Machine learning, part of the AI movement, is one of the biggest developments in computing over the last five years and Desertsci recognises its unparalleled potential in the field of data science.  DesertSci is applying ML to the field of scoring protein-ligand binding affinity.

Our experience with Scorpion, an empirical scoring function developed in collaboration with experts in the field, provides us with unparalleled know-how to build this new ranking method. In addition, our experience with Proasis, our protein-structure database system, provides us with extensive knowledge in working with large quantities of 3D protein-ligand coordinate data.   The development of this new AI technology has the potential to unlock a better understanding of the key factors driving tight ligand binding.   This, in turn, will allow for better ligands to be suggested for chemical synthesis, leading to shorter time-lines to clinical candidates.

We are focused on creating a ranking scheme based purely on protein structure data rather than relying on experimental affinity data, where the latter is severely limited by the availability of high-quality data. Furthermore, our strategy focuses on creating a ranking scheme based on non-covalent interactions, network descriptors, and protein flexibility.

Our ML technologies are a work in progress but early results are extremely promising.  We are developing the technology at a fast pace and look forward to testing our methods extensively amongst the Desertsci user community.

Watch this space …