K Mer Counter Calculator

A k-mer is any contiguous subsequence of length k drawn from a nucleotide string. Sliding a window one base at a time, a sequence of length L yields exactly L - k + 1 k-mers, while the four-letter DNA alphabet permits 4 to the power k distinct words: 64 for k = 3, about 4.2 million for k = 11, and roughly 4.4 trillion for k = 21. Comparing the observed unique k-mers against that ceiling gives a linguistic complexity fraction, which falls near zero for a poly-A tract or a tandem repeat and rises toward 1 only for sequences long and random enough to sample the space. Because DNA is double-stranded and biological features have no preferred strand, most production tools count canonical k-mers, collapsing each word with its reverse complement so that ATG and CAT are tallied as one. K-mer size is the decisive tuning parameter. Short k values of 15 to 21 tolerate low coverage and base-call error but generate spurious overlaps in repetitive genomes; long k values of 31 to 127 resolve repeats but demand deeper coverage and far more memory. Bacterial assemblies commonly use k = 21 to 31, human assemblies 51 to 101, and assemblers such as SPAdes run several k values and merge the resulting graphs. K-mer histograms also support alignment-free genome size estimation, since total k-mers divided by the coverage peak approximates genome size, with heterozygosity and repeat content readable from the curve shape. Two caveats matter: sequencing errors produce a large tail of unique low-count k-mers, and k should be odd in canonical counting so no word can be its own reverse complement.

Counting fixed-length words predates genomics. It descends from the n-gram statistics Claude Shannon used in his 1948 estimate of the entropy of written English and from the substring indexing methods of early computer science. Its genomic career began with assembly theory: Pavel Pevzner, Haixu Tang and Michael Waterman set out the de Bruijn graph formulation of fragment assembly in a 2001 PNAS paper, in which reads are shattered into overlapping k-mers and reconstruction becomes the search for an Eulerian path through a graph whose edges are k-mers and whose nodes are their k-1 prefixes and suffixes. That reframing suited the short, high-volume reads that arrived with pyrosequencing and then Illumina sequencing later in the decade. Short-read assemblers built directly on the idea: Velvet from Daniel Zerbino and Ewan Birney at the European Bioinformatics Institute in 2008, ABySS and SOAPdenovo shortly after, and SPAdes from the St Petersburg group in 2012, which popularised multi-k assembly. Counting k-mers at genome scale became its own engineering problem, answered by Jellyfish from Guillaume Marcais and Carl Kingsford in 2011 using a lock-free hash table, and later by disk-partitioning counters such as KMC. Alignment-free applications followed and now dominate: GenomeScope profiles genomes straight from k-mer histograms, Kraken classifies metagenomic reads by k-mer content, and Mash compares whole genomes through MinHash sketches of their k-mer sets.