| Description: |
The number of new materials and applications of these materials is experiencing rapid growth. Today, increased computational power and the established use of automated machine learning approaches make data science tools available, which provide an overview of the chemical space, support the choice of appropriate materials, and predict specific properties of materials for the desired application. Among the different data science tools, graph theory approaches, where data generated from numerous real-world applications are represented as a graph (network) of connected objects, has been widely used in a variety of scientific fields such as social sciences, health informatics, biological sciences, agricultural sciences, and economics. In this work, we describe applying a particular graph theory approach, social network analysis (SNA), to the metal-organic framework (MOF). To demonstrate MOF materials, we construct a social network called MOFSocialNet from geometrical MOFs descriptors in the CoRE-MOFs database. The MOFSocialNet is an undirected, weighted, and heterogeneous social network; following the construction of this graph, a set of social network analysis processes is conducted to extract valuable knowledge from the MOFs data using graph machine learning algorithms. Community detection is one of the well-known SNA techniques employed on the MOFSocialNet to extract the most similar MOF communities. To evaluate whether the properties of new MOFs can be predicted using MOF communities, we randomly chose three from the CoRE MOFs database. For these MOFs, we excluded the crystal density as input during featurization and placed the MOFs within the MOFSocialNet. The crystal density of the new MOFs is predicted by simply averaging the crystal density of the ten nearest neighbors. Additionally, communities extracted from MOFSocialNet can be leveraged to predict MOF gas adsorption properties for CO2 and CH4. |