Welcome to the Github page for the implementation of KInGAN!
Kinase CycleGAN Development of a model that is capable of producing specific molecules with desired characteristics is highly dependent on three factors, namely:
- Convenient and informative representation of a molecule in a space that is usable by machine learning models
- Machine learning models that can approximate a learned distribution and generate novel samples from this distribution, ideally picking up on the most important characteristics that define the distribution
- Specific and diverse selectivity scores that assist the model in differentiating desired molecules from undesirable ones.
Recent advancements in generative machine learning research has strengthened our ability to create these three factors, allowing for the possibility of a generative model that can create kinase inhibiting molecules. In order to achieve this goal, we propose the following model for the generation of kinase inhibiting molecules by altering the structure of existing molecules to give them the kinase inhibition characteristic. The success of this experiment is highly dependent on the three factors above.
Here are some figures from the data exploration
