The view of materials as systems has paved the path for a hierarchical approach to materials-by- design. The emergence of powerful cyberinfrastructure have created unprecedented opportunities to develop data-driven approaches to material design and discovery. In the context of predicting microstructure-sensitive mechanical behaviors of materials, a core challenge is to perform high-fidelity computations with low computational cost. By way of consequence, high throughput approaches are emerging as a means to project mechanical behaviors from finer scale calculations to a coarser scale, and augmenting them with data-driven (e.g., machine-learning (ML)) framework. We present a strategy to decode the role structure (grain size and texture) and damage tolerance (property) in hexagonal close-packed materials with application to magnesium alloys. Such a strategy provides a pathway toward efficient data-driven damage prediction.