This study analyzed whether lymphoblastoid cell lines (LCLs) can be used to identify differential gene expression pathways that distinguish high and low suicide risk in bipolar disorder (BD) patients, and whether machine learning (ML) models can accurately predict suicide risk based on these molecular signatures. Researchers conducted RNA sequencing on LCLs derived from 20 Caucasian BD patients (some of whom later died by suicide, while others had no suicide attempts or family history). Analysis revealed 841 differentially expressed genes (DEGs) between the ‘SUICIDE’ and ‘NON-SUICIDE’ groups. These genes were enriched for primary immunodeficiency, ion channels, cell adhesion, neural function, cardiovascular regulation, and aging, suggesting multifaceted involvement in DEGs linked to suicidality. Notable genes included LCK, KCNN2, and GRIA1, which point to convergent immune, cardiovascular, and neural mechanisms. Using feature selection, the researchers generated a top 10 gene predictor panel and trained five ML algorithms. Cross-validated performance was markedly high, with accuracies of 99.67% for logistic regression, 95.67% for random forest, and 99% for the neural network model. Interestingly, several identified genes overlapped with those previously linked to lithium responsiveness, highlighting the shared biological pathways that may contribute to both treatment resistance and suicidality in BD. The researchers propose that using blood based biomarkers in translational psychiatry could augment clinical risk stratification for suicidality in BD.