Code_and_Results_for_LLPS_Clust

LLPS-Clust is a clustering algorithm motivated by liquid–liquid phase separation (LLPS). It builds a k-nearest-neighbour affinity graph, applies Gaussian edge weights, and then emulates a physical phase-separation process: nucleation → surface-tension-driven growth → cluster fusion (union-find) → Ostwald ripening → late nucleation. The algorithm is density- and shape-robust, needs no prescribed cluster count, and marks low-density or ambiguous points as an "unassigned / dilute phase" (-1) rather than forcing them into a cluster.