Linezolid is an essential drug for treating multidrug-resistant and extensively drug-resistant tuberculosis, yet its long-term use is frequently limited by peripheral neuropathy, which in severe cases may be irreversible. To investigate biological factors underlying susceptibility, we analyzed transcriptomic profiles of peripheral blood mononuclear cells from 51 MDR/XDR-TB patients before and after six months of linezolid treatment, stratifying individuals into asymptomatic, mild, and severe neuropathy groups. Patients who later developed severe neuropathy already exhibited distinct transcriptional signatures at baseline, including evidence of immune dysregulation and impaired antioxidant defense, with downregulation of genes involved in immune receptor activity and the pentose phosphate pathway. Following treatment, these patients showed limited transcriptional responses compared with other groups, most notably a marked downregulation of IL10, a central anti-inflammatory cytokine, suggesting maladaptive systemic regulation. Using a machine learning–based feature selection approach, we derived an 11-gene predictive model that achieved strong performance with a cross-validated AUC of 0.93. Model interpretation using SHAP values highlighted HP and HSPA1B as the most influential predictors. Overall, our findings suggest that severe linezolid-induced peripheral neuropathy is predisposed by baseline immune and metabolic alterations and that patients with this complication fail to mount appropriate systemic responses during treatment.