We found that along one of the axes in the 2D projection, the trajectories met at a few points, which on visual examination showed that helix 1 and helix 3 were completely formed for those data points in all trajectories. However, along the second dimension, the trajectories were still slightly separated except for around the native state where they met again. Notice that in our reduced space separation implies real distinction of conformations. Needless to say, coincidence does not necessarily imply agreement of conformations. Trajectories and 3 had more similarities with each other than with Trajectory 1 in the second projected axis. The most notable common feature found across trajectories using nMDS on all input spaces was the Garenoxacin Mesylate hydrate competition between local and global structure formation. If the protein formed all three helices very early like in trajectories 2 and 3, it spent a long time exploring non-native two-helix or collapsed conformations before dissociating and locking into the correct global structure. However, small changes in folding time such as this are not significant for small proteins. To understand the competition between global arrangement and local structure formation, it is important to study folding trajectories of larger proteins. To this end, we need at least four orders of magnitude faster computational speed. In conclusion, we have shown that nMDS can achieve high compression of MD data while preserving the salient features of the underlying trajectories. We also showed that PCA is a good tool that can be applied to the data as a first step to check for any structure present in the projected data. Our analysis has convincingly been able to pick out similarities and distinguishing features of different MD folding trajectories better than any cut-off dependent clustering method. While conventional clustering methods produced unstable clusters, nMDS produced a stable representation of all trajectories showing densely populated regions of the phase space clearly. Also, unlike tight clustering which produces a large number of clusters whose Clevudine interrelationships are not known, nMDS gives a clear picture of the relationships between data points across time. Investigations of villin headpiece folding using nMDS have shown that the three villin trajectories analyzed here explore significantly different portions of the conformational space barring a few similarities such as rapid formation of most secondary structure elements and a similar flipping transition towards the end of the folding process.