20–40% of lung cancer patients develop bone metastasis (BM) with significantly decreased overall survival. Currently, BM is mainly diagnosed by computerized tomography (CT) scan or magnetic resonance imaging (MRI) when symptom develops. Novel biomarkers with higher prediction value of BM are needed. Plasma-derived exosomal microRNAs had been isolated and sequenced of total 30 non-small cell lung cancer (NSCLC) patients including 16 with bone metastasis and 14 without bone metastasis. Hierarchical clustering based on the total miRNA profile can clearly separate cancer patients and healthy individuals (H), but not patients with (BM +) or without (BM−) BM. Weight Co-expression network of miRNAs (WGCNA) analyses identified three consensus clusters (A, B, C) of highly correlated miRNAs, among which cluster B (144 miRNAs) showed significantly differential expression in lung cancer patients, especially in BM + group. Pathway analysis of cluster B miRNAs revealed enrichment in metabolic pathways that may involve in preconditioning of the metastatic niche. Three differentially expressed miRNAs between BM + and BM− patients within cluster B were identified as miR-574-5p, a suppressor of Wnt/β-catenin pathway, was down-regulated, while miR-328-3p and miR-423-3p, two activators of the same pathway, were up-regulated in BM + patients. Cluster A miRNAs (n = 49) also showed trend of upregulation in BM + patients. Interestingly, pathway analysis indicated that 43 of them are associated with chromosome14, which has been suggested to promote epithelial-mesenchymal transition (EMT) and bone metastasis.