Paper 4 & 5: Normalization - bcb420-2025/Keren_Zhang GitHub Wiki

Table of Contents Paper 4: "Selecting between-sample RNA-Seq normalization methods from the perspective of their assumptions" Introduction RNA-Seq and Normalization Fundamentals Common Normalization Methods Assumptions Behind Normalization Methods Impact of Assumptions on Normalization Efficacy Evaluation of Normalization Methods Recommendations and Guidelines Paper 5: "A scaling normalization method for differential expression analysis of RNA-seq data" Introduction Background and Motivation TMM Normalization Method Results and Practical Applications Implications for RNA-seq Analysis Conclusions References

Normalization in RNA-Seq is critical for accurate data analysis, particularly when comparing gene expression across samples. The effectiveness of normalization methods largely depends on the underlying assumptions each method makes about the data. This paper examines various normalization methods and the assumptions they rely on, offering guidance on selecting appropriate methods based on specific experimental conditions.

• Background: RNA-Seq is a powerful tool for studying gene expression under various conditions. However, differences in sequencing depth and other technical variabilities necessitate the use of normalization methods to ensure meaningful comparisons.
• Importance of Assumptions: The assumptions underlying different normalization methods can significantly impact their effectiveness. Misunderstandings or incorrect assumptions can lead to errors in subsequent analyses.

• Gene Expression Quantification: In RNA-Seq, the number of reads mapped to a gene indicates its expression level. Normalization adjusts these read counts to account for factors like sequencing depth, gene length, and GC content.
• Types of Effects: Effects needing normalization are categorized into two types: within-sample (e.g., gene length, GC content) and between-sample (e.g., sequencing depth).

• Total Count and RPKM/FPKM: These methods adjust read counts by total reads per sample or factor in gene length, suitable when the total RNA output per sample is consistent.
• TMM and Quantile Normalization: These methods adjust based on the distribution of read counts, aiming to make the distribution of counts across samples similar.

• Same Total Expression: Some methods assume that the total mRNA expression across conditions is the same. This assumption may not hold if highly expressed genes dominate the read counts.
• Symmetry in Differential Expression: Methods like TMM and quantile normalization assume a balance in the number of up- and down-regulated genes across conditions. Violations of this assumption can skew normalization results.

• Influence of Highly Expressed Genes: The presence of a few highly expressed genes can disproportionately affect total read counts, leading to incorrect normalization if not properly accounted for.
• Global Shifts in Expression: Global increases or decreases in expression across all genes in a sample can lead to incorrect assumptions about constant total mRNA, impacting normalization effectiveness.

• Experimental Validation: The paper highlights the importance of validating normalization methods under controlled conditions to ensure assumptions hold true.
• Case Studies: Examples from studies on organisms like Mus musculus show how different normalization methods perform under varying conditions of gene expression.

• Choice of Normalization Method: The selection of a normalization method should consider the specific biological and technical conditions of the experiment. No single method is universally superior.
• Critical Analysis of Assumptions: Researchers are encouraged to critically analyze the assumptions each normalization method makes about their data to avoid pitfalls in gene expression analysis.

RNA-seq is a powerful tool for studying the transcriptome, providing detailed insights into gene expression. The paper by Robinson and Oshlack addresses the critical role of normalization in RNA-seq data analysis, specifically introducing the Trimmed Mean of M-values (TMM) method for normalization.

• Context: The complexity of the transcriptional architecture and the massive data generated by RNA-seq necessitate effective normalization techniques to identify biologically significant expression changes across different conditions.
• Normalization Need: Traditional methods like RPKM adjust for gene length and sequencing depth, but do not account for the dynamic range of expression levels across samples, potentially skewing differential expression (DE) analysis.

• Challenges with Current Methods: Existing normalization methods, while useful, often fail to account for the variability in gene expression introduced by different RNA populations across samples.
• Advantages of RNA-seq over Microarrays: Unlike microarrays, RNA-seq can identify splicing variants and allele-specific expression, but it still requires robust normalization to accurately measure these features.

• Concept: TMM estimates scale factors by assuming that most genes are not differentially expressed across the conditions studied. It involves calculating the trimmed mean of log expression ratios (M-values) to adjust for differences in RNA production between samples.
• Methodology: The method focuses on scaling the read counts to compensate for compositional differences in RNA samples, effectively reducing the influence of highly expressed genes that could dominate the analysis.

• Comparison with Other Methods: TMM normalization showed improved performance in identifying DE genes compared to traditional methods, which often overestimate the number of DE genes due to normalization issues.
• Case Studies: Applications of TMM to liver versus kidney datasets demonstrated more balanced identification of DE genes, highlighting its effectiveness in practical scenarios.

• Biological Relevance: Accurate normalization is crucial for identifying true biological variations in gene expression, which is essential for downstream analyses like understanding disease mechanisms or developmental stages.
• Technical Considerations: The method is robust across different types of RNA-seq data, including those with high variability in gene expression levels and different sequencing depths.

• Normalization is Essential: The study reinforces the necessity of normalization in RNA-seq data analysis, showing that even advanced sequencing technologies are prone to biases that can mislead biological interpretations.
• Future Directions: The paper suggests further development of normalization methods that can adapt to the increasing complexity and scale of RNA-seq data in various biological contexts.

• Evans, C., Hardin, J., & Stoebel, D. M. (2018). Selecting between-sample RNA-Seq normalization methods from the perspective of their assumptions. Briefings in Bioinformatics, 19(5), 776-792. DOI: 10.1093/bib/bbx008
• Robinson, M. D., & Oshlack, A. (2010). A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biology, 11, R25. https://genomebiology.biomedcentral.com/articles/10.1186/gb-2010-11-3-r25