Shape analysis of high-throughput transcriptomics experiment data.

Area of Science:

• Genomics
• Statistical Bioinformatics
• Computational Biology

Background:

• High-throughput transcriptome technologies generate large datasets requiring robust statistical analysis.
• Current methods often assume Gaussian distribution, which is difficult to verify, especially with small sample sizes.
• Assessing distributional assumptions is crucial for reliable interpretation of transcriptomic data.

Purpose of the Study:

• To introduce L-moments statistics for exploratory data analysis in high-throughput transcriptomics.
• To assess distributional assumptions and perform hypothesis testing using L-moments.
• To develop an algorithm for identifying genes with non-typical expression distributions.

Main Methods:

• Utilized L-moments statistics for exploratory data analysis and assumption assessment.
• Employed L-moments ratios to evaluate skewness and kurtosis of transcriptomic data.
• Developed an algorithm for identifying outlier gene expression distributions.
• Applied the framework to RNA-sequencing (RNA-seq) data.

Main Results:

• L-moments provide a robust framework for assessing distributional shapes (skewness, kurtosis) in transcriptomic data.
• An algorithm was proposed to effectively identify genes with distributions deviating from the norm.
• The L-moments approach demonstrated robustness for differential expression analysis in RNA-seq data.
• Methods utilizing L-moments as weights for t-tests proved reliable.

Conclusions:

• L-moments statistics offer a valuable tool for robust analysis of high-throughput transcriptomic data.
• This framework enhances the assessment of distributional assumptions, crucial for data integrity.
• The proposed methods are particularly useful for identifying atypical gene expression patterns and ensuring reliable differential expression analysis.