Identification and validation of soft tissue sarcoma-specific transcriptomic model for predicting radioresistance.

Abstract

Purpose: We aimed to identify the transcriptomic signatures of soft tissue sarcoma (STS) related to radioresistance and establish a model to predict radioresistance.

Materials and methods: Nine STS cell lines were cultured. Adenosine triphosphate-based viability was determined 5 days after irradiation with 8 Gy of X-rays in a single fraction. Radiosensitive and radioresistant groups were stratified according to the survival rates. Whole transcriptomic sequencing analysis was performed and differentially expressed genes (DEGs) were identified between the radiosensitive and radioresistant groups. For model generation, a cohort of 59 patients with sarcomas from The Cancer Genome Atlas (TCGA) was used. DEGs of the responder and non-responder groups according to the radiotherapy-best response were identified. The overlapping DEGs between those from TCGA data and the STS cell line were subjected to linear regression to develop a formula, namely the STS-specific radioresistance index (STS-RRI), and its performance was compared with that of the previously established radiosensitivity index (RSI).

Results: We selected thirteen overlapping DEGs and established STS-RRI using seven of them: STS-RRI = 1.5185 × MYO16-0.01575 × MYH11 + 3.900375 × KCTD16 + 0.105375 × SYNPO2-0.777375 × MYPN-0.849875 × PCSK6-0.700125 × LTK + 39.4635. Delong's test revealed that the STS-RRI performed better at stratifying responder and non-responder in TCGA cohort than the RSI (p = .002). The progression-free survival curves of the TCGA cohort were significantly discriminated by STS-RRI (p = .013) but not by RSI (p = .241).

Conclusion: We developed the STS-RRI to predict the radioresistance of patients with STS in the TCGA dataset, showing a higher performance than RSI.