The present work focuses on analyses of the autoignition delay time predicted by the large eddy simulation (LES) method by applying different subgrid scales (SGS) models and two different discretization schemes. The analysed flow configuration is a two-phase chemically reacting turbulent flow with monodispersed evaporating fuel droplets. The impact of numerical procedure is investigated in a 3D flow domain with a temporally evolving mixing layer that constituted between the streams of fuel and oxidizer that moved in opposite directions. The upper stream of cold gas carries a dispersed fuel spray (ethanol at 300 K). The lower stream is a hot air at 1000 K. Three commonly used in LES, SGS models are investigated, namely: classical Smagorinsky model, model proposed by Vreman and the #27;-model proposed by Nicoud. Additionally, the impact of two discretization schemes, i.e., total variation diminishing (TVD) and weighted essentially nonoscillatory (WENO) is analysed. The analysis shows that SGS model and discretization scheme can play a crucial role in the predictions of the autoignition time. It is observed that for TVD scheme the impact of SGS model is rather small. On the contrary, when the WENO scheme is applied the results are much more dependent on the SGS model.