Hello all,

I am confronted to difficult analysis task (at least for me) on my trajectory file. 

I am studying coarse-grained molecules that are initially configured in 2 aggregates using droplet formers, on each sides of a rectangular box (one at x= 0.25, the other x = 0.75 along the longest dimension of this rectangular box). Everywhere else inside the box are solvent beads.

My goal is to track in time the number of beads in both aggregates, that will evolve continuously. The tricky part is that I observe also a drift of the droplet themselves, so classifiying beads as belonging to an aggregate is not easy as I can't select the initial COM of the aggregates. 
I tried using a friend-of-friend algorithm (all beads below a cutoff distance X from a another bead are considered belonging to a same cluster) but it seems to struggle a bit too.

Do you have recommandations on how to track the clustered bead numbers accurately in time ?

Kind regards,
Moncef