The theoretical predictions of endpoints related to nanomaterials are attractive and more efficient alternatives
for their experimental determinations. Such type of calculations for the "usual" substances (i.e. non nanomaterials) can be
carried out with molecular graphs. However, in the case of nanomaterials, descriptors traditionally used for the
quantitative structure - property/activity relationships (QSPRs/QSARs) do not provide reliable results since the molecular
structure of nanomaterials, as a rule, cannot be expressed by the molecular graph. Innovative principles of computational
prediction of endpoints related to nanomaterials extracted from available eclectic data (technological attributes, conditions
of the synthesis, etc.) are suggested, applied to two different sets of data, and discussed in this work.