Abstract
BACKGROUND: Carboplatin/paclitaxel (CP), with or without sorafenib, result in objective response rates of 18-20 % in unselected chemotherapy-naïve patients. Molecular predictors of survival and response to CP-based chemotherapy in metastatic melanoma (MM) are critical to improving the therapeutic index. Intergroup trial E2603 randomized MM patients to CP with or without sorafenib. Expression data were collected from pre-treatment formalin-fixed paraffin-embedded (FFPE) tumor tissues from 115 of 823 patients enrolled on E2603. The selected patients were balanced across treatment arms, BRAF status, and clinical outcome. We generated data using Nanostring array (microRNA (miRNA) expression) and DNA-mediated annealing, selection, extension and ligation (DASL)/Illumina microarrays (HT12 v4) (mRNA expression) with protocols optimized for FFPE samples. Integrative computational analysis was performed using a novel Tree-guided Recursive Cluster Selection (T-ReCS) [1] algorithm to select the most informative features/genes, followed by TargetScan miRNA target prediction (Human v6.2) and mirConnX [2] for network inference. RESULTS: T-ReCS identified PLXNB1 as negatively associated with progression-free survival (PFS) and miR-659-3p as the primary miRNA associated positively with PFS. miR-659-3p was differentially expressed based on PFS but not based on treatment arm, BRAF or NRAS status. Dichotomized by median PFS (less vs greater than 4 months), miR-659-3p expression was significantly different. High miR-659-3p expression distinguished patients with responsive disease (complete or partial response) from patients with stable disease. miR-659-3p predicted gene targets include NFIX, which is a transcription factor known to interact with c-Jun and AP-1 in the context of developmental processes and disease. CONCLUSIONS: This novel integrative analysis implicates miR-659-3p as a candidate predictive biomarker for MM patients treated with platinum-based chemotherapy and may serve to improve patient selection.