Seurat function argument values in scRNA-seq data analysis: potential pitfalls and refinements for biological...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Biological data volume is rapidly increasing, necessitating robust processing methods.
• Standard bioinformatics pipelines for scRNA-seq data may oversimplify complex biological signals.
• The biological rationale behind common mathematical transformations is often underexplored.

Purpose of the Study:

• To critically evaluate standard preprocessing, dimensionality reduction, integration, and clustering methods for scRNA-seq data.
• To emphasize the importance of biological context in interpreting results from mathematical data processing.
• To propose an integrated bioinformatics and biology approach for deeper biological insights.

Main Methods:

• Analysis of common scRNA-seq data processing steps: preprocessing (LogNormalize, CLR, RC), dimensionality reduction, integration, and clustering.
• Review of mathematical transformations and their implications for biological data.
• Application of an integrated biology and bioinformatics framework.

Main Results:

• Standard methods like normalization and scaling (LogNormalize, CLR, RC) lack clear biological justification in many contexts.
• Dimensionality reduction may discard biologically relevant minor patterns.
• Current integration and clustering methods require re-evaluation for biological data.

Conclusions:

• Blind application of mathematical methods to scRNA-seq data can lead to flawed biological hypotheses.
• An integrated approach combining biological expertise with bioinformatics is crucial for accurate data interpretation.
• Further research is needed to refine and validate bioinformatics tools for biological discovery.