Visualization of multi-property landscapes for compound selection and optimization

Chemical space visualization: transforming multidimensional chemical spaces into similarity-based molecular networks

BACKGROUND

The concept of chemical space is of fundamental relevance for medicinal chemistry and chemical informatics. Multidimensional chemical space representations are coordinate-based. Chemical space networks (CSNs) have been introduced as a coordinate-free representation.

RESULTS

A computational approach is presented for the transformation of multidimensional chemical space into CSNs. The design of transformation CSNs (TRANS-CSNs) is based upon a similarity function that directly reflects distance relationships in original multidimensional space.

CONCLUSION

TRANS-CSNs provide an immediate visualization of coordinate-based chemical space and do not require the use of dimensionality reduction techniques. At low network density, TRANS-CSNs are readily interpretable and make it possible to evaluate structure-activity relationship information originating from multidimensional chemical space.

Recent developments in SAR visualization

Shown is a section of an SAR network. Nodes represent compounds and are colored by potency and edges indicate pair-wise similarity relationships.