DOI: 10.1200/oa-26-00017 ISSN: 2994-9750

GemSarc: A Novel Predictive Gene Signature for Gemcitabine Efficacy in Advanced Bone and Soft Tissue Sarcomas

Ankur Sheel, Erin Fetzer, Yue Yu, Swati Satturwar, Raven Vella, James L. Chen, Valerie P. Grignol, Joal D. Beane, Marium Husain, Gabriel Tinoco, David J. Konieczkowski, Karl E. Haglund, Carl D. Quinion, John H. Alexander, Joel L. Mayerson, Hans Iwenofu, Lianbo Yu, Sameek Roychowdhury, Raphael E. Pollock, David A. Liebner
Show PDF Cite

Purpose

There are no validated predictive biomarkers for chemotherapy response in advanced soft-tissue sarcoma. We sought to develop and validate an RNA-based gene signature predictive of gemcitabine response.

Methods

Patients treated with gemcitabine-based chemotherapy were identified from The Ohio State University (OSU) Sarcoma Registry (ClinicalTrials.gov identifier: NCT02677961 ). Responders were defined as patients with a time to gemcitabine discontinuation greater than the cohort median, based on RECIST-defined progression. Gene expression from Tempus RNA sequencing data was correlated with gemcitabine half maximal inhibitory concentration from sarcoma cell lines to identify predictive transcripts. Differentially expressed genes between responders and nonresponders were intersected, and a random forest regression model was trained to generate the GemSarc score. The model was externally validated using Oncology Research Information Exchange Network (ORIEN) and Total Cancer Care (TCC) data sets.

Results

Thirty-one patients met inclusion criteria. Unsupervised clustering revealed three transcriptomic subgroups, with prolonged duration of therapy in clusters enriched for DNA replication and repair pathways. The GemSarc model identified 19 genes that predicted gemcitabine sensitivity. High GemSarc scores were associated with improved clinical benefit in the OSU discovery cohort (hazard ratio [HR], 0.23 [95% CI, 0.08 to 0.66]), with consistent validation in two independent cohorts (ORIEN: HR, 0.30; TCC: HR, 0.17). The model also demonstrated strong discriminatory performance, with AUCs of 0.88 and 0.79 in the ORIEN and TCC cohorts, respectively. GemSarc did not predict overall survival or progression-free survival in patients with sarcoma from The Cancer Genome Atlas data. GemSarc also failed to identify responders to other cytotoxic agents (AUC <0.63).

Conclusion

GemSarc is a validated RNA-based predictive biomarker that identifies patients with advanced sarcomas likely to benefit from gemcitabine-based therapy. Its implementation may optimize patient selection and guide biomarker-driven clinical trials in sarcoma.

More from our Archive