Genomic Data Quality Impacts Automated Detection of Lateral Gene Transfer in Fungi

Abstract

Lateral gene transfer (LGT, also known as horizontal gene transfer), an atypical mechanism of transferring genes between species, has almost become the default explanation for genes that display an unexpected composition or phylogeny. Numerous methods of detecting LGT events all rely on two fundamental strategies: primary structure composition or gene tree/species tree comparisons. Discouragingly, the results of these different approaches rarely coincide. With the wealth of genome data now available, detection of laterally transferred genes is increasingly being attempted in large uncurated eukaryotic datasets. However, detection methods depend greatly on the quality of the underlying genomic data, which are typically complex for eukaryotes. Furthermore, given the automated nature of genomic data collection, it is typically impractical to manually verify all protein or gene models, orthology predictions, and multiple sequence alignments, requiring researchers to accept a substantial margin of error in their datasets. Using a test case comprising plant-associated genomes across the fungal kingdom, this study reveals that composition- and phylogeny-based methods have little statistical power to detect laterally transferred genes. In particular, phylogenetic methods reveal extreme levels of topological variation in fungal gene trees, the vast majority of which show departures from the canonical species tree. Therefore, it is inherently challenging to detect LGT events in typical eukaryotic genomes. This finding is in striking contrast to the large number of claims for laterally transferred genes in eukaryotic species that routinely appear in the literature, and questions how many of these proposed examples are statistically well supported.

Orthologous peramine and pyrrolopyrazine‐producing biosynthetic gene clusters in Metarhizium rileyi, Metarhizium majus and Cladonia grayi

“…This suggests these sequences may be approaching the mutational divergence limit for orthologous peramine synthetase‐encoding genes. An alternative hypothesis is that of lateral gene transfer, but this is not supported by the perA sequence phylogeny, and we lack sufficient genome data from PPZ ‐containing species for a more conclusive analysis (Dupont and Cox, ).…”

Too Much Eukaryote LGT

“…It is apparent neither to me nor to others why eukaryote LGT should be preferred over both vertical inheritance and gene loss as the first line of interpretation for eukaryote gene trees, while less spectacular mechanisms that conform to the rules of eukaryote genetics such as vertical inheritance within species, gene duplication, and differential loss are implicitly penalized as unlikely processes. Why should gene duplication and gene loss in addition to the ever‐present existence of random phylogenetic error be viewed as unlikely causes for unexpected database search results or unexpected branches in trees?…”

Inferring synteny between genome assemblies: a systematic evaluation

“…Analysis of synteny in closely related species is now the norm for every new published genome. However, assembly quality comes into question as it has been demonstrated to affect subsequent analysis such as annotation or rate of lateral transfer [32, 33]. …”