The Connected Subgraph Fingerprint (CSFP) is a novel fingerprint method which, in contrast to other methods, captures all connected subgraphs as structural features of a compound. This property gives the CSFP a complete feature space and high adaptive potential. Apart from surpassing common methods in standard similarity-driven virtual screening settings, the CSFP has substantial structural advantages when applied to combinatorial fragment spaces or in machine learning.
The CSFP and four derivatives are available as part of the Python module CSFPy. The user can load molecules from file or via a SMILES string, generate fingerprints and derive similarity measurements. The module is free for noncommercial and acadamic research for Linux, Windows and Mac OS X. Non-acadamic users can get an evaluation license free of charge.
The CSFPy module is part of the AMD tools software bundle. To download the module, register at https://software.zbh.uni-hamburg.de.