TitleThe total microbiome functions in bacteria and fungi
Publication TypeJournal Article
Year of Publication2020
AuthorsStarke, R, Čapek, P, Morais, D, Callister, SJ, Jehmlich, N
JournalJournal of Proteomics

Unveiling the relationship between phylogeny and function of the microbiome is crucial to determine its con-tribution to ecosystem functioning. However, while there is a considerable amount of information on microbialphylogenetic diversity, our understanding of its relationship to functional diversity is still scarce. Here wepredicted the total microbiome functions of bacteria and fungi on Earth using the total known functions fromlevel 3 of KEGG Orthology by modelling the increase of functions with increasing diversity of bacteria or fungi.For bacteria and fungi, the unsaturated model described the data significantly better (for bothP< 2.2e-16),suggesting the presence of two types of functions. Widespread functions ubiquitous in every living organism thatmake up two thirds of our current knowledge of microbiome functions are separated from rare functions fromspecialised enzymes present in only a few species. Given previous estimates on species richness, we predicted aglobal total of 35.5 million functions in bacteria and 3.2 million in fungi; of which only 0.02% and 0.14% areknown today. Our approach highlights the necessity of novel and more sophisticated methods to unveil theentirety of rare functions to fully understand the involvement of the microbiome in ecosystem functioning.Significance:The functionality of and within a microbial community is generally inferred based on the taxo-nomic annotation of the organism. However, our understanding of functional diversity and how it relates totaxonomy is still limited. Here we predict the total microbiome functionality in bacteria and fungi on Earth usingknown and annotated protein-coding sequences in species accumulation curves. Our estimates reveal that themajority of functionality (> 99%) could be assigned to yet unknown and rare functions, highlighting that ourcurrent knowledge is incomplete and functional inference is thus lackluster