COPD Disease Classification Using Network Embedding with Synthetic Relationships

Chronic obstructive pulmonary disease (COPD), a progressive and non-reversible lung disease causing obstructed air-flow from the lungs, often occurs with other diseases not restricted to the respiratory system. Therefore, it is important to understand interaction between genes and diseases to uncover the real causes of a disease. In this paper, we propose to automatically classify COPD diseases, using network of gene disease relationships. We simplify interaction between COPD, COPD multimorbidities, and related genes as a bi-partite network, and apply network embedding together with machine learning classifiers to classify diseases into different categories. Our experiments confirm that adding synthetic edges in a strategic way statistically enhances quality of node embedding and improve COPD disease classification performance.