It is well known that ordinary differential equations (ODEs) are widely used to model complex dynamic systems in many areas of science and technology. For example, ODE models have been used in the study of HIV viral dynamics and reconstruction of transcription factor activity for gene regulation. Although ODE models can often be proposed based on expert knowledge of the dynamic process of interest, the values of the ODE parameters are rarely known. Estimating these parameters from observational (noisy) data is an important but challenging statistical problem, because most ODEs have no analytic solutions, and solving ODEs numerically is computationally intensive. In this talk, we first review some popular estimating procedures for ODEs. Then, we propose a semi-parametric mixed-effects ODEs model. A semi-parametric efficient approach is provided for estimating parameters of ODEs. The efficiency of proposed approach is illustrated by analyzing a PK/PD model and a HIV model.