Artificial Intelligence has a broad array of applications in nucleic acid drug development, including nucleic acid structure prediction and drug design. Scientists can utilize AI-assisted platforms to aid in designing and optimizing nucleic acid drug sequences, as well as predicting the three-dimensional structures of nucleic acid molecules. This encompasses non-coding RNAs, which perform various cellular functions, and mRNAs that encode proteins.

Fig.1 Schematic representation of artificial intelligence (AI) in the drug discovery process. (Rehman A U, et al.,2024)

Advantages of AI Technology

• AI technology has significant potential for application in areas such as molecular design, virtual screening, and compound optimization.
• The integration of AI technology and automation facilitates the efficient synthesis and continuous manufacturing of drug molecules, significantly reducing the need for human resources in candidate discovery.
• AI can effectively utilize big data to analyze personalized medical models and create individualized treatment plans.

Foundations of Artificial Intelligence in Nucleic Acid Drug Discovery

Artificial Intelligence Algorithms

In the current field of nucleic acid drug discovery and development, the application of artificial intelligence (AI) algorithms exhibits diverse characteristics, including machine learning (ML) and deep learning (DL). AI algorithms are extensively utilized in various stages, from drug target identification to molecular design, and from clinical trial data analysis to decision support for personalized treatment. These technologies have significantly enhanced the efficiency and effectiveness of nucleic acid drug development through their powerful data processing and pattern recognition capabilities.

Nucleic Acid Database

Nucleic acid databases encompass specialized databases for nucleic acid structures, databases for macromolecules that include nucleic acid structures, and databases for protein-nucleic acid complex structures.

Nucleic Acid Characterization Methods

In the context of contemporary artificial intelligence algorithms, which primarily utilize vector processing techniques, the characterization of nucleic acids relies on their structural and property features. Nucleic acid data is converted into vectors while preserving the original information to the greatest extent possible. The characterization of nucleic acid data can be categorized into features based on sequence information, physicochemical properties, and secondary and tertiary structures.

Artificial Intelligence in Nucleic Acid Drug Development

Nucleic Acid Structure Prediction

AI algorithms can predict the secondary and tertiary structures of nucleic acids from extensive nucleic acid sequence data, a task that is challenging to achieve using traditional experimental methods. By employing AI models such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) in deep learning (DL), scientists can predict nucleic acid structures with greater accuracy and speed, offering valuable insights for drug design.

Nucleic Acid Targeting Prediction

The AI algorithm effectively identifies and validates optimal RNA or DNA targets by analyzing extensive genomic, structural, and bioinformatic data. This process not only enhances the accuracy and validity of the targets but also significantly shortens the research and development (R&D) cycle. By utilizing machine learning models and deep learning algorithms, AI can predict the binding properties of targets and drug candidates, optimizing drug design to maximize therapeutic efficacy while minimizing side effects.

Nucleic Acid Drug Design

In the nucleic acid drug design phase, AI technology can aid scientists in achieving more precise molecular design and optimization. Examples include generating novel nucleic acid sequences using Generative Adversarial Networks (GANs), enhancing the chemical properties and biological activities of drugs through Reinforcement Learning (RL), and expediting the new drug discovery process via Transfer Learning (TL).

Publication Data