The Message Passing Interface standard MPI offers collective communication routines to perform commonly required operations on groups of processes. The usage of these operations is highly recommended due to their simplified and compact interface and their optimized performance. This paper describes a pattern matching approach to detect clusters of point-to-point communication functions in program traces, which may resemble collective operations. The extracted information represents an important indicator for performance tuning, if point-to-point operations can be replaced by their collective counterparts. The paper describes the pattern matching and a series of measurements, which underline the feasibility of this idea.