Fast and Accurate Distance-based Phylogenetic Placement using Divide and Conquer

Abstract

Phylogenetic placement of query samples on an existing phylogeny is increasingly used in microbiome analyses and other biological studies. As the size of available reference trees used in microbiome analyses continues to grow, there is a growing need for methods that place sequences on ultra-large trees with high accuracy.In this paper, we build on our previous work and introduce APPLES-2, a distance-based phylogenetic placement method. In extensive studies, we show that APPLES-2 is more accurate and much more scalable than APPLES and even some of the leading maximum likelihood methods. This scalability is owed to a divide-and-conquer technique that limits distance calculation and phylogenetic placement to parts of the tree most relevant to each query. The increased scalability and accuracy enables us to study the effectiveness of APPLES-2 for placing microbial genomes on a dataset of 10,575 microbial species using subsets of 381 marker genes. APPLES-2 has very high accuracy in this setting, placing 97% of query genomes within three branches of the optimal position in the species tree using 50 marker genes. As a proof of concept, we also tested APPLES-2 for placing scaffolds produced by metagenomic assembly. It was able to place 3318 scaffolds on the backbone tree with 10575 taxa in less than 3 hours. The accuracy of these placement was high as long as a scaffold included tens of marker genes.

SCAMPP+FastTree: Improving Scalability for Likelihood-based Phylogenetic Placement

“…We use the published FastTree-2 results (tree and numeric parameters estimated under Minimum Evolution) on the nt78 datasets [21]. The scripts to calculate delta error was adapted from a published script [1], and borrow utilities from newick_utils. Further details are provided in the published scripts here.…”

“…The most accurate of the current phylogenetic placement methods operate using maximum likelihood, and so require the query sequences to already be aligned to the backbone sequences; examples of such methods include pplacer [9] and EPA-ng [3]. Other types of methods have been developed for phylogenetic placement that are not based on maximum likelihood; a recent example is APPLES-2 [1], which uses branch lengths and distances computed between the query sequence and the leaves in the backbone tree to place each query sequence. APPLES-2 is much faster than pplacer and EPA-ng and has lower memory requirements, but is not as accurate as pplacer.…”

“…The first is a software package [4], which computes the numerical model parameters on the backbone tree (required for the maximum likelihood placement of query sequences) using FastTree 2 [14] instead of RAxML. This substitution allows pplacer to run on larger datasets without producing negative infinity log likelihood values, as shown in [1, 21].…”

<strong></strong> Recent Progress on Methods for Estimating and Updating Large Phylogenies

“…Other phylogenetic placement methods have been developed that seek to improve scalability to larger trees or reduce running time (e.g., UShER (Turakhia et al, 2021), EPA-ng (Barbera et al, 2019), APPLES (Balaban et al, 2020), and APPLES-2 (Balaban et al, 2021)). EPA-ng is likelihood-based and has been optimized for "batch processing" of query sequences (so that the cost of performing phylogenetic placement of a large number of query sequences is much less than the cost of placing them one-by-one).…”

“…EPA-ng has slightly reduced accuracy compared to pplacer. APPLES is a very fast distance-based method; recent studies (Balaban et al, 2020(Balaban et al, , 2021Wedell et al, 2021) showed that APPLES can run on trees with 200,000 leaves and is much faster than both pplacer and EPA-ng. APPLES-2 is an improvement on APPLES with respect to accuracy and running time, and also scales to at least 200,000 sequences.…”