Integrated machine learning highlights MYBL2 as a prognostic marker and potential therapeutic target in hepatocellular carcinoma
Why hepatocellular carcinoma needs new prognostic markers
A study found that MYBL2 expression in hepatocellular carcinoma (HCC) is associated with poor prognosis, and that adding expression data to conventional staging improves survival prediction. Because HCC treatment outcomes remain poor, new biomarkers are needed to support diagnosis, prognosis, and treatment selection. The researchers focused on MYBL2, which is involved in cell-cycle regulation. They examined whether adding its expression data to existing staging information could improve prognostic prediction, and how it relates to cancer-cell growth and drug sensitivity.
MYBL2 assessed with machine learning and experiments
The researchers applied LASSO-based variable selection to The Cancer Genome Atlas (TCGA) data and validated the prognostic association of MYBL2 in independent cohorts from the International Cancer Genome Consortium (ICGC) and the Gene Expression Omnibus (GEO). They also incorporated single-cell transcriptomic analyses and analyses of public CRISPR dependency-screening data. In analyses distinguishing tumor from normal liver tissue, the area under the receiver operating characteristic curve (AUC) for MYBL2 was 0.968. This measures tissue-classification performance in the analyzed data; it does not establish the accuracy of an early-detection test for patients.
High MYBL2 expression was associated with higher tumor grade, microvascular invasion, hepatitis B virus (HBV) positivity, nonresponse to transarterial chemoembolization (TACE), and poorer survival outcomes. A model combining American Joint Committee on Cancer (AJCC) stage with MYBL2 expression had 1-, 3-, and 5-year survival-prediction AUCs of 0.73, 0.77, and 0.72, respectively, compared with 0.65, 0.69, and 0.67 for stage alone. Improved predictive performance should not be interpreted as improved patient survival.
MYBL2 was positively correlated with alpha-fetoprotein (AFP), MKI67, PCNA, and BIRC5. Analysis of public Cancer Dependency Map (DepMap) data found that growth was inhibited in 21 of 23 HCC cell lines after MYBL2 was removed. In pharmacogenomic analyses, high MYBL2 expression was associated with a lower half-maximal inhibitory concentration (IC50) for sorafenib, indicating greater sensitivity. A separate transcriptomic analysis also found a correlation with indicators of an immunosuppressive microenvironment. Neither finding establishes a causal relationship with patients’ drug responses or immune regulation.
miR-29a regulation and the need for clinical validation
In protein analyses and luciferase reporter experiments using HepG2 cells, the researchers presented evidence that the microRNA miR-29a binds directly to the 3′ untranslated region of MYBL2 and suppresses its expression. The mouse experiment included 16 animals divided into four groups of four. In a 25-week model of liver cancer induced by a Western diet and carbon tetrachloride, mice with miR-29a overexpression did not develop tumors, unlike wild-type mice under the same conditions; MYBL2 expression and fibrosis were also reduced.
However, these mice were engineered to overexpress miR-29a continuously from development. This differs from administering a treatment to animals with established tumors. The study also lacked a rescue experiment to determine whether restoring MYBL2 expression reverses the effects, so the full antitumor effect cannot be attributed solely to MYBL2 suppression. The patient cohorts were retrospective, and the sorafenib-sensitivity finding came from computational analysis based on cell-line data. Prospective clinical validation and further preclinical experiments are needed before these results can lead to a diagnostic kit or treatment.
The findings and limitations are described in the full paper.
BACKGROUND: Hepatocellular carcinoma (HCC) presents with poor treatment outcomes, creating an urgent need for novel biomarkers to improve diagnosis, prognosis, and precision medicine. While the MYB family of oncogenes is implicated in cancer, the role and regulatory mechanisms of its member, particularly MYB proto‑oncogene like 2 (MYBL2), remain underexplored in HCC. Therefore, this study aimed to systematically validate the clinical significance of MYBL2, elucidate its functional role in tumor progression and drug sensitivity, and identify its upstream regulatory mechanisms using an integrative machine learning and experimental framework.
METHODS: We applied an integrative pipeline combining LASSO‑based feature selection on TCGA and GEO cohorts, single‑cell transcriptomics, pharmacogenomic surveys, and CRISPR dependency screens. These computational approaches were complemented by experimental validation.
RESULTS: MYBL2 robustly discriminated tumor from normal liver (AUC = 0.968), and high expression was associated with adverse features, including higher grade, microvascular invasion, HBV positivity, non‑response to TACE, and worse survival. A nomogram combining MYBL2 with AJCC stage improved 1‑, 3‑, and 5‑year AUCs versus stage alone. MYBL2 correlated with proliferative biomarkers (AFP, MKI67, PCNA, BIRC5) and CRISPR knockout inhibited growth in most HCC lines. High MYBL2 expression was associated with greater sensitivity to sorafenib in pharmacogenomic screens and was linked to an immunosuppressive microenvironment and higher MSI. Mechanistically, miR‑29a was shown to suppress MYBL2 translation by directly binding to its 3'UTR.
CONCLUSIONS: MYBL2 is a potent diagnostic and prognostic biomarker in HCC that also predicts sorafenib sensitivity. Our findings establish a clear regulatory link where MYBL2 is a direct and functionally important target of the tumor‑suppressive miR‑29a. This positions MYBL2 as a tractable target for miR‑29a‑based therapeutic strategies, warranting clinical validation.
MYBL2 is a candidate prognostic marker that could add molecular information to existing staging for patients with HCC. Evaluating its practical use requires prospective patient cohorts to establish whether measuring MYBL2 helps stratify risk and guide care. Its potential use for identifying patients for sorafenib treatment also remains unvalidated. In therapeutic development, the miR-29a–MYBL2 regulatory relationship could provide a starting point for further experiments. However, rescue experiments restoring MYBL2 are needed to isolate its contribution, along with studies testing miR-29a delivery in already established tumors. These results alone do not provide a basis for expecting improved survival or quality of life for patients.