Radiogenomic method combining DNA methylation profiles and magnetic resonance imaging radiomics predicts patient prognosis in skull base chordoma.

Abstract

Background: Chordoma is a rare malignant bone tumor exhibiting poor survival and prognosis. Hence, it is crucial to develop a convenient and effective prognostic classification method for the rehabilitation and management of patients with chordoma. In this study, we combined DNA methylation profiles and magnetic resonance imaging (MRI) images to generate a radiogenomic signature to assess its effectiveness for prognosis classification in patients with skull base chordoma.

Results: DNA methylation profiles from chordoma tissue samples of 40 patients were factorized into eight DNA methylation signatures. Among them, Signature 4 was identified as the prognosis-specific signature. Based on the Signature 4 loading values, the patients were categorized into low-signature (LLG) and high-signature (HLG) loading groups. HLG patients had higher progression-free survival times than LLG patients. Combined analysis with external single-cell RNA-seq data revealed higher tumor cell proportions and lower natural killer cell proportions in the HLG than in the LLG. Additionally, 2,553 radiomic features were extracted from T1, T2, and enhanced T1 MRI images of the patients, and a radiogenomic signature comprising 14 radiomic features was developed. In a validation cohort of 122 patients, the radiogenomic signature successfully distinguished between the two groups (P = 0.027). Furthermore, the existence of Signature 4 was confirmed in an additional dataset of 14 patients.

Conclusion: We developed a prognostic radiogenomic signature using a radiogenomic classification method, which leverages MRI images to extract features that reflect the DNA methylation signature associated with prognosis, enabling the stratification of patients based on their prognostic risk. This method offers the advantages of being noninvasive and convenient.