A New Tool for Research in QSPR/QSAR
The software aimed to construct QSPR/QSAR models by means of optimal descriptors calculated using SMILES notations. In fact the software produce such values of correlation weights of SMILES fragment which produce maximal values of special criterions. These can be first correlation coefficient between endpoint and optimal descriptor (in the case of training – test scheme); and second criterion (in the case of training-calibration-test scheme) calculated as
where R(training) and R(calibration) are correlation coefficients for endpoint under consideration and descriptor calculated as
for the training and calibration set. CW(SF) is a correlation weight of SMILES fragment.
This software can be used for any datasets which are represented as SMILES notations. In other words, these can be organic, inorganic, coordination compounds. In particular the approach has been tested for modeling fullerene C60 solubility in organic solvents. The elucidation of molecular structure has been defined as SMILES notations of these solvents.
RR=R (training) + R (calibration) – abs[ R (training)-R (calibration)]*alpha
2
2
2
2
DCW =......CW(SF)
S
Overview
