Carbon nanotubes (CNT) have unparalleled mechanical properties, spanning several orders of magnitude over both length and time scales. Computational and experimental results vary greatly, partly due to the multitude of variables. Coupling physics-based molecular dynamics (MD) with informatics methodologies is proposed to navigate the large problem space. The adaptive intermolecular reactive empirical bond order (AIREBO) is used to model short range, long range and torsional interactions. A powerful approach that has not been used to study CNT mechanical properties is the derivation of descriptors and quantitative structure property relationships (QSPRs). For the study of defected single-walled CNTs (SWCNT), two descriptors were identified as critical: the density of non-sp2 hybridized carbons and the density of methyl groups functionalizing the surface. It is believed that both of these descriptors can be experimentally measured, paving the way for closed-loop computational-experimental development. Informatics can facilitate discovery of hidden knowledge. Further evaluation of the critical descriptors selected for Poisson’s ratio lead to the discovery that Poisson’s ratio has strain-varying nonlinear elastic behavior. CNT effectiveness in composites is based both on intrinsic mechanical properties and interfacial load transfer. In double-walled CNTs, inter-wall bonds are surface defects that decrease the intrinsic properties but also improve load transfer. …