Pareto Optimal Compression of Genomic Dictionaries, with or without Random Access in Main Memory
Motivation: A Genomic Dictionary, i.e., the set of the k-mers appearing in a genome, is a fundamental source of genomic information: its collection is the first step in strategic computational methods ranging from assembly to sequence comparison and phylogeny. Unfortunately, it is costly to store. This motivates some recent studies regarding the compression of those k-mer sets. However, such an area does not have the maturity of genomic compression, lacking an homogeneous and methodologically sound experimental foundation that allows to fairly compare the relative merits of the available solutions, and that takes into account also the rich choices of compression methods that can be used. Results: We provide such a foundation here, supporting it with an extensive set of experiments that use reference datasets and a carefully selected set of representative data compressors. Our results highlight the spectrum of compressor choices one has in terms of Pareto Optimality of compression vs. post-processing, this latter being important when the Dictionary needs to be decompressed many times. In addition to the useful indications, not available elsewhere, that this study offers to the researchers interested in storing k-mer dictionaries in compressed form, a software system that can be readily used to explore the Pareto Optimal solutions available r a given Dictionary is also provided. Availability: The software system is available at https://github.com/GenGrim76/Pareto-Optimal-GDC, together with user manuals and installation instructions. Contact: raffaele.giancarlo@unipa.it Supplementary information: Additional data are available in the Supplementary Material.
READ FULL TEXT