Local sequence and sequencing depth dependent accuracy of RNA-seq reads

Area of Science:

• Genomics
• Bioinformatics
• Statistical Modeling

Background:

• RNA sequencing (RNA-seq) technology is prone to biases and spurious effects.
• These issues lead to non-uniform sequencing read counts across gene base positions.
• Accurate modeling requires addressing base-level non-uniformity and leveraging read count properties for precise mean and variance estimation.

Purpose of the Study:

• To investigate the impact of various factors on RNA-seq accuracy.
• To develop improved RNA-seq read modeling strategies.
• To compare different modeling approaches for RNA-seq data.

Main Methods:

• Analysis of base-level overdispersion rates in RNA-seq data.
• Investigating the relationship between sequencing depth and overdispersion.
• Assessing the influence of local sequences on overdispersion.
• Developing a beta-binomial model with a dynamic overdispersion rate.

Main Results:

• Overdispersion rate decreases with increased sequencing depth at the base level.
• Local sequence significantly influences overdispersion, but this effect diminishes after accounting for sequencing depth.
• A beta-binomial model with a dynamic overdispersion rate was developed for modeling RNA-seq read counts.

Conclusions:

• The study offers insights into position-level overdispersion in RNA-seq.
• Key factors influencing overdispersion include sequencing depth, local sequence, and preparation protocols.
• Findings can enhance RNA-seq quality control and statistical downstream analysis methods.