Machine learning-driven discovery of anoikis-related biomarkers in Adult T-Cell Leukemia/Lymphoma subtypes

Abstract

Adult T-Cell Leukemia/Lymphoma (ATLL) is a malignancy that arises from T-cells infected with the human T-cell lymphotropic virus type 1 (HTLV-1). The disease is characterized by uncontrolled proliferation and reduced apoptosis of malignant T cells, which contributes to tumor progression and resistance to therapy. Anoikis is a specific form of programmed cell death triggered by the loss of cell–matrix or cell–cell adhesion, playing a critical role in preventing detached cells from surviving and forming tumors. Dysregulation of anoikis has been implicated in cancer metastasis and therapeutic resistance across various malignancies; however, its role in ATLL remains largely unexplored. To our knowledge, this is the first study to investigate anoikis-related genes in ATLL subtypes, particularly across its major subtypes: acute, chronic, and smoldering. In this study, we explored anoikis-related differentially expressed genes to identify those specifically associated with each subtype. We then applied Least Absolute Shrinkage and Selection Operator (LASSO) regression to select the most informative features. Subsequently, we employed decision trees, random forest, extreme gradient boosting, support vector machine, and logistic regression algorithms to identify classifier genes distinguishing each ATLL subtype from asymptomatic carriers. The identified biomarkers include SMARCE1 and CASP3 for acute, TGFΒ1 and MTA1 for chronic, and CXCL1 and LGALS8 for smoldering subtypes. These genes are involved in cell adhesion, survival signaling, and apoptosis—key processes in cellular homeostasis and oncogenesis. Our findings provide novel insights into the molecular mechanisms linking anoikis to ATLL subtypes and highlight potential therapeutic targets.