Artificial Intelligence (AI)-based DNA-encoded Glycan Library (DEGL) Design
Overview
AI-based DEGL design represents a cutting-edge approach that integrates advanced computational techniques with DNA encoding technology to revolutionize glycan research. This methodology leverages AI algorithms to design and optimize expansive glycan libraries, providing powerful tools for high-throughput screening and targeted discovery. By utilizing AI for data analysis, structure prediction, and virtual screening, researchers can efficiently explore vast chemical spaces and identify promising glycan candidates with enhanced biological activity. DNA encoding enables precise tracking and management of these glycans, facilitating large-scale synthesis and detailed profiling.
AI-Driven DEGLs: Precision, Speed, Discover
CD BioGlyco provides one-stop DEGL Design services with our extensive experience and expertise. Our DEGL design services include but are not limited to Chemistry-driven DEGL Design, Biology-driven DEGL Design, and AI-based DEGL design.
Experimental Data Analysis and Pattern Recognition
We leverage AI to quickly process and analyze large amounts of data obtained from experiments, extracting useful information and patterns that can be deciphered from DEGL's complex data for the benefit of DEGL design. At the same time, we use machine learning algorithms to enable AI to recognize the relationship between glycan molecules and biological targets and discover potential binding modes and interactions.
Structure Prediction and Optimization
Our experts use AI tools to predict the structure of glycan molecules and their binding modes to targets. These predictions help optimize the design of glycans for better biological activity. We also optimize the structures of glycan molecules to improve their binding affinity and stability. For example, by adjusting the functional group or spatial structure of the glycan to enhance the interaction with the target.
Screening and Evaluation
We utilize AI tools to efficiently screen potential active glycan molecules in DEGL, reducing the number of experiments and resource consumption. The most promising candidate molecules are prioritized by AI prediction combined with affinity. Moreover, we predict the biological activities and functions of glycan molecules based on existing data and models to provide targeted candidate molecules for experiments.
We offer a highly comprehensive DEGL design service by integrating these three aspects, resulting in significantly enhanced efficiency and quality of the designed DEGL.
Workflow
Applications
- AI algorithms are used to screen compounds in designed glycan libraries to discover potential drug candidates and improve the efficiency and accuracy of drug discovery in DEGL.
- AI-based DEGL design can examine the interactions between glycan molecules and biological targets, assist in the discovery of novel biomarkers for DEGL construction, and enhance the inclusiveness of glycan molecules covered by the designed DEGL.
- AI-driven design generates sugar molecules with novel structures and functions for synthetic biology applications and biomaterials development.
Advantages
- AI algorithms analyze and predict large amounts of DEGL data to efficiently identify and screen glycan molecules that bind specifically to the target, thus improving the hit rate of compounds against the target in DEGL.
- We design glycan molecules with specific functional groups according to specific biological goals and research needs, thereby increasing the relevance and validity of the library.
- AI-driven design can generate glycan libraries with a wide range of structural diversity, exploring the broader glycan chemical space.
Publication Data
Technologies: Machine learning, DNA-encoded library (DEL)
Journal: ACS omega
Published: 203
IF: 2.931
Results: This article focuses on the use of deep learning methods to discover tumor-targeting small organic ligands from DELs. The authors developed machine learning models and applied them to DEL data for hit finding and hit-to-lead processes. Their approach enabled the discovery of new hits against CAIX and the successful translation of DEL selection data into lead compounds with good in vivo biodistribution and excellent accumulation in CAIX-positive tumors. The authors also applied the model to a commercial catalog to discover novel and structurally diverse hits with high hit rates. The results show that the DEL-derived machine learning model can be successfully generalized beyond its training range and provide accurate predictions for non-DEL compounds. The authors also discuss ways to further extend the model, including building probabilistic regression models based on DEL screening results and improving model performance through hyperparameter search.
Fig.1 The DEL training dataset mentioned in the article. (Torng, et al., 2023)
Frequently Asked Questions
- How is the accuracy of AI predictions ensured?
- Comprehensive and representative training datasets should be ensured, including structural, functional, and binding affinity data of various glycan molecules. At the same time, noise and errors in the data are removed to ensure data quality and the accuracy and consistency of the training data. The data is divided into training and test sets using cross-validation techniques to ensure that the model performs well even on untested data.
- How much data is needed to train an AI model?
- The amount of data required to train an AI model depends on the complexity of the model and the difficulty of the task. In general, more high-quality data can improve the model's performance, but data cleaning and preprocessing are also needed to ensure the model's accuracy.
CD BioGlyco is dedicated to keeping abreast of the latest advances in DEGL design so that we can provide innovative AI-based DEGL design solutions that meet the unique needs of each client. If you need more information about our AI-based DEGL design solutions, please feel free to contact us!
References
- Torng, W.; et al. Deep learning approach for the discovery of tumor-targeting small organic ligands from DNA-encoded chemical libraries. ACS omega. 2023, 8(28): 25090-25100.
- Mathez, G.; et al. Novel inhibitors of SARS-CoV-2 RNA identified through virtual screening. Journal of Chemical Information and Modeling. 2024.
For research use only. Not intended for any clinical use.
Quick Links
Related Solutions
