Achieving accurate prediction of free energy using molecular dynamics simulations has been a central topic in computational biophysics in the past few decades. However, it is very challenging due to slow environment responses in biological systems. We introduce the orthogonal space sampling method to actively sample the environment responses along with the change of a predefined event order parameter. To demonstrate the effectiveness and predictive power of this method, we present results from various studies: alchemical free energy simulations, trans-membrane permeation of small molecules, host-guest binding, and protein folding and conformational change.