Model-based systems engineering approaches are commonly used for the development of various heterogeneous mechatronic systems including nowadays popular Unmanned Aerial Vehicles (UAV). As a rule, the dependability analysis is carried out during the system integration phase of the UAV development. Therefore, fundamental design drawbacks might be uncovered too late leading to serious and costly rework. This paper presents a new methodology for the early dependability analysis of a UAV system applicable during the design phase. The proposed dependability analysis is based on the Dual-graph Error Propagation Model (DEPM), a stochastic model of system design aspects that influence error propagation processes: control flow, data flow, and component-level reliability properties. A new domain-specific Systems Modeling Language (SysML) profile and the transformation algorithm from the SysML model to the DEPM are introduced as two key parts of this methodology as well as the main research results of this paper. The relevant technical details of the modeling workflow are demonstrated with a case study UAV that explains how to design a UAV using the introduced SysML profile, transform the SysML model into the DEPM, and evaluate dependability properties.