A k-mer-Based Approach for Phylogenetic Classification of Taxa in Environmental Genomic Data

Abstract

In the age of genome sequencing, whole genome data is readily and frequently generated, leading to a wealth of new information that can be used to advance various fields of research. New approaches, such as alignment-free phylogenetic methods that utilize k-mer-based distance scoring, are becoming increasingly popular given their ability to rapidly generate phylogenetic information from whole genome data. However, these methods have not yet been tested using environmental data, which often tends to be highly fragmented and incomplete. Here we compare the results of one alignment-free approach (which utilizes the D 2 statistic) to traditional multi-gene maximum likelihood trees in three algal groups that have high quality genome data available. In addition, we simulate lower-quality, fragmented genome data using these algae to test method robustness to genome quality and completeness. Finally, we apply the alignment-free approach to environmental metagenome assembled genome data of unclassified Saccharibacteria and Trebouxiophyte algae, and single-cell amplified data from uncultured marine stramenopiles to demonstrate its utility with real datasets. We find that in all instances, the alignment-free method produces phylogenies that are comparable, and often more informative, than those created using the traditional multi-gene approach. The k-mer-based method performs well even when there is significant missing data, that includes marker genes traditionally used for tree reconstruction. Our results demonstrate the value of alignment-free approaches for classifying novel, often cryptic or rare, species, that may not be culturable or are difficult to access using single-cell methods, but fill important gaps in the tree of life.

Estimating Genome-wide Phylogenies Using Probabilistic Topic Modeling

“…Many authors (cf. Edwards et al 2002;Chan et al 2014;Balaban et al 2022;Van Etten et al 2023;Aledo 2022;Dylus et al 2023) declared that k-mers from unaligned data deliver good phylogenetic information, but our analysis shows that k-mers derived from unaligned sequences contain less information than sequences augmented with alignment information. We agree that unaligned data analysis is a great first and quick step to evaluate phylogenetic relationships, but if great accuracy is needed, then aligned sequences will be necessary.…”

“…Alignment-free methods for evolutionary analysis have been reviewed (Vinga and Almeida 2003; Zielezinski et al 2019) and their robustness investigated (Chan et al 2014; Bernard et al 2016). For example, they can be used to derive distances to be summarized into phylogenies (Edwards et al 2002; Chan et al 2014; Balaban et al 2022; Van Etten et al 2023). Alignment-free approaches for phylogeny inference have attracted considerable attention in recent years (Dylus et al 2023; Aledo 2022; Bernard et al 2021; Zielezinski et al 2017).…”

Diverse fates of ancient horizontal gene transfers in extremophilic red algae

“…These algae are also evolutionarily distinct, having diverged shortly after the emergence of the Galdieria clade approximately 1 billion years ago (Figure A20B; Yoon et al, 2017). Furthermore, each isolate represents the earliest divergence within its respective clade (Rossoni et al, 2019; Van Etten, Stephens, & Bhattacharya, 2023). Despite having indistinguishable morphology and nearly identical genome size and features (e.g., few introns, sparse non‐coding DNA, similar GC content; Rossoni et al, 2019), these divergent organisms (recently established as different species; Park et al, 2023) have likely resided on opposite sides of our planet for hundreds of millions of years, possibly since the breakup of the last supercontinent, Pangaea (Correia & Murphy, 2020).…”

Application and Comparison of Machine Learning and Database-Based Methods in Taxonomic Classification of High-Throughput Sequencing Data

“…In the face of intricate and varied genomic data, k-mer taxonomic classification techniques have evolved into an essential instrument for researchers. Their high efficiency and generality have led to their extensive application in diverse settings and fields, ranging from environmental genomics to metagenomics, epidemiology, and origin tracking, to name a few ( Van Etten et al 2023 ). More importantly, k-mer methods do not rely on specific genes or genomic structures, which is particularly beneficial for genomes without detailed annotations or those that are highly heterogeneous.…”