Predicting chromosome 1p/19q codeletion by RNA expression profile: a comparison of current prediction models

Area of Science:

• Oncology
• Genetics
• Bioinformatics

Background:

• Chromosome 1p/19q codeletion is a critical genetic marker for glioma classification, included in the 2016 WHO guidelines.
• Current detection methods like fluorescent in situ hybridization (FISH) have limitations impacting clinical decisions.
• RNA sequencing presents a potential alternative for determining 1p/19q status.

Purpose of the Study:

• To explore and evaluate computational methods using RNA sequencing data for detecting 1p/19q codeletion status in gliomas.
• To compare the accuracy, sensitivity, and specificity of various RNA sequencing-based tools.

Main Methods:

• Utilized TCGA (n=692) and REMBRANDT (n=222) cohorts with known 1p/19q status for training and validation.
• Assessed five computational tools: TSPairs, GSVA, PAM, Caret, and smoother.
• Compared tools based on prediction accuracy, sensitivity, and specificity.

Main Results:

• In the TCGA cohort, GSVA achieved 98.4% accuracy, while smoother reached 97.8%.
• In the REMBRANDT cohort, the smoother method demonstrated the highest accuracy at 98.6%.
• Both GSVA and smoother showed high sensitivity and specificity in predicting 1p/19q status.

Conclusions:

• The smoother method was identified as the most stable and accurate RNA sequencing-based approach for predicting 1p/19q codeletion.
• This computational method shows promise as a viable alternative for clinical application in glioma management.