Targeted Molecular Generation driven by Fingerprint Embeddings

Identifier:  TFG:8162

Handle: https://hdl.handle.net/20.500.11797/TFG8162

Authors:  Reverté Àlvaro, Berta

Abstract:
This project explores the adaptation of a diffusion model for directed molecular generation by incorporating Mol2Vec embeddings. We adjusted the original model to include these embeddings, which provide detailed molecular context, in the generation process. Using graphical neural networks (GNNs), the model was trained to generate molecules with the desired properties. Our analysis shows potential, although further optimisation is needed. This approach has applications in pharmaceuticals, metabolomics, agriculture, and materials science, enabling more accurate and efficient molecular design.