peptide sequence by the protein as a reward. A mutation policy model is trained, using the state and the reward, to generate modifications that increase the presentation score.