Extracellular vesicle (EV) microRNAs are of major interest as potential diagnostic biomarkers in all cancer types. This study aims to identify miRNA profiles of shed microvesicles (sMVs) and exosomes (Exos) secreted from the isogenic colorectal cancer (CRC) cell lines SW480 and SW620 and evaluate their ability to predict CRC. Deep sequencing of miRNAs in parental cell lysates (CLs) and highly-purified sMVs and Exos was performed. We focused on miRNAs enriched in EVs and dysregulated miRNAs in metastatic cells (SW620) relative to primary cancer cells (SW480). We investigated the ability of EV miRNA signatures to predict CRC tumours using 594 tumours (representing different pathological stages) and 11 normal samples obtained from TCGA. In SW480 and SW620 cells we identified 345 miRNAs, of which 61 and 73 were upregulated and downregulated in SW620-CLs compared to SW480-CLs, respectively. Selective distribution of cellular miRNAs into EVs results in distinct miRNA signatures for sMVs and Exos in each cell line. Cross cell line comparisons of EV miRNA profiles reveal a subset of miRNAs critical in CRC progression from primary carcinoma to metastasis. Many miRNAs non-detectable (<5 TPM) in CLs were significantly enriched (>1000 TPM) in secreted EVs. Strikingly, miR-7641 which is non-detectable in SW480-CL but upregulated in SW620-CL is highly enriched in EVs secreted from both cell lines. Pearson correlation analysis demonstrated that EV miRNA profiles can be used to predict CRC tumours with ~96% accuracy. Our findings suggest that EV miRNA profiles from CRC cell lines may allow prediction of CRC tumours, and that miR-7641 may serve as an attractive candidate for the specific, non-invasive diagnosis and prognosis of CRC.