Coev-web: a web platform designed to simulate and evaluate coevolving positions along a phylogenetic tree

Abstract

Background: Available methods to simulate nucleotide or amino acid data typically use Markov models to simulate each position independently. These approaches are not appropriate to assess the performance of combinatorial and probabilistic methods that look for coevolving positions in nucleotide or amino acid sequences.

Results: We have developed a web-based platform that gives a user-friendly access to two phylogenetic-based methods implementing the Coev model: the evaluation of coevolving scores and the simulation of coevolving positions. We have also extended the capabilities of the Coev model to allow for the generalization of the alphabet used in the Markov model, which can now analyse both nucleotide and amino acid data sets. The simulation of coevolving positions is novel and builds upon the developments of the Coev model. It allows user to simulate pairs of dependent nucleotide or amino acid positions.

Conclusions: The main focus of our paper is the new simulation method we present for coevolving positions. The implementation of this method is embedded within the web platform Coev-web that is freely accessible at http://coev.vital-it.ch/, and was tested in most modern web browsers.

Fig. 1

Simulation interface of the Coev-web platform. Part a shows the job submission page while part b gives an example of the results obtained for 15 pairs

Fig. 2

Correlation between the number of lineages with double substitutions and the likelihood difference between Coev model and independent model for amino acid and nucleotide sequences. The X axis reflects the number of lineages with double substitutions. In both plots the same tree is used and it is composed of 100 leaves. The likelihood difference increases as X increases. The likelihood difference represented by ΔAIC shows that the Coev model is preferred to the independent model for amino acid and nucleotide sequences especially when X is big. (1.) The combinations used for nucleotide experiment are Adenine-Adenine (AA) and Thymine-Thymine (TT). (2.) The combinations used for the amino acid experiment are Alanine-Alanine (AA) and Threonine-Threonine (TT)

Fig. 3

Simulation. We present the simulation steps of two nucleotides pairs along a phylogenetic tree of 4 leafs. In red we highlight the nucleotide changes. (1.) We randomly pick a state at the root. (2.) We assign internal node states using the transition probability matrix P(t)=e ^Qt where Q is the Coev instantaneous rate matrix and t is the branch length [7]. (3.) The simulated pairs are the pairs assigned to the leafs of the phylogenetic tree

Fig. 4

Simulation plots. We plot the proportion of combinations simulated, that belong to the profile {AA, CC}, against d/s ratio along different branch lengths (a for 0.1; b for 0.5; c for 1; d for 5). Each box plot is obtained by varying the r ₁ and r ₂ rates within the range [1,100] and by randomly picking an ancestral state from the frequency vector issued from the matrix Q