Models of gene gain and gene loss for probabilistic reconstruction of gene content in the last universal common ancestor of life

Abstract

Background: The problem of probabilistic inference of gene content in the last common ancestor of several extant species with completely sequenced genomes is: for each gene that is conserved in all or some of the genomes, assign the probability that its ancestral gene was present in the genome of their last common ancestor.

Results: We have developed a family of models of gene gain and gene loss in evolution, and applied the maximum-likelihood approach that uses phylogenetic tree of prokaryotes and the record of orthologous relationships between their genes to infer the gene content of LUCA, the Last Universal Common Ancestor of all currently living cellular organisms. The crucial parameter, the ratio of gene losses and gene gains, was estimated from the data and was higher in models that take account of the number of in-paralogs in genomes than in models that treat gene presences and absences as a binary trait.

Conclusion: While the numbers of genes that are placed confidently into LUCA are similar in the ML methods and in previously published methods that use various parsimony-based approaches, the identities of genes themselves are different. Most of the models of either kind treat the genes found in many existing genomes in a similar way, assigning to them high probabilities of being ancestral ("high ancestrality"). The ML models are more likely than others to assign high ancestrality to the genes that are relatively rare in the present-day genomes.

Figure 1

Distribution of all COGs under models B2 and M2, as well as high-ancestrality COGs (LUCA-ML and LUCA1.0), by the number of genomes in which they are present.

Figure 2

Probability distribution of the COG ancestrality under various models. The first panel shows the frequency of COGs with the different probability of occurrence at LUCA, and the second panel shows the number of COGs above the different probability thresholds.

Figure 3

The reference species tree. The reference species tree was constructed using the sequences of 11 universal genes COG00081, COG00093, COG00096, COG00097, COG00099, COG00102, COG00103, COG00197, COG00244, COG00256, COG00533.

Figure 4

A toy example showing the parameters for calculation of the likelihood function. The elements of the phyletic vector of a gene are shown at the leaves. Branch lengths t₁, t₂, ⋯, t₈ are known.

Figure 5

The strategy of editing phyletic vectors on the basis of information about HGT. A horizontal gene transfer event (red arrow) from ancestral species A to species B is inferred on the basis of comparison of the topologies of the species tree and the sequence family tree (not shown in this figure). The bottom vector is obtained by replacing all the presences by absences in all descendants of B, and the inference of ancestral state can be updated on this edited vector.