the human cavitome
Off-target and side-effect prediction with Innophore’s point-cloud technology
What is CavitOmiX?
When we discuss cavitomix as a field of study, it covers the integrated art and science of deciphering the binding and interaction locations within biomolecules for a specific organism.
The CavitOmiX copilot is an innovative tool specifically designed to predict side effects of drugs, while addressing the challenges in this field with an unique approach.
Traditionally, it was believed that drugs bind to similar proteins, and thus, the landscape of similar proteins was screened to predict potential side effects. However, this approach often failed to predict unexpected side effects. The reason for this is that drugs actually bind to similar binding sites, not similar proteins. This means that even proteins with completely different sequences and structures can bind the same molecule leading to unexpected off-target effects that cannot be predicted by merely screening the similar landscape of the target proteins. Instead of focusing on similar proteins, we focus on similar binding sites across different proteins. By doing so, we can predict unexpected side effects caused by drugs binding to proteins with entirely different sequences and structures.Over 42,000 distinct human proteins and their binding sites
The base of CavitOmiX is a meticulously constructed and curated dataset containing the proteins found in the human proteome, including splicing variants. This dataset was created by integrating state-of-the-art modeling tools such as AlphaFold 2TM, OpenFold, and ESMFoldTM, which are part of the BioNeMo platform from NVIDIATM. To ensure high quality and comparability of the dataset, we also employed homology modeling using Innophore’s CavitOmiX platform.
Off-target effects and drug repurposing
We analyze binding sites through point clouds, enabling us to precisely map various physico-chemical properties such as electrostatics, hydrophobicity, accessibility, hydrogen-binding potential, elasticity, and many more. Our point clouds are an excellent resource for machine learning and AI applications. Our AI matching algorithm allows us to compare each point cloud in the human proteome with all the others, creating an interaction map. This map enables us to identify potential new drug binding sites, which may indicate off-target effects or suggest drug repurposing opportunities.