Developing Graph based Chemical Representation for Synthetic Lipid and Evaluating its Application for AI-based Predication for siRNA delivery
sun, zhuorui.
2021
-
Thesis (M.S.)--Tufts
University, 2021.
Submitted to the Dept. of Biomedical Engineering.
Advisor: Qiaobing Xu.
Committee: Mark Cronin-Golomb, and Soha Hassoun.
Keyword: Biomedical engineering.
Therapeutic nucleic acid-based macromolecules play a critical role in preventing and treating different diseases. One of the most significant challenges is ... read moredeveloping a safe and efficient delivery system. Lipid nanoparticles (LNP) based delivery system is a versatile and promising tool for biomacromolecules delivery. Combinatorial library approach has been used to synthesize new lipids with diverse chemical structures. Together with high throughput library screening, lipids with good performance (e.g. high delivery efficiency and low toxicity) were identified from thousands of candidates. However, large amount of work is required for chemical syntheses and library screening. With the development of machine learning, a deep learning model can learn from the experimental data, which may help predict the property of new input after training. Herein, I developed an Artificial Intelligence (AI) based method to predict the siRNA delivery efficiency of LNPs based on the experimental data from literature. I represented lipids into Graph data structure with node feature as input data and implemented a trained Graph Neural Networks model to predict the lipid delivery efficiency. This pilot study provides new insights into screening lipid with high performance based on the experimental data, and can be applied to predict the property of a new LNP even just based on its chemical structure. With the AI prediction model, I can test and predicate the LNPs even without actually synthesizing the lipid. In addition, this study provided a lipid representation method and a database of synthetic lipids which can be used in other advanced models in the future.read less - ID:
- 02871951f
- To Cite:
- TARC Citation Guide EndNote
- Usage:
- Detailed Rights
