A new tool for the automatic generation of kinetic models called Genesys is presented. A rule-based network generation methodology creates dedicated kinetic models based on the chemical knowledge of the end-user by iterating over all user-defined reaction families. The latter are constrained as much as possible to avoid the creation of unimportant species and insignificant reactions. Through integration with existing open-source chemo-informatics libraries the methodology is not constrained to specific chemical elements or to specific chemistries. The separation of all chemical data from the network generation code is accomplished by adopting a sub-molecular pattern language called SMARTS, and enables the use of quantitative structure property relationships such as group contribution methods. To demonstrate this, a Benson group additivity method for the estimation of ideal gas phase thermodynamic properties is described and compared to a benchmark database. A kinetic group additivity scheme for the estimation of Arrhenius parameters is implemented and validated. (C) 2012 Elsevier B.V. All rights reserved.