Shape analyses aim at inferring semantic invariants related to the data-structures that programs manipulate. To achieve that, they typically abstract the set of reachable states. By contrast, abstractions for transformation relations between input states and output states not only provide a finer description of program executions but also enable the composition of the effect of program fragments so as to make the analysis modular. However, few logics can efficiently capture such transformation relations. In this paper, we propose to use connectors inspired by separation logic to describe memory state transformations and to represent procedure summaries. Based on this abstraction, we design a top-down interprocedural analysis using shape transformation relations as procedure summaries. Finally, we report on implementation and evaluation.