This research paper presents MetaPOL, a novel approach to enhancing physical security needs of indoor facilities through digital twin construction for human Patterns of Life (PoL) simulation. Our methodology integrates an agent-based simulation component developed in AnyLogic, complemented by an interactive, immersive Virtual Reality (VR) metaverse developed in Unity. Human users are able to interact with the VR environment, which replicates a secure indoor facility via a Meta Quest Pro VR headset. In instances where access to motion sensor data from the secure indoor facility is restricted due to security and safety concerns, the agent-based simulator is able to generate human movement data based on predefined behavior rules. Next, two deep neural networks are trained as surrogate models on the movement data generated by simulations of the agent-based model. These trained networks infer waypoint stay duration and next waypoint selection for Non-Player Characters (NPCs) within the immersive VR simulator. The VR simulator is designed for training facility users in appropriately responding to emergencies and deploying safeguards against insider threat scenarios. This approach is particularly valuable for scenarios that are too hazardous or expensive to reproduce or enact in real-world physical environments for mobility sensor-based data collection. We have presented a case study of deploying extit{MetaPoL} for insider threat scenario assessment in the indoor layout of a real nuclear reactor facility. Our methodology demonstrates how integration of agent-based simulation, metaverse technology, and deep neural networks can be harnessed by digital twins and leveraged for preparedness and responsiveness within secure indoor facilities.