The verification of dependability and performance requirements plays a critical role during the development of complex heterogeneous system-of-systems in the automotive, transportation, and aerospace domains. The dependability and performability aspects of the system are often analyzed either by manually constructing formal models or by model transformations provided by systems engineering toolchains. However, for system-level dependability and performance verification, various models from multiple formalisms shall be integrated into a single analysis model as heterogeneous system-of-systems development often relies on different engineering languages and toolchains for different components and subsystems. This paper presents a novel approach to integrating various engineering models into a common extra-functional analysis framework. We utilize an existing intermediate formal language to provide formal semantics to the engineering languages and also introduce a language extension to capture the stochastic aspects of the system behavior. We also introduce an efficient simulation-based analysis approach using statistical model checking and Bayesian machine learning. We developed a prototype implementation to support our analysis approach, which enables the integrated evaluation of different architecture models and languages using model transformation. We evaluated our analysis approach and prototype implementation using case studies inspired by aerospace, transportation, and IoT systems.