With the ability to perform all-atoms Molecular Dynamics (MD) simulations of complex biological systems on the micro- and even millisecond time-scales, the need to extract the interesting features of the molecular behavior inherent in the resulting trajectories has become more pressing. For the large molecules, short simulations on the order of nanoseconds are often considered to be metastable and quasi-harmonic, representing small fluctuations around a single local minimum in the free energy landscape; longer simulations, on the order of microseconds, must be treated as non-equilibrium trajectories, as they often include large, anharmonic transitions between more than one minima that can lead to significant changes in the protein topology. Because of their complexity, these long simulations are generally subjected to extensive, detailed analysis of many parameters (distances, angles, etc.), often causing the interesting dynamics to be lost in a sea of minutiae. We present a new analysis method that can be used to simultaneously identify, visualize, and compare complex events in MD trajectories of proteins. The statistical approach uses sliding window principal component analysis (sw-PCA) to identify collective motions that are large but transient, which is followed by projection techniques to compare motions between trajectories. We illustrate the method by analyzing microsecond MD simulations of the bacterial leucine transporter LeuT in complex with the substrates leucine, valine, and alanine that have been shown to produce different transport phenotypes. In all three systems we identified transient, hundred nanosecond time-scale collective motions in the intracellular domains, and found that these motions were coupled to different, substrate-specific, conformational changes in the primary substrate site. Our results indicate that the method can be a powerful tool in the analysis of all-atoms MD simulations as both system size and trajectory length increase.