Sensitivity Analysis of QSAR Models for Assessing Novel Military Compounds

Reliable estimates of physical and biochemical properties of novel energetic compounds are essential before making the investment to synthesize, scale-up, and manufacture a new material for use in either military or civilian applications. Quantitative Structure-Activity Relationship QSAR software tools are available for predicting the physicochemical properties and environmental impacts of these emerging materials. The uncertainty and variability in melting point, solubility, half-lives, and related properties as a means of determining whether QSAR tools could provide meaningful results were evaluated. In particular, the octanol-water partition coefficient Kow or log P was estimated for several proposed compounds. Log P was selected both because it typically can be measured with a high degree of certainty and because it correlates highly with water solubility and bioaccumulation. This study tested 1 the variability in QSAR model predictions resulting from potential structural variants in emerging chemicals and 2 the uncertainty from six different commercial Kow calculators KOWWIN, MarvinSketch, ACDLabs, CLogP, SPARC, and ALOGPs. Analyses were performed on three military compounds hexahydro-1,3,5-trinitro-1,3,5-triazine RDX, butanetriol trinitrate BTTN, and 2,4,6-trinitrotoluene TNT and two pesticides 1- chloro-3-ethylamino-5-isopropylamino-2,4,6-triazine atrazine and dichlorodiphenyldichloroethylene DDE. Analyses of these compounds revealed that the uncertainty due to structural variations can be several orders of magnitude. Variability among the five software packages was as high as 10 orders of magnitude for emerging materials although lower for more well-studied chemicals such as DDE and atrazine. The magnitude of the uncertainty suggests use of existing QSAR models for emerging energetic materials is not appropriate.