Estimating the k-mer Coverage Frequencies in Genomic Datasets: A Comparative Assessment of the State-of-the-art

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Estimating k-mer abundance and unique k-mers is crucial for genome analysis.
• Applications include genome size prediction, assembly preprocessing, and error rate estimation.
• Various cardinality estimation methods for sequencing data exist.

Purpose of the Study:

• To comparatively assess k-mer frequency estimation programs: ntCard, KmerGenie, KmerStream, and Khmer.
• To evaluate their relative merits and demerits for bioinformatics applications.

Main Methods:

• Rigorous experimental analysis of miscounts/error rates for various k values.
• Evaluation of runtime, scalability, memory, CPU utilization, and parallelism.
• Comparative assessment of ntCard, KmerGenie, KmerStream, and Khmer.

Main Results:

• ntCard demonstrated superior accuracy in estimating F0, f1, and full k-mer abundance histograms.
• ntCard was the fastest tool evaluated.
• ntCard exhibited higher memory requirements compared to KmerGenie.

Conclusions:

• Provides a roadmap for users to select appropriate streaming algorithms for k-mer frequency estimation.
• Highlights areas for future research, including open questions and technique combinations.