Study design requirements for RNA sequencing-based breast cancer diagnostics

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Molecular subtyping of breast cancer offers superior prognostication over routine biomarkers.
• Current clinical practice lacks widespread implementation of molecular subtyping.
• Translational success hinges on accurate subtype prediction models with high sensitivity and specificity.

Purpose of the Study:

• To evaluate the impact of sample size and RNA sequencing read count on breast cancer molecular subtyping accuracy.
• To inform the rational design of translational studies for sequencing-based cancer diagnostics.

Main Methods:

• Subsampling techniques were employed to assess the influence of training sample size and RNA sequencing read depth.
• Classification accuracy for molecular subtypes and routine biomarkers was evaluated using both unsupervised and supervised models.
• The study analyzed the relationship between sample size, read count, and prediction accuracy.

Main Results:

• Molecular subtype classification accuracy improved with sample size, reaching 0.93 at N=750, with marginal gains beyond N=350 (0.92 accuracy).
• Prediction accuracy for routine biomarkers, Estrogen Receptor (ER) and Her2, reached 0.94 and 0.92 respectively at N=200.
• Subtype classification accuracy increased with RNA sequencing library size, plateauing at 5 million reads.

Conclusions:

• Adequate sample size and RNA sequencing depth are critical for developing accurate molecular subtyping models.
• Well-designed translational studies are essential for implementing sequencing-based diagnostics in clinical settings.
• This research provides crucial data for optimizing study design in precision oncology.