ACO-Driven Chimeric RNA Signatures: A Novel Biomarker Panel for Metastasis Destination Prediction

Background Metastasis remains the leading cause of cancer-related mortality, with organ-specific tropism critically influencing prognosis and therapeutic decision-making. Despite advances in genomic profiling, robust biomarkers capable of predicting metastatic destination remain elusive. Chimeric RNAs—transcripts formed by the fusion of two distinct genes—have emerged as promising cancer biomarkers, yet their potential for predicting organ-specific metastasis has not been systematically explored. Methods We developed an Ant Colony Optimization (ACO)-driven computational framework to discover chimeric RNA signatures predictive of metastatic destination. Gene expression data from 289 metastatic samples across 15 anatomical sites were obtained from the GEO database (GSE205154) and normalized using TPM. ACO was employed to identify tissue-specific chimeric RNA pairs through iterative pheromone-guided feature selection, with discriminative power evaluated via Random Forest classifiers and 3-fold cross-validation. The predictive performance of the identified signatures was assessed using 5-fold cross-validation, and feature importance was determined by Gini impurity. Results ACO identified 15 tissue-specific chimeric RNA signatures: 5 for liver, 5 for peritoneum, and 5 for lung. The top-ranked signatures included ENSG00000111640 + ENSG00000167526 for lung (accuracy = 0.9594) and ENSG00000163631 + ENSG00000173432 for liver (importance = 0.1756). The chimeric RNA panel achieved an overall accuracy of 91.46% in predicting metastatic destination. Liver metastasis exhibited high precision (0.90) and recall (1.00), while peritoneum metastasis showed high precision (1.00) but lower recall (0.20). Feature importance analysis revealed that liver-specific chimeric RNAs played a dominant role in discriminating metastatic sites. Conclusions We present a novel ACO-driven framework for discovering chimeric RNA signatures that predict metastatic destination with high accuracy (91.46%). These findings establish chimeric RNAs as promising biomarkers for organ-specific metastasis and provide a foundation for developing diagnostic assays for personalized metastasis prediction. Future work will focus on experimental validation and expansion to other cancer types.