Hello science enthusiasts! Today we have an exciting topic to dive into: AI-driven citizen science and recombinant adeno-associated virus (RAAV) drug formulation development with the assistance of ChatGPT. Sorry for the long title, but if you are watching this, I am going to assume that you have a high-level understanding of ChatGPT.
Before we get started, let’s first understand at a basic level what recombinant adeno-associated virus drug products are all about. Recombinant adeno-associated virus (RAAV) is a type of virus used in gene therapy, a cutting-edge field that holds great promise for treating genetic diseases. Our RAAV drug products are customized formulations that carry therapeutic genes to target cells in the body, aiming to correct or replace faulty genes responsible for diseases.
Developing these drug formulations is a complex and time-consuming task requiring years of expertise. Traditional methods involve running numerous experiments to test different formulation conditions, making it an arduous process. However, with the help of ChatGPT, an AI language model, I embarked on a proof-of-concept project to demonstrate the power of AI-driven citizen science and RAAV drug formulation development.
Citizen science is all about involving non-experts like you and me in scientific research. There are a variety of citizen science projects open for the public to participate in, such as Zooniverse, SciStarter, and Project BudBurst. With AI-driven solutions, citizen scientists can actively participate in drug development and biopharmaceutical research with limited domain knowledge.
Now, let’s talk about the methodology of this project. I generated a synthetic dataset in Python using the pandas, numpy, and random libraries. This dataset included numerical data types for vector concentration, cryoprotectant concentration, lipoprotectant concentration, buffer pH, buffer concentration, and preservative concentration. It also included categorical data types for cryoprotectant type, lipoprotectant type, buffer type, bulking agent type, preservative type, and lethality.
To build and train the neural network, I used an artificial neural network, which is a type of machine learning model inspired by the human brain. It mimics the way biological neurons signal to one another. The neural network takes in the dataset and predicts the lethality of different RAAV drug formulations. The numerical variables and categorical variables are used as features for the neural network.
To optimize the neural network, I performed hyperparameter optimization, which involved trying different layer and feature combinations. ChatGPT helped me change the code so that I could automatically try different combinations. However, the accuracy score of the model was only around 50 percent with the synthetic data. Further improvements and validation with real-world experimental data are crucial.
In conclusion, this project showcases the potential of AI-driven citizen science and RAAV drug formulation development. With the assistance of ChatGPT, even individuals with limited coding experience can contribute to scientific research. However, it is important to note that this project is a proof of concept, and further research and validation are needed.
Thank you for watching, and I hope this video was helpful for those interested in citizen science, coding projects, or the gene therapy space. If you are interested, please like, comment, and subscribe for more content about citizen science projects. Have a great day!
