A SMILES string encodes a molecular graph as a sequence of atom and bond symbols, with parentheses for branches, matching digits for ring closures and additional symbols for charges and stereochemistry. For example, specifies ethanol. A molecule may have several valid SMILES traversals, so the raw character sequence is not itself a unique physical descriptor.
CCOTwo numerical encodings are:
- A molecular fingerprint: a fixed-length vector indicating local substructures, for example hashed circular neighborhoods around atoms. Each coordinate is a bit or a count. The same atom–bond graph should produce the same fingerprint regardless of its SMILES traversal.
- A molecular descriptor vector: real-valued quantities such as molar mass, atom counts, formal charge, hydrogen bond donor and hydrogen bond acceptor counts, or topological indices. Standardize these coordinates using the training set before fitting a scale-sensitive model.
Articles by others on the same topic
There are currently no matching articles.