# Identifying sorafenib benefit among patients with hepatocellular carcinoma: A transcriptomic and genomic approach

**Authors:** Sun Young Yim, Hayeon Kim, Tae Hyung Kim, Sang-Hee Kang, Youngwoo Lee, Eunho Choi, Yang Jae Yoo, Seong Hee Kang, Young-Sun Lee, Young Kul Jung, Yeon Seok Seo, Hyung Joon Yim, Jong Eun Yeon, Kyung Suk Yang, Yitao Tang, Bowha Sohn, Yun Seong Jeong, Hyewon Park, Han Liang, Ju-Seog Lee, Ji Hoon Kim

PMC · DOI: 10.1016/j.jhepr.2026.101742 · JHEP Reports · 2026-01-27

## TL;DR

This study identifies a 50-gene signature called KUSS50 that can predict which patients with liver cancer will benefit from sorafenib treatment, helping to personalize therapy and improve outcomes.

## Contribution

The study introduces KUSS50, a novel 50-gene signature for predicting sorafenib benefit in hepatocellular carcinoma.

## Key findings

- The KUSS50 gene signature showed high predictive accuracy in identifying sorafenib beneficiaries (AUC: 87.1% and 90.8% in validation cohorts).
- KUSS50-defined subtypes are associated with distinct molecular features like ferroptosis activation and β-catenin mutations.
- The KUSS50 signature enables personalized treatment selection and may guide new therapeutic strategies for HCC.

## Abstract

Sorafenib has been a cornerstone of hepatocellular carcinoma (HCC) therapy; however, its efficacy is limited, and identifying patients who will benefit from sorafenib is challenging. We aimed to identify predictive biomarkers of sorafenib benefit in patients with HCC.

Gene expression data from 33 HCC tumors treated with sorafenib were analyzed to construct a prediction model aimed at identifying patients with greater benefit from sorafenib treatment. The robustness of the predictor was validated using gene expression data from two phase III clinical trials, IMbrave150 and STORM.

The analysis of transcriptome data revealed a 50-gene signature, the KUSS50 (Korea University Sorafenib Signature with 50 genes), that exhibited high predictive power in identifying patients who benefited from sorafenib treatment in a training cohort. Validation in two independent cohorts – IMbrave150 (n = 48) and BIOSTORM (n = 67) –demonstrated high specificity for predicting sorafenib benefit (AUC: 87.1%, p = 1.8 × 10-4 and 90.8%, p = 1.0 × 10-7, respectively). Genomic analyses identified distinct molecular characteristics associated with the KUSS50-defined benefit subtype, including an increased mutation rate and activation of ferroptosis, suggesting increased baseline ferroptotic activity in these HCCs, which may sensitize them to sorafenib. The benefit subtype also overlapped with previously defined HCC subtypes associated with stemness and aggressiveness. Conversely, the non-benefit subtype correlated with β-catenin mutations and increased tumor purity, underscoring its biological significance.

The KUSS50 is a clinically actionable biomarker that may optimize patient selection for sorafenib treatment in HCC, potentially improving outcomes. Further exploration of the underlying biology of KUSS50-defined subtypes – particularly the role of ferroptosis in sorafenib sensitivity – may yield additional therapeutic insights.

This study identifies the KUSS50, a novel 50-gene signature, as a predictive biomarker for identifying patients with hepatocellular carcinoma (HCC) who are likely to benefit from sorafenib treatment. The findings have significant implications for the clinical management of HCC, particularly in optimizing treatment strategies and enhancing patient outcomes. The ability to predict the benefit of sorafenib treatment with high specificity allows for more personalized therapy, reducing unnecessary exposure to ineffective treatments. This approach can be directly applied by clinicians to improve treatment selection, ultimately leading to better patient outcomes. Additionally, understanding the molecular mechanisms underlying the KUSS50-defined subtypes may pave the way for new therapeutic strategies and interventions aimed at improving the efficacy of sorafenib and other treatments in patients with HCC.

Image 1

•The KUSS50 is a novel 50-gene signature predicting the benefit of sorafenib treatment in HCC.•High predictive accuracy was validated across multiple independent cohorts.•KUSS50-defined subtypes have distinct molecular characteristics linked to treatment outcomes.•The KUSS50 signature enables personalized sorafenib therapy, improving patient selection.•Insights offer new therapeutic avenues for enhancing HCC treatment efficacy.

The KUSS50 is a novel 50-gene signature predicting the benefit of sorafenib treatment in HCC.

High predictive accuracy was validated across multiple independent cohorts.

KUSS50-defined subtypes have distinct molecular characteristics linked to treatment outcomes.

The KUSS50 signature enables personalized sorafenib therapy, improving patient selection.

Insights offer new therapeutic avenues for enhancing HCC treatment efficacy.

## Linked entities

- **Chemicals:** sorafenib (PubChem CID 216239)
- **Diseases:** hepatocellular carcinoma (MONDO:0007256), HCC (MONDO:0007256)

## Full-text entities

- **Genes:** SLTM (SAFB like transcription modulator) [NCBI Gene 79811] {aka Met}, RAF1 (Raf-1 proto-oncogene, serine/threonine kinase) [NCBI Gene 5894] {aka CMD1NN, CRAF, NS5, Raf-1, c-Raf}, STMN1 (stathmin 1) [NCBI Gene 3925] {aka C1orf215, LAP18, Lag, OP18, PP17, PP19}, APOB (apolipoprotein B) [NCBI Gene 338] {aka FCHL2, FLDB, LDLCQ4, apoB-100, apoB-48}, BAP1 (BRCA1 associated deubiquitinase 1) [NCBI Gene 8314] {aka HUCEP-13, KURIS, TPDS1, UBM2, UCHL2, UVM2}, AFP (alpha fetoprotein) [NCBI Gene 174] {aka AFPD, FETA, HPAFP}, LPAR2 (lysophosphatidic acid receptor 2) [NCBI Gene 9170] {aka EDG-4, EDG4, LPA-2, LPA2}, CTNNB1 (catenin beta 1) [NCBI Gene 1499] {aka CTNNB, EVR7, MRD19, NEDSDV, armadillo}, ACSL4 (acyl-CoA synthetase long chain family member 4) [NCBI Gene 2182] {aka ACS4, FACL4, LACS4, MRX63, MRX68, XLID63}, AXIN1 (axin 1) [NCBI Gene 8312] {aka AXIN, CMDOH, PPP1R49}, ALB (albumin) [NCBI Gene 213] {aka FDAHT, HSA, PRO0883, PRO0903, PRO1341}, ERBB2 (erb-b2 receptor tyrosine kinase 2) [NCBI Gene 2064] {aka CD340, HER-2, HER-2/neu, HER2, MLN 19, MLN-19}, ANGPT2 (angiopoietin 2) [NCBI Gene 285] {aka AGPT2, ANG2, LMPHM10}, HGF (hepatocyte growth factor) [NCBI Gene 3082] {aka DFNB39, F-TCF, HGFB, HPTA, SF}, VEGFA (vascular endothelial growth factor A) [NCBI Gene 7422] {aka L-VEGF, MVCD1, VEGF, VPF}, TP53 (tumor protein p53) [NCBI Gene 7157] {aka BCC7, BMFS5, LFS1, P53, TRP53}, BRAF (B-Raf proto-oncogene, serine/threonine kinase) [NCBI Gene 673] {aka B-RAF1, B-raf, BRAF-1, BRAF1, NS7, RAFB1}, UGT1A9 (UDP glucuronosyltransferase family 1 member A9) [NCBI Gene 54600] {aka HLUGP4, LUGP4, UDPGT, UDPGT 1-9, UGT-1I, UGT1-09}, FAM47A (family with sequence similarity 47 member A) [NCBI Gene 158724], SLC7A11 (solute carrier family 7 member 11) [NCBI Gene 23657] {aka CCBR1, xCT}, STAT3 (signal transducer and activator of transcription 3) [NCBI Gene 6774] {aka ADMIO, ADMIO1, APRF, HIES}
- **Diseases:** chronic (MESH:D002908), necrosis (MESH:D009336), HCC (MESH:D006528), hypertension (MESH:D006973), metastasis (MESH:D009362), cytotoxic (MESH:D064420), diarrhea (MESH:D003967), fatigue (MESH:D005221), hemorrhage (MESH:D006470), metabolic dysfunction (MESH:D008659), extrahepatic (MESH:D001651), cirrhosis (MESH:D005355), liver diseases (MESH:D008107), hepatitis B or C virus infection (MESH:D006509), Cancer (MESH:D009369), hand and foot skin reactions (MESH:D060831)
- **Chemicals:** lipid (MESH:D008055), glutathione (MESH:D005978), reactive oxygen species (MESH:D017382), formalin (MESH:D005557), Cystine (MESH:D003553), polyunsaturated fatty acid (MESH:D005231), H&amp;E (MESH:D006371), regorafenib (MESH:C559147), IMbrave150 (-), atezolizumab (MESH:C000594389), iron (MESH:D007501), bevacizumab (MESH:D000068258), Sorafenib (MESH:D000077157), paraffin (MESH:D010232), lipid hydroperoxides (MESH:D008054)
- **Species:** Homo sapiens (human, species) [taxon 9606]
- **Cell lines:** S2 — Drosophila melanogaster (Fruit fly), Spontaneously immortalized cell line (CVCL_Z232)

## Full text

_Full body text omitted from this summary view._ Fetch the complete paper as Markdown: https://tomesphere.com/paper/PMC12969672/full.md

## Figures

6 figures with captions in the complete paper: https://tomesphere.com/paper/PMC12969672/full.md

## References

50 references — full list in the complete paper: https://tomesphere.com/paper/PMC12969672/full.md

---
Source: https://tomesphere.com/paper/PMC12969672