Molecule and drug Discovery Application For Three Diseases

Main phases that i worked on for this projects:

Data Preparation

In this phase, we meticulously handled the data. The key steps involved are:

1. Data Import: We imported the ChEMBL datasets relevant to our three target diseases.
2. Data Cleaning: Extensive cleaning was performed to ensure data quality and consistency.
3. Feature Engineering: Relevant features were extracted from the datasets to facilitate modeling.

Modeling Phase

The modeling phase had two distinct objectives:

Molecule Generation

• We explored two deep learning algorithms: Variational Autoencoder (VAE) and MolGann.
• After thorough evaluation, VAE was selected as the preferred algorithm and applied to all three datasets.

Bioactivity Prediction

• We employed various machine learning algorithms to predict two key aspects:

  • PIC50 Value and Bioactivity Class: Predicting the PIC50 value for bioactivit and Classifying compounds based on their bioactivity.

• The best-performing model for each dataset was identified and saved for deployment using Streamlit.