Lipid Nanoparticle Delivery

Lipid nanoparticles are promising delivery mechanisms for a variety of drugs into the cytoplasm. They have been employed to deliver a wide range of nucleic acids, most notably the mRNA in Pfizer’s and Moderna’s COVID-19 vaccines. The main component and most important factor in improving delivery efficiency are the ionizable lipids, for which there is a narrow interval of acceptable pKas, usually between 6.1 and 6.7. Lipids must be synthesized to be screened, a process that can take months and many expensive reagents. Our project would allow for more efficient screening of ionizable lipids by creating a model to predict their pKa and delivery efficiency. Hence, researchers would only need to synthesize the lipids that the model considers a hit, greatly reducing the time and resources spent on low-quality lipids. We tried approaching the problem by creating a database from patents and publications. We then tried different structure encoding and supervised learning approaches, including kNN, random forest, and XGBoost. This resulted in a reasonably good model, considering factors like limited dataset and noisy data. We also hope to make the final product accessible for researchers without a coding background; as such, we have a user-friendly website where individuals can enter in SMILE sequences of ionizable lipids and a pKa value will be returned. At the moment, the website predictor is not fully connected to our model but we hope to improve upon it in the future.

Built With

• css
• html
• javascript
• python
• scikit-learn