Optimizing Sample Size for Supervised Machine Learning with Bulk Transcriptomic Sequencing: A Learning Curve...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate sample classification using transcriptomics data is vital for personalized medicine.
• Current sample size calculation methods may not be suitable for supervised machine learning (ML) classification.
• A methodological gap exists in determining adequate sample size for ML-based transcriptomics analysis.

Purpose of the Study:

• To develop and evaluate a novel computational approach for establishing the power-versus-sample-size relationship in transcriptomics studies.
• To address the limitations of existing methods for sample size determination in the context of ML classification.
• To facilitate the use of ML in transcriptomics for personalized medicine.

Main Methods:

• A novel computational approach employing data augmentation and fitting a learning curve to establish the power-versus-sample-size relationship.
• Comprehensive performance evaluation using microRNA and RNA sequencing data.
• Consideration of diverse data characteristics and algorithm configurations.

Main Results:

• The developed approach effectively establishes the power-versus-sample-size relationship for transcriptomics data.
• Performance was validated across various data types (miRNA, RNA-seq) and ML algorithms.
• The method provides a robust framework for sample size estimation in ML-driven transcriptomics.

Conclusions:

• The novel computational approach bridges a critical methodological gap in sample size determination for ML-based transcriptomics.
• This method enhances statistical power and optimizes resource allocation in transcriptomics studies.
• Availability of code on GitHub promotes accessibility, reproducibility, and accelerates the translation of transcriptomics findings into clinical applications for personalized treatment.