Molecular Feature-Based Classification of Retroperitoneal Liposarcoma: A Prospective Cohort Study

Reviewer #1 (Public review):

Summary:

In this study, Xiao et al. classified retroperitoneal liposarcoma (RPLS) patients into two subgroups based on whole transcriptome sequencing of 88 patients. The G1 group was characterized by active metabolism, while the G2 group exhibited high scores in cell cycle regulation and DNA damage repair. The G2 group also displayed more aggressive molecular features and had worse clinical outcomes compared to G1. Using a machine learning model, the authors simplified the classification system, identifying LEP and PTTG1 as the key molecular markers distinguishing the two RPLS subgroups. Finally, they validated these markers in a larger cohort of 241 RPLS patients using immunohistochemistry. Overall, the manuscript is clear and well-organized, with its significance rooted in the large sample size and the development of a classification method.

Weakness:

(1) While the authors suggest that LEP and PTTG1 serve as molecular markers for the two RPLS groups, the process through which these genes were selected remains unclear. The authors should provide a detailed explanation of the selection process.

(2) To ensure the broader applicability of LEP and PTTG1 as classification markers, the authors should validate their findings in one or two external datasets.

(3) Since molecular subtyping is often used to guide personalized treatment strategies, it is recommended that the authors evaluate therapeutic responses in the two distinct groups. Additionally, they should validate these predictions using cell lines or primary cells.

Reviewer #2 (Public review):

Surgical resection remains the most effective treatment for retroperitoneal liposarcoma. However, postoperative recurrence is very common and is considered the main cause of disease-related death. Considering the importance and effectiveness of precision medicine, the identification of molecular characteristics is particularly important for the prognosis assessment and individualized treatment of RPLS. In this work, the authors described the gene expression map of RPLS and illustrated an innovative strategy of molecular classification. Through the pathway enrichment of differentially expressed genes, characteristic abnormal biological processes were identified, and RPLS patients were simply categorized based on the two major abnormal biological processes. Subsequently, the classification strategy was further simplified through nonnegative matrix factorization. The authors finally narrowed the classification indicators to two characteristic molecules LEP and PTTG1, and constructed novel molecular prognosis models that presented obviously a great area under the curve. A relatively interpretable logistic regression model was selected to obtain the risk scoring formula, and its clinical relevance and prognostic evaluation efficiency were verified by immunohistochemistry. Recently, prognostic model construction has been a hot topic in the field of oncology. The interesting point of this study is that it effectively screened characteristic molecules and practically simplified the typing strategy on the basis of ensuring high matching clinical relevance. Overall, the study is well-designed and will serve as a valuable resource for RPLS research.