Confidence intervals for Markov chain transition probabilities based on next generation sequencing reads data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Markov chains (MC) are established for molecular sequence modeling.
• Traditional methods for estimating MC transition probabilities and confidence intervals rely on long sequence data.
• Next-generation sequencing (NGS) generates short reads, creating new data characteristics where classical asymptotic distribution theory is insufficient.

Purpose of the Study:

• To develop and present asymptotic distributions for MC statistics based on NGS data.
• To establish theoretical confidence intervals for MC transition probabilities using NGS short reads.
• To provide a robust statistical tool for analyzing NGS data.

Main Methods:

• Derivation of asymptotic distributions for MC statistics tailored to NGS data.
• Scaling statistics by effective coverage to align with distributions from long sequences.
• Application of derived properties to construct theoretical confidence regions for transition probabilities.

Main Results:

• The asymptotic distributions for MC statistics based on NGS data were successfully presented.
• Theoretical confidence intervals for MC transition probabilities were derived and validated.
• The derived confidence intervals demonstrated high accuracy when compared to simulated and real datasets, and the parametric bootstrap method.

Conclusions:

• The developed asymptotic distributions and theoretical confidence intervals are highly accurate for NGS data.
• This provides a powerful and reliable tool for statistical analysis of next-generation sequencing data.
• The findings bridge the gap in statistical methodology for MC modeling with short-read sequencing data.