A transcriptomic five-gene signature for sorafenib response prediction in hepatocellular carcinoma: clinical promise and limitations

Among primary liver cancers, hepatocellular carcinoma (HCC), which arises from malignant hepatocyte-related cells, is the most common type, constituting the leading cause of mortality among patients with chronic liver disease (1). This tumor poses a major challenge in digestive oncology due to its high incidence, high mortality, and marked biological heterogeneity. In most cases, HCC arises in the context of chronic liver injury and cirrhosis, complicating both treatment selection and prognosis. Despite advances in surveillance strategies and therapeutic options, a substantial proportion of patients are still diagnosed at advanced stages, when curative interventions such as surgical resection, local ablation, or liver transplantation are no longer feasible (2). An updated algorithm for recommended treatments based on the Barcelona Clinic Liver Cancer (BCLC) staging has recently been published (3).

For intermediate-stage unresectable disease, transarterial chemoembolization (TACE) remains a standard approach, providing survival benefit in well-selected patients (4). However, many individuals eventually progress beyond the window for locoregional therapies, underscoring the importance of systemic treatment strategies. Systemic therapy with tyrosine kinase inhibitors (TKIs) and immunomodulatory agents plays a central role in the modern management of patients with advanced HCC (5). Sorafenib has been the TKI cornerstone of first-line systemic therapy for these patients for more than a decade, and in 2026 still remains an option among the combination of drugs used as first- and second-line treatments for advanced HCC (3). However, sorafenib is being replaced by other TKIs, such as lenvatinib and regorafenib, which have also shown positive results in phase III clinical trials for unresectable HCC (6). More recently, immune checkpoint inhibitors (ICIs) targeting the PD-1/PD-L1 pathway have become preferred treatment options, including combinations such as atezolizumab plus bevacizumab, which showed clinically significant survival benefits, in terms of overall survival (OS), compared with sorafenib in the clinical trial IMbrave150 (7), the combination of nivolumab and ipilimumab evaluated in the CheckMate 9DW trial (8) and durvalumab and tremelimumab in the HIMALAYA trial (9). Nevertheless, the clinical benefit of both sorafenib and more recently approved systemic regimens remains limited and highly heterogeneous among patients, with frequent resistance mechanisms and significant adverse effects. Objective tumor responses remain relatively infrequent, and many patients discontinue therapy because of toxicity or disease progression (10,11).

Consequently, predicting therapeutic response to sorafenib in HCC is critically important, especially in certain geographic regions where immunotherapy may still be difficult to access. Nevertheless, robust biomarkers that identify patients most likely to benefit from sorafenib remain lacking (12). Although several attempts have been made to identify serum (13,14) and genetic (15) biomarkers that could relate pharmacological response to prognostic signatures, no current protocols have reached routine clinical practice. Traditional prognostic frameworks, such as the BCLC classification (16) and tumor-node-metastasis (TNM) staging (17), provide useful clinical stratification but are limited by the pronounced molecular and microenvironmental heterogeneity of HCC. Increasing evidence indicates that interindividual differences in tumor biology, immune circumstances, and stromal interactions strongly influence treatment outcomes.

In this context, attention has increasingly focused on the tumor microenvironment (TME), a complex ecosystem comprising malignant cells, immune infiltrates, cancer-associated fibroblasts, endothelial cells, pericytes, and extracellular matrix components. Depending on its composition and functional state, the TME may exert either tumor-suppressing or tumor-promoting effects (18). The TME is not only a passive scaffold but also an active driver of therapeutic response and immune evasion. Indeed, microenvironmental interactions have been implicated in modulating response to sorafenib, as shown by studies demonstrating that suppression of Sema3C sensitizes HCC to sorafenib in vivo (19). However, the molecular mechanisms linking the liver microenvironment to sorafenib resistance remain incompletely understood.

To address this unmet need, the study by Zhao et al. (20) sought to identify novel genetic biomarkers and transcriptomic signatures associated with differential sensitivity to sorafenib by integrating large-scale data from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO). They also examined the relationship between this molecular signature and immune cell infiltration to define patient subgroups most likely to benefit from sorafenib treatment.

This study makes a significant contribution to precision medicine in HCC by developing and validating a five-gene predictive signature (DNASE1L3, ACSL6, ACAN, BRSK1, and CD68) associated with sorafenib sensitivity and patient survival (Figure 1). To construct this model, the authors first performed an in silico estimation of drug response by integrating transcriptomic profiles with pharmacogenomic data from the Genomics of Drug Sensitivity in Cancer (GDSC) database. Notably, lower estimated half-maximal inhibitory concentration (IC₅₀) values correlated with improved survival, supporting the relevance of this computational approach for patient stratification.

Figure 1 Schematic representation of the methods and main results of the study by Zhao et al. (20). DEG, differentially expressed gene; GDSC, Genomics of Drug Sensitivity in Cancer; LASSO, Least Absolute Shrinkage and Selection Operator; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas; WGCNA, Weighted Gene Co-expression Network Analysis.

In a second step, Weighted Gene Co-expression Network Analysis (WGCNA) was used to identify biologically coherent gene clusters associated with sorafenib sensitivity. In parallel, differential expression analyses compared tumor versus normal tissues and patients who responded to sorafenib with those who did not. The intersection of these datasets refined the candidate gene set, yielding 56 common genes. Functional annotation using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses highlighted key processes involved in therapeutic response, including apoptosis, proliferation, xenobiotic metabolism, and drug resistance mechanisms.

Subsequently, prognostic genes were further filtered using univariate Cox regression followed by Least Absolute Shrinkage and Selection Operator (LASSO) regression, yielding the final five-gene signature. The derived risk score (RS) stratified patients into high- and low-risk groups with significant differences in OS, as confirmed by Kaplan-Meier analyses. The robustness of the proposed model was further supported by validation in an independent cohort (GSE76427), in which it retained acceptable predictive performance, with area under the curve (AUC) values exceeding 0.7 for 1-, 3-, and 5-year survival.

To enhance clinical applicability, the authors developed a nomogram that integrates the molecular RS with established clinicopathological variables, particularly pathological stage. Such integrative approaches may provide clinicians with more accurate tools for pre-treatment stratification and therapeutic decision-making.

In our opinion, a particularly valuable aspect of the study is the incorporation of immune TME analyses, which enhances the model’s biological interpretability. High-risk tumors were characterized as immunologically “cold”, exhibiting an immunosuppressive phenotype with increased infiltration by regulatory T cells and M2-like macrophages. Conversely, low-risk tumors displayed an immune “hot” profile enriched for cytotoxic CD8⁺ T cells (21). Notably, CD68, a marker of tumor-associated macrophages, was highly expressed in the high-risk group, supporting the hypothesis that macrophage-driven immunosuppression contributes to therapeutic resistance (22). These findings suggest that sorafenib sensitivity may reflect not only tumor-intrinsic genetic factors but also broader immune microenvironmental states.

In the study by Zhao et al. (20), we could identify several methodological strengths. Its comprehensive bioinformatic framework integrates multiple independent datasets, increasing robustness and reducing cohort-specific bias. The combination of linear models for microarray data (LIMMA), WGCNA, Cox regression, and LASSO provides a rigorous strategy for biomarker discovery, while reducing the signature to five genes enhances translational feasibility. A particularly positive and interesting aspect that we would like to highlight is that this study does not limit itself to prediction but also performs a TME analysis, using tools widely used to characterize the TME with transcriptomic datasets, such as CIBERSORT (Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts), a computational deconvolution algorithm that estimates the relative proportions of different immune cell types within tumor tissues based on gene expression data, and ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data), which infers the overall presence of stromal and immune components and provides an indirect estimate of tumor purity.

Nevertheless, in our opinion, important limitations must be acknowledged. First, the study is retrospective and based exclusively on in silico estimates. Sorafenib sensitivity is inferred indirectly from pharmacogenomic data derived from cancer cell lines; therefore, reported IC₅₀ values should be interpreted with caution, as they are computational approximations rather than direct clinical measurements. While such approaches are widely used for large-scale gene screening, they do not fully capture the complexity of drug response in patients. Additionally, clinical heterogeneity across datasets, such as differences in etiology, liver function, or prior treatments, is not fully controlled. Although external validation was performed, public cohorts often lack complete clinical annotation. Nevertheless, despite the inherent limitations of this type of studies, they often constitute the only practical approach for screening thousands of genes prior to undertaking costly experimental validation.

Unfortunately, the study by Zhao et al. (20) lacks such a functional evaluation of the identified genes in the context of sorafenib resistance mechanisms. This is a critical limitation because the absence of in vitro and in vivo validation restricts mechanistic interpretation and translational relevance. Although the role of CD68-positive tumor-associated macrophages in reduced sorafenib sensitivity is relatively well supported (23,24), the mechanistic contribution of other genes, such as ACAN or BRSK1, remains insufficiently characterized. Further experimental studies will be essential to establish causal relationships and confirm their potential as therapeutic targets.

The rapidly evolving systemic treatment landscape for advanced HCC poses challenges for the immediate clinical application of sorafenib-specific biomarkers. Although immunotherapy-based combinations are increasingly replacing sorafenib in first-line settings, sufficiently large and well-annotated datasets of patients treated with newer regimens are not yet widely available, limiting the development of comparable predictive models.

The biological heterogeneity of HCC continues to drive the development of personalized medicine strategies. Despite therapeutic advances, sorafenib remains widely used across many clinical contexts, and the proposed five-gene signature provides a robust framework for patient stratification. By linking transcriptomic predictors to distinct immune microenvironmental states, this work reinforces the promise of molecular classifiers to refine prognostic assessment, guide therapeutic decision-making, and advance precision oncology in HCC.

As systemic treatment options expand and the field moves towards individualized therapy, molecular stratification tools are likely to become increasingly important not only for prognostic refinement but also for optimizing treatment selection and sequencing. More precisely, gene-expression signatures may help address a key limitation in current HCC management: the lack of reliable predictors of therapeutic benefit (12). While clinical frameworks such as BCLC remain essential for treatment allocation (16) and staging systems such as TNM provide additional prognostic information (25), they do not fully capture the underlying biological diversity. This diversity determines response or resistance to systemic agents, which is increasingly recognized as a hallmark of HCC heterogeneity (17). Molecular classifiers could therefore serve as complementary tools, enabling clinicians to better distinguish patients likely to derive meaningful benefit from TKIs such as sorafenib from those who might require alternative approaches.

Nevertheless, several barriers must be overcome before such signatures can be implemented in routine practice. A major challenge is the need for prospective validation in well-characterized cohorts with documented clinical responses. Retrospective analyses of public datasets provide valuable hypothesis-generating evidence, but clinical translation ultimately requires validation on standardized platforms and with real-world treatment outcomes. Furthermore, technical difficulties, including tumor sampling variability, intratumoral heterogeneity, and differences among transcriptomic profiling methodologies, may affect reproducibility across institutions.

Another important future direction is integrating transcriptomic signatures with additional layers of biological information. HCC is increasingly recognized as a disease shaped by multidimensional interactions among genetic alterations, epigenetic regulation, metabolic rewiring, and immune microenvironmental dynamics (15,18,26). Predictive models that incorporate immune profiling, circulating biomarkers, radiomics, and clinical variables may outperform gene-expression signatures alone.

Moreover, the rapidly evolving systemic therapy landscape raises important questions regarding the current clinical relevance of sorafenib-specific biomarkers. Although immunotherapy-based combinations have become the preferred first-line treatment for advanced HCC, as supported by recent clinical trials such as IMbrave150 (7), CheckMate 9DW (8), and HIMALAYA, mentioned above (9), TKIs, including sorafenib and mainly lenvatinib, continue to play a substantial role in disease management (5). These agents remain particularly relevant in patients with contraindications to immunotherapy, in settings with limited access to newer agents, or as later-line options (27). Importantly, molecular insights derived from sorafenib response may extend beyond this specific drug, providing broader insights into tumor biology, including aggressiveness, immunosuppression, and therapeutic vulnerability.

Furthermore, the association between the proposed gene signature and the tumor immune microenvironment underscores its potential relevance in the era of immunotherapy. The identified molecular features may reflect underlying immune states that influence not only response to sorafenib but also sensitivity to immune-based treatments. In this context, transcriptomic signatures incorporating immune-related components could contribute to patient stratification for current combination therapies and to a more profound understanding of their mechanisms of action.

Finally, the link between risk stratification and immune contexture identified in this study suggests opportunities for rational design of combination strategies. Patients with “cold” tumors, characterized by macrophage-driven immunosuppression, may benefit from therapies that modulate the microenvironment or overcome immune exclusion, whereas “hot” tumors enriched in cytotoxic T cells may be better candidates for immunotherapy-based regimens (21). Tumor-associated macrophages, often identified by CD68 expression, have been linked to treatment resistance and immune evasion in HCC (24).

Thus, transcriptomic biomarkers, such as the one proposed in the study, may ultimately contribute not only to treatment selection but also to the development of novel combinatorial approaches tailored to specific immunological states or molecular vulnerabilities. In particular, components of the identified gene signature may point to distinct therapeutic avenues. For instance, CD68 reflects the presence of tumor-associated macrophages, which are key mediators of immunosuppression and may be targets for therapies aimed at modulating the TME or enhancing responses to ICIs. In addition, other genes within the signature, such as DNASE1L3, ACSL6, ACAN, and BRSK1, may be associated with biological processes including regulation of apoptosis, lipid metabolism, extracellular matrix organization, and stress-response signaling pathways, all of which could influence tumor progression and therapeutic resistance. Although these associations remain largely exploratory, they raise the possibility that such genes could serve as entry points for identifying novel drug targets or for the rational design of combination strategies integrating targeted agents with immunotherapy. However, it is important to emphasize that these potential applications remain speculative at present, as noted above. Only through such validation will it be possible to translate these transcriptomic findings into clinically actionable strategies.

Conclusions

In summary, studies such as that by Zhao et al. (20) provide an important foundation for precision oncology in HCC. They established a sorafenib response-related prognostic risk prediction model in HCC based on five signature genes (DNASE1L3, ACSL6, ACAN, BRSK1, and CD68), which had high predictive precision on sorafenib response. Moreover, they found significant correlations between risk groups and immunity. It should be recognized that, although further prospective validation and functional characterization are needed, molecular signatures that integrate prognostic and immune-related information hold substantial promise for refining patient stratification, improving therapeutic personalization, and advancing clinical management of this highly heterogeneous malignancy.

Acknowledgments

The authors express their gratitude to Mrs. Mariar Franco for her support with administrative tasks.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Journal of Gastrointestinal Oncology. The article has undergone external peer review.

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-1-0139/prf

Funding: This study received funding from various sources, including the CIBERehd (grant No. CB06/04/0023) and Instituto de Salud Carlos III (ISCIII) (grant No. PI25/00628), co-funded by the European Union; “Junta de Castilla y Leon” (Nos. SA113P23 and EDU/1868/2022), Spain.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-1-0139/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kumar A, Acharya SK, Singh SP, et al. The Indian National Association for Study of the Liver (INASL) Consensus on Prevention, Diagnosis and Management of Hepatocellular Carcinoma in India: The Puri Recommendations. J Clin Exp Hepatol 2014;4:S3-S26. [Crossref] [PubMed]
2. Nagalapuram V, Popuri N, Nipp RD, et al. Management of Advanced Hepatocellular Carcinoma: A Review and Practical Guide. JCO Oncol Pract 2025;21:1247-56. [Crossref] [PubMed]
3. Reig M, Sanduzzi-Zamparelli M, Forner A, et al. BCLC strategy for prognosis prediction and treatment recommendations: The 2026 update. J Hepatol 2026;84:631-54. [Crossref] [PubMed]
4. Ye K, Zhou J, Lin J, et al. Efficacy of TACE plus tyrosine kinase inhibitors and immune checkpoint inhibitors in patients with unresectable hepatocellular carcinoma: a systematic review and meta-analysis. Front Oncol 2025;15:1707622. [Crossref] [PubMed]
5. European Association for the Study of the Liver. EASL Clinical Practice Guidelines on the management of hepatocellular carcinoma. J Hepatol 2025;82:315-74. [Crossref] [PubMed]
6. Yang X, Zhang Q, Li D, et al. A Multifunctional Nanodrug Increases the Therapeutic Sensitivity of Lenvatinib to Hepatocellular Carcinoma by Inhibiting the Stemness of Hepatic Cancer Stem Cells. Adv Healthc Mater 2024;13:e2401398. [Crossref] [PubMed]
7. Cheng AL, Qin S, Ikeda M, et al. Updated efficacy and safety data from IMbrave150: Atezolizumab plus bevacizumab vs. sorafenib for unresectable hepatocellular carcinoma. J Hepatol 2022;76:862-73.
8. Yau T, Galle PR, Decaens T, et al. Nivolumab plus ipilimumab versus lenvatinib or sorafenib as first-line treatment for unresectable hepatocellular carcinoma (CheckMate 9DW): an open-label, randomised, phase 3 trial. Lancet 2025;405:1851-64. [Crossref] [PubMed]
9. Rimassa L, Chan SL, Sangro B, et al. Five-year overall survival update from the HIMALAYA study of tremelimumab plus durvalumab in unresectable HCC. J Hepatol 2025;83:899-908. [Crossref] [PubMed]
10. Shi Y, Cui D, Xia L, et al. Efficacy and safety of lenvatinib plus gefitinib in lenvatinib-resistant hepatocellular carcinomas: a prospective, single-arm exploratory trial. Signal Transduct Target Ther 2024;9:359. [Crossref] [PubMed]
11. Zhang W, Hong X, Xiao Y, et al. Sorafenib resistance and therapeutic strategies in hepatocellular carcinoma. Biochim Biophys Acta Rev Cancer 2025;1880:189310. [Crossref] [PubMed]
12. Galun D, Mijac D, Filipovic A, et al. Precision Medicine for Hepatocellular Carcinoma: Clinical Perspective. J Pers Med 2022;12:149. [Crossref] [PubMed]
13. Narahara S, Watanabe T, Nagaoka K, et al. Clusterin and Related Scoring Index as Potential Early Predictors of Response to Sorafenib in Hepatocellular Carcinoma. Hepatol Commun 2022;6:1198-212. [Crossref] [PubMed]
14. Song S, Bai M, Li X, et al. Early predictive value of circulating biomarkers for sorafenib in advanced hepatocellular carcinoma. Expert Rev Mol Diagn 2022;22:361-78. [Crossref] [PubMed]
15. Sanchez-Martin A, Macias RIR, Muñoz-Bellvís L, et al. A three-gene resistome signature as a prognostic tool in hepatocellular carcinoma. Ann Hepatol 2026;31:102179. [Crossref] [PubMed]
16. Llovet JM, Brú C, Bruix J. Prognosis of hepatocellular carcinoma: the BCLC staging classification. Semin Liver Dis 1999;19:329-38. [Crossref] [PubMed]
17. Safri F, Nguyen R, Zerehpooshnesfchi S, et al. Heterogeneity of hepatocellular carcinoma: from mechanisms to clinical implications. Cancer Gene Ther 2024;31:1105-12. [Crossref] [PubMed]
18. de Visser KE, Joyce JA. The evolving tumor microenvironment: From cancer initiation to metastatic outgrowth. Cancer Cell 2023;41:374-403. [Crossref] [PubMed]
19. Peng H, Yang M, Feng K, et al. Semaphorin 3C (Sema3C) reshapes stromal microenvironment to promote hepatocellular carcinoma progression. Signal Transduct Target Ther 2024;9:169. [Crossref] [PubMed]
20. Zhao Y, Dong J, Zhong H, et al. Comprehensive bioinformatic analysis reveals sorafenib response-related prognostic signature in hepatocellular carcinoma. J Gastrointest Oncol 2025;16:2176-92. [Crossref] [PubMed]
21. Donne R, Lujambio A. The liver cancer immune microenvironment: Therapeutic implications for hepatocellular carcinoma. Hepatology 2023;77:1773-96. [Crossref] [PubMed]
22. Li X, Li Z, Zhai J, et al. A Macrophage-Derived 7-Gene Signature Predicts Prognosis and Therapeutic Response in Hepatocellular Carcinoma. IUBMB Life 2025;77:e70079. [Crossref] [PubMed]
23. Sariyar E, Firtina Karagonlar Z. Modelling the Sorafenib-resistant Liver Cancer Microenvironment by Using 3-D Spheroids. Altern Lab Anim 2023;51:301-12. [Crossref] [PubMed]
24. Han JW, Kim JH, Kim DH, et al. Higher Number of Tumor-Infiltrating PD-L1+ Cells Is Related to Better Response to Multikinase Inhibitors in Hepatocellular Carcinoma. Diagnostics (Basel) 2023;13:1453. [Crossref] [PubMed]
25. Amin MB, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging. CA Cancer J Clin 2017;67:93-9.
26. Su L, Zhang G, Kong X. A Novel Five-Gene Signature for Prognosis Prediction in Hepatocellular Carcinoma. Front Oncol 2021;11:642563. [Crossref] [PubMed]
27. Marin JJG, Romero MR, Herraez E, et al. Mechanisms of Pharmacoresistance in Hepatocellular Carcinoma: New Drugs but Old Problems. Semin Liver Dis 2022;42:87-103. [Crossref] [PubMed]