A new molecular structure representation: Spectral weighted molecular (SWM) signals for studies of molecular similarity

A new molecular representation is presented based on a semi-invariant decom position of the 3D molecular structure. The basic approach is Principal Com ponent Analysis on the (x,y,z) atomic coordinates of a molecule, obtaining the atom projections on the three principal axes (the scores). The directio n of each principal axis is uniquely defined, but not the versus. Whereas WHIM descriptors [1] are statistical indices calculated on the scor es, SWM signals are directly obtained by weighting the scores of each axis by the weights defined in the WHIM descriptors framework (mass, polarizabil ity, Mulliken atom charge, van der Waals volumes, electrotopological charge s). Thus a molecule can be represented by a sequential juxtaposition of the weighted scores of three principal axes, obtaining a spectral representati on: the scores are the atom positions along the asis and the weights are th e signal intensities, Similarity analyses based on this representation have been performed on different sets of compounds using the Camberra distance.