Estimation of evolutionary parameters using short, random and partial sequences from mixed samples of anonymous...

Area of Science:

• Evolutionary genetics
• Bioinformatics
• Computational biology

Background:

• Next-generation sequencing (NGS) is widely used but poses challenges for evolutionary biologists estimating genetic parameters from mixed, unlabelled DNA samples.
• Reconstructing full-length haplotypes from short-read sequences can be unreliable, complicating parameter estimation.
• Existing methods struggle with large datasets of short-read sequences from anonymous DNA.

Purpose of the Study:

• To develop novel computational methods for inferring evolutionary genetic parameters from NGS data.
• To address the challenges of analyzing mixed, unlabelled DNA samples with short-read sequences.
• To provide accurate and computationally efficient tools for evolutionary genetic analysis.

Main Methods:

• Least Squares Estimation (LS): A novel approach using nucleotide frequencies at each site without full alignment or phylogeny reconstruction.
• Approximate Bayesian Computation Markov chain Monte Carlo (ABC-MCMC): A novel Bayesian approach using nucleotide frequencies for parameter estimation.
• Simulations were used to evaluate the performance of LS and ABC-MCMC against established methods like BEAST.

Main Results:

• LS method showed poor performance, with bootstrap 95% Confidence Intervals (CIs) often under- or over-estimating true parameter values.
• ABC-MCMC 95% Highest Posterior Density (HPD) intervals demonstrated performance comparable to BEAST, a method using full-length sequences.
• ABC-MCMC HPDs were larger than BEAST's due to information loss from using only sitewise nucleotide frequencies.

Conclusions:

• LS is not recommended as a standalone method for evolutionary parameter estimation.
• ABC-MCMC provides comparable parameter estimates to BEAST but with larger HPDs, making it suitable for large NGS datasets.
• ABC-MCMC offers a significant advantage in computational efficiency, scaling linearly with the number of short-read sequences and independent of the number of full-length sequences.