Toward Precision Surgery in Rectal Cancer: Integrating Machine Learning for Molecular Subtype Identification and Surgical Planning

Abstract

Because rectal cancer is a diverse disease, improving patient outcomes requires precision medicine and individualized surgical techniques. To identify important molecular subtypes that could affect therapeutic decision-making and precision surgery, this study combined unsupervised machine learning techniques with high-throughput microarray analysis. Using R software (v4.0.1), microarray-based gene expression data from the GSE253106 dataset were extracted from the Gene Expression Omnibus (GEO). The molecular subtypes of rectal cancer were identified using unsupervised machine learning techniques. Markov Clustering (MCL) mapped molecular networks involved in DNA repair and microenvironment interactions, while K-Means clustering grouped genes according to similarities in their expression. Density- Based Spatial Clustering of Applications with Noise, or DBSCAN, identified uncommon tumor subtypes linked to aggressive characteristics. The findings showed that rectal cancer had 411 DEGs, 348 of which were upregulated and 63 were downregulated. Genes such as TTTY15, RPS4Y1, and KDM5D were upregulated, whereas IGF2 and INS-IGF2 were downregulated. K-means clustering highlighted immune and metabolic regulation by grouping genes, such as C3AR1, TREM2, IGF2, and APOE. Networks involving DNA repair (FANCA, DAPK1) and the tumor microenvironment (AIF1, C1QA) were mapped using Markov Clustering (MCL). Rare aggressive subtypes were identified using DBSCAN, which also identified PLTP, ISG15, and TYROBP as indicators of immune evasion. TTTY15, KDM5D, and IGF2 were identified by outlier detection, indicating their involvement in tumor progression and treatment response. By highlighting the important molecular subtypes of rectal cancer, this study demonstrates how machine learning can be used to improve precision oncology and surgical techniques. Biomarker-driven treatment strategies may benefit from additional functional validation, which would improve therapeutic results and patient stratification.