Abstract. Molecular shapes play an important role in molecular interactions, e.g., between a protein and a ligand. The `outer' shape of a molecule can be approximated by its solvent excluded surface (SES). In this article we present a new approach to molecular surface alignment which is capable of identifying partial similarities. The approach utilizes an iterative point matching scheme which is applied to the points representing the SES. Our algorithm belongs to the multi-start methods. We first generate a number of initial alignments that are locally optimized by an iterative surface point matching algorithm which tries to maximize the number of matched points while minimizing the distance between the matched points. The algorithm identifies similar surface regions given by the matched surface points. This makes it well suited for multiple alignment of molecular surfaces. The subalgorithm proposed for distributing points uniformly across a surface might be of general interest fo...