Generative Modeling of RNA Sequence Families with Restricted Boltzmann Machines

Area of Science:

• Computational Biology
• Machine Learning
• Bioinformatics

Background:

• Structured RNA molecules, including riboswitches, possess complex sequence dependencies influenced by secondary and tertiary structures.
• Understanding and modeling these dependencies is crucial for predicting RNA function and designing novel RNA molecules.
• Restricted Boltzmann Machines (RBMs) offer a potential framework for capturing these intricate sequence patterns.

Purpose of the Study:

• To explore the application of Restricted Boltzmann Machines (RBMs) for modeling structured RNA sequence families.
• To demonstrate the utility of RBMs in capturing sequence dependencies related to RNA structure and flexibility.
• To provide practical guidance and code examples for implementing RBMs in biological sequence modeling.

Main Methods:

• Mathematical and practical introduction to Restricted Boltzmann Machines (RBMs).
• Development of self-contained code examples for data acquisition, RBM training, and sequence sampling.
• Detailed implementation of algorithms for RBM application in biological sequence modeling.

Main Results:

• RBMs effectively model intricate sequence dependencies in structured RNA.
• RBMs can be successfully applied to the design of allosteric RNA, exemplified by riboswitches.
• Experimental validation confirms RBMs as generative models for specific RNA sequence families (e.g., SAM-I riboswitch aptamer domain).

Conclusions:

• Restricted Boltzmann Machines are a powerful tool for modeling RNA sequence families with complex structural features.
• RBMs facilitate the design of functional RNA molecules, including therapeutically relevant riboswitches.
• The provided computational tools enable researchers to apply RBMs for advanced biological sequence modeling and RNA design.