Incorporating Polymer Representation in DeepChem for Drug Delivery and Material Science Support

Current methods in the DeepChem library struggle to understand complex polymers, limiting their application in drug delivery and materials science. This project proposes a solution: two new representation methods for these molecules. One utilizes a detailed graph structure, capturing the intricate connections within the polymer, while the other leverages a special text format (BigSMILES) for efficient storage and retrieval. By implementing these methods, DeepChem will be significantly better equipped to handle these repetitive and complex molecules. To achieve this, the project will deliver a comprehensive toolkit: new graph-based and BigSMILES featurizers for polymer data, code for data loaders and integration with DeepChem's MoleculeNet framework, trained and evaluated GCN and MAT models utilizing the new representations, and finally, tutorials, documentation, and a public dataset of polymers represented using the new methods – all to empower the scientific community for further research in this field. The proposer has experience in relevant areas like machine learning and working with polymers.