MrBayes tgMC³: a tight GPU implementation of MrBayes

Episodic evolution of a eukaryotic NADK repertoire of ancient provenance

NAD kinase (NADK) is the sole enzyme that phosphorylates nicotinamide adenine dinucleotide (NAD+/NADH) into NADP+/NADPH, which provides the chemical reducing power in anabolic (biosynthetic) pathways. While prokaryotes typically encode a single NADK, eukaryotes encode multiple NADKs. How these different NADK genes are all related to each other and those of prokaryotes is not known. Here we conduct phylogenetic analysis of NADK genes and identify major clade-defining patterns of NADK evolution. First, almost all eukaryotic NADK genes belong to one of two ancient eukaryotic sister clades corresponding to cytosolic (“cyto”) and mitochondrial (“mito”) clades. Secondly, we find that the cyto-clade NADK gene is duplicated in connection with loss of the mito-clade NADK gene in several eukaryotic clades or with acquisition of plastids in Archaeplastida. Thirdly, we find that horizontal gene transfers from proteobacteria have replaced mitochondrial NADK genes in only a few rare cases. Last, we find that the eukaryotic cyto and mito paralogs are unrelated to independent duplications that occurred in sporulating bacteria, once in mycelial Actinobacteria and once in aerobic endospore-forming Firmicutes. Altogether these findings show that the eukaryotic NADK gene repertoire is ancient and evolves episodically with major evolutionary transitions.

MrBayes tgMC3++: A High Performance and Resource-Efficient GPU-Oriented Phylogenetic Analysis Method

MrBayes is a widespread phylogenetic inference tool harnessing empirical evolutionary models and Bayesian statistics. However, the computational cost on the likelihood estimation is very expensive, resulting in undesirably long execution time. Although a number of multi-threaded optimizations have been proposed to speed up MrBayes, there are bottlenecks that severely limit the GPU thread-level parallelism of likelihood estimations. This study proposes a high performance and resource-efficient method for GPU-oriented parallelization of likelihood estimations. Instead of having to rely on empirical programming, the proposed novel decomposition storage model implements high performance data transfers implicitly. In terms of performance improvement, a speedup factor of up to 178 can be achieved on the analysis of simulated datasets by four Tesla K40 cards. In comparison to the other publicly available GPU-oriented MrBayes, the tgMC³++ method (proposed herein) outperforms the tgMC³ (v1.0), nMC³ (v2.1.1) and oMC³ (v1.00) methods by speedup factors of up to 1.6, 1.9 and 2.9, respectively. Moreover, tgMC³++ supports more evolutionary models and gamma categories, which previous GPU-oriented methods fail to take into analysis.