Secure and Focused Hit-Expansion with CHEESE Search
Privacy mode, similarity thresholding, and multi-database searches — new features for precise, confidential molecular discovery across 39B+ molecules.
CHEESE Search: New Features for Hit Expansion
CHEESE is the super-fast molecular similarity search in enumerated databases, offering access to over 39 billion molecules searchable via multiple similarity metrics including 3D shape and 3D electrostatics.
Privacy Mode
CHEESE Search now provides privacy protection, allowing researchers to run molecular searches without leaving any digital trace on the public version of the service. Privacy Mode complements on-premises deployment options for organizations requiring internal hosting.
Similarity Thresholding
The platform enables filtering results by similarity scores, moving beyond broad searches toward precise cutoff specifications. Users can establish shape or electrostatic thresholds and download scored results. This supports identifying the most shape-similar but least electrostatic-similar molecules — or vice versa — for tasks like scaffold hopping.
How It Works
CHEESE uses AI-generated isometric vectors where Euclidean distance reflects molecular similarity. The similarity score is computed as 1 minus the normalized Euclidean distance, on a scale from 0 (minimum similarity) to 1 (maximum similarity), leveraging Asymmetric Distance Computation over product quantisation codes.
- Display and download similarity scores or filter by them
- Multi-database searches showing hit distribution across sources
- Enhanced reproducibility and consistency
- Result decluttering by removing dissimilar molecules
- Comparative shape/electrostatic analysis experiments