DesertSci is developing new AI models based on DeepLearning for better predicting properties such as ligand affinity and binding site flexibility.We are in the position of having both a very large resource of highly curated experimental data and a comprehensive tool box of innovative methods, perfectly suited for AI model building.DesertSci is combining:
- protein-ligand non-bonding interaction data from ViewContacts
- network concepts from Scorpion
- data from clustering of binding site atoms from Proasis4
to generate our next set of predictive tools for structure based drug design.We are also using DeepLearning to train modelsfor selecting hits from our fragment based ligand design software, Viper.
