A recurrence model for laryngeal cancer based on SVM and gene function clustering.

CONCLUSION: A prognostic model was obtained for LC. Several critical genes were unveiled. They could be potentially applied for LC recurrence prediction.

OBJECTIVE: Gene expression data of laryngeal cancer (LC) were analyzed to identify critical genes associated with recurrence.

METHODS: Two gene expression datasets were downloaded from the Gene Expression Omnibus. Dataset GSE27020 is used as the training set, containing 75 non-recurred LC cases and 34 recurred LC cases.

RESULTS: A total of 725 DEGs were identified from the training set. A total of 4126 gene pairs showed significant correlations in non-recurred LC only, corresponding to 533 genes. A total of 7235 gene pairs showed significant correlations in recurred LC only, corresponding to 608 genes. Besides, 1694 gene pairs showed significant correlations in both non-recurred and recurred LC, corresponding to 322 genes. Functional enrichment analysis was performed for the three groups of DEGs. Seven overlapping biological functions were revealed: positive regulation of chondrocyte differentiation, autoimmune thyroid disease, focal adhesion, linoleic acid metabolism, drug metabolism, organic cation transport, and ECM-receptor interaction. Eight feature genes (PDIA3, MYH11, PDK1, SDC3, RPE65, LAMC3, BTK, and UPK1B) were identified. Their prognostic effect was validated by independent test set as well as survival analysis.