Background Circulating biomarkers for cancer have great potential for diagnosis as well as follow up of treatment. MicroRNAs (miRNA) are involved in the expression of a majority of proteins with different cell types having different miRNA expression. The aim of this study was to create a circulating miRNA-based model to discriminate patients with lung cancer from patients with benign lung disease. Methods Samples were collected from patients under investigation for lung cancer at Örebro University hospital. Patients were then divided into groups based on diagnosis, which resulted in NSCLC adenocarcinoma (n=24), NSCLC squamous cell carcinoma (n=13), SCLC (n=4) and a heterogeneous group consisting of different benign lung diseases (n=19). Healthy controls were collected separately (n=17). Circulating miRNA was processed using the extraction-free library preparation miRNA Whole Transcriptome Assay with probes for 2083 human mature miRNAs and analyzed with massive parallel sequencing. Differential expression between groups was estimated using edgeR. MiRNAs that had the highest impact on patient grouping were used in a sPLS discriminant analysis. The resulting classification model was validated using the leave-one-out method. Results The final model for comparison between patients with benign lung disease and patients with lung cancer contained 19 miRNAs. The model had an error rate of 15 % with errors distributed evenly between groups. A sub-analysis of patients with mutations in EGFR (n=5) and KRAS (n=6) was performed showing two distinct patterns in miRNA expression. Conclusion MiRNA shows promise as a circulating biomarker for lung cancer but may not be sufficient as an independent classifier. The predictive power may be improved by using several biomarkers in combination. The difference in expression between tumors with different mutations may be derived from alternate driving processes in these tumors.Sequencing results were standardized as counts per million (cpm), miRNA with cpm < 100 was filtered out and quantile normalization and log2 transformation was performed. Differential expression was analyzed using regression model (R software v. 3.5.1, package edgeR v. 3.24.1) with Benjamini-Hochberg correction. The miRNAs that, after correction, had a significant impact on sample groups were kept and analyzed with sparse partial least squares-regression. The resulting model was validated using leave-one-out method.