The importance of data transformation in RNA-Seq preprocessing for bladder cancer subtyping

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• RNA-Sequencing (RNA-Seq) is vital for gene expression quantification and cancer molecular subtyping.
• The accuracy of RNA-Seq analyses, particularly for prognosis, depends heavily on data preprocessing steps.
• Variations in preprocessing can affect the reliability and validity of molecular subtype classification.

Purpose of the Study:

• To evaluate the influence of RNA-Sequencing preprocessing methods on molecular subtype classification in bladder cancer.
• To benchmark different aligners, quantifiers, normalization, and transformation techniques.
• To highlight the critical role of preprocessing choices in achieving accurate and consistent bladder cancer subtype classification.

Main Methods:

• Benchmarking of various RNA-Seq aligners and quantifiers.
• Evaluation of different normalization and data transformation methods.
• Assessment of classification performance using centroid-based (consensusMIBC, TCGAclas) and distribution-free (LundTax) algorithms.

Main Results:

• Log-transformation is essential for centroid-based classifiers (consensusMIBC, TCGAclas), with non-log-transformed data yielding poor classification rates.
• Distribution-free algorithms, such as LundTax, demonstrate robustness against variations in RNA-Seq preprocessing.
• LundTax consistently provided superior subtype separation compared to consensusMIBC and TCGAclas, irrespective of preprocessing pipeline.

Conclusions:

• Preprocessing choices critically influence RNA-Seq-based molecular subtype classification in bladder cancer.
• Log-transformation is key for specific classifiers, while distribution-free methods offer greater preprocessing independence.
• Further research is needed to validate the robustness and scalability of these findings for objective subtype accuracy assessment.