Building energy flexibility is crucial for improving the local consumption of renewable energy and enhancing building self-sufficiency. The abundant solar energy resource in the tropics presents a great opportunity to reduce carbon emission and achieve net-zero, but the building energy flexibility remains understudied in the region. Hence, this study proposed and implemented a practical control framework based on Model Predictive Control (MPC) that uncovers the energy flexibility potential of a tropical office building with hybrid cooling systems. Considering the impact of data availability on the actual control performance, MPC with alternative data usage configurations were also investigated in actual and virtual end-to-end experiments. It was first demonstrated that the proposed framework effectively regulated the building load. Compared with the baseline control, the PV self-consumption and the building self-sufficiency were respectively improved by 19.5% and 10.6%. Among the three data categories tested (internal disturbance, external disturbance, and system condition), accurate local weather conditions were shown to be the most critical for desirable control results. Moreover, the benefit of higher data granularity under different building characteristics was quantified in the simulation. Based on the systematic experiments, the relationships between the data availability and control performance were established. Accordingly, a data-centric framework was proposed to enhance the reproducibility and scalability of optimal control studies. Future research can be guided to facilitate large-scale real-world implementations.