Efficient techniques for k-mer counting

Abstract

A large number of bioinformatics applications require counting of k-length substrings in genetically important long strings. K-mer counting generates the frequencies of each k-length substring in genome sequences. Genome assembly, repeat detection, multiple sequence alignment, error detection, and many other related applications use k-mer counting as a building block. Many approaches are already available to address the problem. Some of them are time efficient, and some of them are memory efficient. Most of the current solutions use multi-threading to utilize available cores of a machine. A few efficient disk-based algorithms have been devised to reduce required memory. We analyze all available algorithms, and time and memory requirements of those implementations. We improve time consumption by devising a novel algorithm to this problem. Our results show that this new algorithm outperforms previous best-known algorithms.