ntCard

ntCard estimates k-mer frequencies and constructs k-mer frequency histograms from sequencing datasets to support genome size estimation, sequencing error assessment, and optimization of downstream k-mer–based analyses.

Key Features:

• Streaming Algorithm: ntCard uses a streaming algorithm to process large sequencing datasets without loading the entire dataset into memory.
• Hashing Technique: It uses the ntHash algorithm to compute hash values for k-mers to enable rapid k-mer frequency estimation.
• Sampling and Statistical Modeling: ntCard samples hash values to build a reduced-representation multiplicity table of sampled k-mers and applies a statistical model to reconstruct the population k-mer frequency distribution.

Scientific Applications:

• K-mer frequency estimation: ntCard provides rapid estimation of k-mer coverage frequencies for genome size prediction, sequencing error rate assessment, and tuning of parameters for k-mer–based tools.
• Benchmarking on large datasets: ntCard has been benchmarked on datasets ranging from 480 GB to 2.4 TB, including whole-genome shotgun sequencing of human and white spruce, achieving more than 15× faster k-mer frequency estimates than other algorithms with similar memory usage and higher accuracy.

Methodology:

ntCard computes ntHash hash values for streamed sequences, samples those hash values to construct a multiplicity table representing the sample distribution of k-mers, and applies a statistical model to infer the population k-mer frequency distribution.