Structure-property linkages are at the heart of materials science and engineering. The conventional method of rigorous experimentation and validation is time and cost-intensive. In this thesis, we use concepts from materials informatics to develop quantitative structure-property correlations in materials. We employ two-phase dual phase (DP) steels as a model material system. In particular, we study the correlation between the microstructure and the damage initiation in DP steel. First, we present a simple and unique approach to estimate the intrinsic dimensionality of two-phase microstructure data using principal component analysis (PCA) and multi-dimensional scaling (MDS) techniques. Next, we develop a microstructure and work hardening sensitive Reduced Order Model (ROM) for predicting damage initiation using random forest regression. A general framework using a statistical fitting model to rank the severity of damage initiation in microstructures with various morphologies is introduced. Finally, we propose a way to address the issues of generating a large volume of statistically similar and physically aware microstructures from small data sets using generative adversarial networks. The overall framework is an approach to reduce the materials design to deployment time and contribute to accelerated materials development.