RNA Tertiary Structure Ranking for Dynamic Binding Affinity Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The prediction of RNA tertiary structures is challenging due to the scarcity of experimentally solved structures in databases and the dynamic nature of RNAs, necessitating methods that can generate a collection of possible structures and account for thermodynamic contributions.
Innovation Solution
A method involving RNA secondary structure prediction models, followed by a RNA tertiary structure generator, and thermodynamic molecular machine learning models to rank and determine structural binding affinity of small-molecules, utilizing generative artificial intelligence and molecular dynamics simulations to predict thermodynamically ranked RNA tertiary structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black-box AI models are used for structure prediction from databases, then protein structure prediction accuracy is improved, but RNA tertiary structure prediction remains challenging due to severe lack of training data
Solution Approach 1:
The patent uses RNA secondary structures as an intermediary representation to bridge the gap between sequence and tertiary structure prediction. By first predicting secondary structures (which have more available training data) and then using them as input for tertiary structure prediction, the method overcomes the severe lack of RNA tertiary structure training data. This two-stage approach with secondary structure as mediator enables accurate tertiary structure prediction despite limited direct training examples.
2Device complexity
If a single conformation is predicted, then the prediction process is simplified, but it fails to account for the inherently dynamic nature of biomolecules especially RNAs with more rotatable bonds
Solution Approach 1:
The patent implements dynamics by generating an ensemble of multiple RNA tertiary structure conformations rather than a single static structure. The system samples multiple possible tertiary structures from the predicted secondary structures, capturing the dynamic nature of RNA molecules with their numerous rotatable bonds. This ensemble approach maintains reliability by representing the full conformational space while managing complexity through efficient sampling methods.
3Reliability
If multiple RNA tertiary structures are generated to account for dynamic nature, then accuracy of structure representation is improved, but computational complexity and resource requirements increase
Solution Approach 1:
The patent segments the structure prediction process into distinct stages: secondary structure prediction followed by tertiary structure generation. By dividing the complex task of direct tertiary structure prediction from sequence into manageable segments, the system can generate multiple accurate tertiary structures while controlling computational complexity. Each segment uses specialized algorithms optimized for that specific prediction level, making the overall process more efficient despite generating ensembles.
4Measurement precision
If thermodynamic ranking is implemented to account for energetic and entropic contributions, then accuracy of structure prediction is improved, but computational requirements and model complexity increase
Solution Approach 1:
The patent replaces complex explicit thermodynamic calculations with a machine learning model that has been trained to predict thermodynamic properties. Instead of implementing detailed mechanical models of energetic and entropic contributions, the system uses a trained neural network to directly predict thermodynamic rankings of structures. This substitution maintains high accuracy in accounting for thermodynamic factors while significantly reducing model complexity and computational requirements.
Data Source
AI summary
A method may include obtaining a ribonucleic acid (RNA) sequence from a protein data bank comprising a plurality of RNA sequences. The method may also include generating a plurality of candidate RNA secondary structures by passing the RNA sequence through at least one RNA secondary structure prediction model. The method may further include assembling, via a RNA tertiary structure generator, the plurality of candidate RNA secondary structures into a plurality of RNA tertiary structures. The method may also include ranking the plurality of RNA tertiary structures by implementing a plurality of thermodynamic molecular machine learning models. The method may further include determining a structural binding affinity of small-molecule based on the ranked plurality of RNA tertiary structures.