Differentiable Folding for Nearest Neighbor Model Optimization

2025-03-12

AI Summary

This work addresses the challenge of efficiently fitting approximately 13,000 parameters in the RNA nearest-neighbor thermodynamic model to experimental and structural data. We introduce the first differentiable RNA folding framework, built upon classic algorithms—including Nussinov and maximum expected accuracy (MEA)—to enable automatic differentiation and gradient backpropagation. The framework supports end-to-end, gradient-driven parameter optimization, joint training on heterogeneous data sources (e.g., RNAometer stability measurements and structural annotations), flexible loss design, and integration with deep learning pipelines. The learned parameter set achieves over 23 orders-of-magnitude higher predicted probabilities on real RNA family sequence–structure pairs compared to Turner2004, significantly outperforming existing baselines. This establishes a new paradigm for RNA secondary structure prediction: thermodynamically grounded, highly accurate, interpretable, and scalable.

Technology Category

• Natural Language Processing: Learning & Optimization for NLP
• Machine Learning: Structured Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

The Nearest Neighbor model is the $ extit{de facto}$ thermodynamic model of RNA secondary structure formation and is a cornerstone of RNA structure prediction and sequence design. The current functional form (Turner 2004) contains $approx13,000$ underlying thermodynamic parameters, and fitting these to both experimental and structural data is computationally challenging. Here, we leverage recent advances in $ extit{differentiable folding}$, a method for directly computing gradients of the RNA folding algorithms, to devise an efficient, scalable, and flexible means of parameter optimization that uses known RNA structures and thermodynamic experiments. Our method yields a significantly improved parameter set that outperforms existing baselines on all metrics, including an increase in the average predicted probability of ground-truth sequence-structure pairs for a single RNA family by over 23 orders of magnitude. Our framework provides a path towards drastically improved RNA models, enabling the flexible incorporation of new experimental data, definition of novel loss terms, large training sets, and even treatment as a module in larger deep learning pipelines. We make available a new database, RNAometer, with experimentally-determined stabilities for small RNA model systems.

Problem

• Optimizing thermodynamic parameters for RNA secondary structure prediction.
• Improving RNA model accuracy using differentiable folding techniques.
• Enhancing RNA sequence-structure prediction with new experimental data.

Innovation

• Differentiable folding for RNA parameter optimization
• Scalable method using RNA structures and experiments
• Improved RNA models with new experimental data