In early-stage non-small-cell lung cancer (NSCLC), a substantial proportion of patients can be cured by surgery. Development of distant metastases is the most frequent cause of therapeutic failure. The possibility to accurately predict a patient's risk for developing distant metastasis would help to identify patients that are candidates for further intervention such as conventional adjuvant chemotherapy or experimental drugs. Current molecular biology techniques allow the genome-wide screening for differentially expressed genes; and adequate bioinformatics approaches are developed at a rapid pace to improve prognosis prediction. Genes associated with metastasis do not necessarily play a role in disease pathogenesis but rather reflect the activation of specific signal-transduction pathways that are associated with enhanced migration and invasion capability. In our own work, we have identified several genes (e.g. thymosin beta-4, elF4A1), including a novel non-coding RNA (MALAT-1) to be expressed at significantly higher levels in stage-I and stage-II NSCLC primary tumours that subsequently metastasised. As a consequence, patients with high-level expression of these genes were shown to have significantly worse survival compared to patients with low-level expression of these genes. These data support the hypothesis that gene-expression patterns in primary tumours determine the tumours' likelihood to metastasise. In the near future, this information will be used for tailored therapy approaches for patients with early-stage NSCLC.