Estimating repeat spectra and genome length from low-coverage genome skims with RESPECT

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome sequencing costs are decreasing faster than assembly and finishing.
• Lightly sampled genomes (genome-skims) offer potential for genomic ecology.
• K-mer based approaches show promise for species identification and phylogenetic placement.

Purpose of the Study:

• To address limitations in estimating genomic parameters from k-mer spectra.
• To develop a robust method for estimating genome length, coverage, and repeat structure from low-coverage data.
• To improve the accuracy of genomic parameter estimation in the presence of contaminants.

Main Methods:

• Theoretical and empirical analysis of k-mer spectra limitations.
• Development of a novel constrained optimization approach: Spline Linear Programming.
• Empirical learning of constraints for the optimization method.
• Application of the REPeat SPECTra Estimation (RESPECT) method to simulated and real sequencing data.

Main Results:

• Identified fundamental limitations in estimating k-mer spectra due to ill-conditioned systems.
• RESPECT achieved 2.2% error in length estimation on simulated 1X coverage data, a significant improvement over 27% error from previous methods.
• RESPECT demonstrated a median error of 4% for length estimates in shotgun sequenced samples with contaminants, compared to 80% for other methods.

Conclusions:

• Low-pass genomic sequencing can provide reliable estimates of genome length and repeat content.
• The developed constrained optimization approach overcomes limitations in k-mer spectral analysis.
• The RESPECT software offers a significant advancement for genomic parameter estimation, particularly in challenging datasets.