To investigate the description of molecular shape, a heterogeneous data set
of 200 compounds was designed. The compounds included many disparate molec
ular shapes, and a total of 64 descriptors was calculated including topolog
ical indices, surface areas, molecular volumes, measures of molecular dimen
sions, and ratios of moments of inertia. Cluster analysis on the variables
indicated that there are four major categories of steric descriptors, those
for molecular bulk, dimensions, and two clusters representing varying aspe
cts of shape. To describe molecular shape more efficiently, a novel lineari
ty index based upon ratios of moments of inertia and the molecular weight i
s proposed.