Increasing evidence has indicated that microRNAs (miRNAs) can functionally interact with environmental factors (EFs) to affect and determine human diseases. Uncovering the potential associations between diseases and miRNA–EF interactions could benefit the understanding of the underlying disease mechanism at miRNA and EF levels, miRNA signatures identification, and drug repurposing. In this study, based on the assumption that similar miRNAs (EFs) tend to interact with similar EFs (miRNAs) in the context of a given disease and under the framework of random walk with restart (RWR), a novel method of miREFRWR was developed to uncover the hidden disease-related miRNA–EF interactions by implementing random walks on an miRNA similarity network and EF similarity network, respectively. miREFRWR was evaluated by leave-one-out cross-validation, which achieved an AUC of 0.9500. It has been demonstrated that miREFRWR can effectively identify potential interactions in all the test classes, even if these test samples only share either EFs or miRNAs with the training samples. Furthermore, many predictive results for acute promyelocytic leukemia and breast cancer (67 and 10 interactions out of the top 1% predictions, respectively) have been verified by independent experimental studies. It is anticipated that miREFRWR could be a useful and important biological resource for biomedical research.