Biological Intuition in Alignment-Free Methods: Response to Posada

Alignment-Free Analysis of Whole-Genome Sequences From Symbiodiniaceae Reveals Different Phylogenetic Signals in Distinct Regions

Dinoflagellates of the family Symbiodiniaceae are predominantly essential symbionts of corals and other marine organisms. Recent research reveals extensive genome sequence divergence among Symbiodiniaceae taxa and high phylogenetic diversity hidden behind subtly different cell morphologies. Using an alignment-free phylogenetic approach based on sub-sequences of fixed length k (i.e. k-mers), we assessed the phylogenetic signal among whole-genome sequences from 16 Symbiodiniaceae taxa (including the genera of Symbiodinium, Breviolum, Cladocopium, Durusdinium and Fugacium) and two strains of Polarella glacialis as outgroup. Based on phylogenetic trees inferred from k-mers in distinct genomic regions (i.e. repeat-masked genome sequences, protein-coding sequences, introns and repeats) and in protein sequences, the phylogenetic signal associated with protein-coding DNA and the encoded amino acids is largely consistent with the Symbiodiniaceae phylogeny based on established markers, such as large subunit rRNA. The other genome sequences (introns and repeats) exhibit distinct phylogenetic signals, supporting the expected differential evolutionary pressure acting on these regions. Our analysis of conserved core k-mers revealed the prevalence of conserved k-mers (>95% core 23-mers among all 18 genomes) in annotated repeats and non-genic regions of the genomes. We observed 180 distinct repeat types that are significantly enriched in genomes of the symbiotic versus free-living Symbiodinium taxa, suggesting an enhanced activity of transposable elements linked to the symbiotic lifestyle. We provide evidence that representation of alignment-free phylogenies as dynamic networks enhances the ability to generate new hypotheses about genome evolution in Symbiodiniaceae. These results demonstrate the potential of alignment-free phylogenetic methods as a scalable approach for inferring comprehensive, unbiased whole-genome phylogenies of dinoflagellates and more broadly of microbial eukaryotes.

Success of alignment-free oligonucleotide (k-mer) analysis confirms relative importance of genomes not genes in speciation and phylogeny

The utility of DNA sequence substrings (k-mers) in alignment-free phylogenetic classification, including that of bacteria and viruses, is increasingly recognized. However, its biological basis eludes many 21st century practitioners. A path from the 19th century recognition of the informational basis of heredity to the modern era can be discerned. Crick’s DNA ‘unpairing postulate’ predicted that recombinational pairing of homologous DNAs during meiosis would be mediated by short k-mers in the loops of stem-loop structures extruded from classical duplex helices. The complementary ‘kissing’ duplex loops – like tRNA anticodon–codon k-mer duplexes – would seed a more extensive pairing that would then extend until limited by lack of homology or other factors. Indeed, this became the principle behind alignment-based methods that assessed similarity by degree of DNA–DNA reassociation in vitro. These are now seen as less sensitive than alignment-free methods that are closely consistent, both theoretically and mechanistically, with chromosomal anti-recombination models for the initiation of divergence into new species. The analytical power of k-mer differences supports the theses that evolutionary advance sometimes serves the needs of nucleic acids (genomes) rather than proteins (genes), and that such differences can play a role in early speciation events.

k-mer Similarity, Networks of Microbial Genomes, and Taxonomic Rank

Microbial genomes have been shaped by parent-to-offspring (vertical) descent and lateral genetic transfer. These processes can be distinguished by alignment-based inference and comparison of phylogenetic trees for individual gene families, but this approach is not scalable to whole-genome sequences, and a tree-like structure does not adequately capture how these processes impact microbial physiology. Here we adopted alignment-free approaches based on k-mer statistics to infer phylogenomic networks involving 2,783 completely sequenced bacterial and archaeal genomes and compared the contributions of rRNA, protein-coding, and plasmid sequences to these networks. Our results show that the phylogenomic signal arising from ribosomal RNAs is strong and extends broadly across all taxa, whereas that from plasmids is strong but restricted to closely related groups, particularly Proteobacteria. However, the signal from the other chromosomal regions is restricted in breadth. We show that mean k-mer similarity can correlate with taxonomic rank. We also link the implicated k-mers to genome annotation (thus, functions) and define core k-mers (thus, core functions) in specific phyletic groups. Highly conserved functions in most phyla include amino acid metabolism and transport as well as energy production and conversion. Intracellular trafficking and secretion are the most prominent core functions among Spirochaetes, whereas energy production and conversion are not highly conserved among the largely parasitic or commensal Tenericutes. These observations suggest that differential conservation of functions relates to niche specialization and evolutionary diversification of microbes. Our results demonstrate that k-mer approaches can be used to efficiently identify phylogenomic signals and conserved core functions at the multigenome scale. IMPORTANCE Genome evolution of microbes involves parent-to-offspring descent, and lateral genetic transfer that convolutes the phylogenomic signal. This study investigated phylogenomic signals among thousands of microbial genomes based on short subsequences without using multiple-sequence alignment. The signal from ribosomal RNAs is strong across all taxa, and the signal of plasmids is strong only in closely related groups, particularly Proteobacteria. However, the signal from other chromosomal regions (∼99% of the genomes) is remarkably restricted in breadth. The similarity of subsequences is found to correlate with taxonomic rank and informs on conserved and differential core functions relative to niche specialization and evolutionary diversification of microbes. These results provide a comprehensive, alignment-free view of microbial genome evolution as a network, beyond a tree-like structure.

Necessary relations for nucleotide frequencies

Genome composition analysis of di-, tri- and tetra-nucleotide frequencies is known to be evolutionarily informative, and useful in metagenomic studies, where binning of raw sequence data is often an important first step. Patterns appearing in genome composition analysis may be due to evolutionary processes or purely mathematical relations. For example, the total number of dinucleotides in a sequence is equal to the sum of the individual totals of the sixteen types of dinucleotide, and this is entirely independent of any assumptions made regarding mutation or selection, or indeed any physical or chemical process. Before any statistical analysis can be attempted, a knowledge of all necessary mathematical relations is required. I show that 25% of di-, tri- and tetra-nucleotide frequencies can be written as simple sums and differences of the remainder. The vast majority of organisms have circular genomes, for which these relations are exact and necessary. In the case of linear molecules, the absolute error is very nearly zero, and does not grow with contiguous sequence length. As a result of the new, necessary relations presented here, the foundations of the statistical analysis of di-, tri- and tetra-nucleotide frequencies, and k-mer analysis in general, need to be revisited.

Inferring phylogenies of evolving sequences without multiple sequence alignment

Alignment-free methods, in which shared properties of sub-sequences (e.g. identity or match length) are extracted and used to compute a distance matrix, have recently been explored for phylogenetic inference. However, the scalability and robustness of these methods to key evolutionary processes remain to be investigated. Here, using simulated sequence sets of various sizes in both nucleotides and amino acids, we systematically assess the accuracy of phylogenetic inference using an alignment-free approach, based on D₂ statistics, under different evolutionary scenarios. We find that compared to a multiple sequence alignment approach, D₂ methods are more robust against among-site rate heterogeneity, compositional biases, genetic rearrangements and insertions/deletions, but are more sensitive to recent sequence divergence and sequence truncation. Across diverse empirical datasets, the alignment-free methods perform well for sequences sharing low divergence, at greater computation speed. Our findings provide strong evidence for the scalability and the potential use of alignment-free methods in large-scale phylogenomics.

Molecular phylogenetics before sequences: oligonucleotide catalogs as k-mer spectra

From 1971 to 1985, Carl Woese and colleagues generated oligonucleotide catalogs of 16S/18S rRNAs from more than 400 organisms. Using these incomplete and imperfect data, Carl and his colleagues developed unprecedented insights into the structure, function, and evolution of the large RNA components of the translational apparatus. They recognized a third domain of life, revealed the phylogenetic backbone of bacteria (and its limitations), delineated taxa, and explored the tempo and mode of microbial evolution. For these discoveries to have stood the test of time, oligonucleotide catalogs must carry significant phylogenetic signal; they thus bear re-examination in view of the current interest in alignment-free phylogenetics based on k-mers. Here we consider the aims, successes, and limitations of this early phase of molecular phylogenetics. We computationally generate oligonucleotide sets (e-catalogs) from 16S/18S rRNA sequences, calculate pairwise distances between them based on D₂ statistics, compute distance trees, and compare their performance against alignment-based and k-mer trees. Although the catalogs themselves were superseded by full-length sequences, this stage in the development of computational molecular biology remains instructive for us today.