Evomorph: Evolutionary morphometrics tool using Aegla singularis data

CABRERA JM; GIRI F

Bahia Blanca

Congreso; VI Argentinian Conference on Bioinformatics and Computational Biology; 2015

Asociacion Argentina de BIoinformática

BackgroundMacroevolutionary processes are difficult to measure as evidence in living organism. Computational simulation is an acceptable way to evaluate and understand those processes over major scales [1,2]. You can simulate evolution in a population varying different parameters (effective population size, phenotypic variance, and heritability), under different types of evolutionary mechanisms (drift, selection, etc.). In this work, we present an informatics tool to study evolutionary aspects using shape data of any species.Materials and methodsEvomorph was developed as an R package. This package include geometric morphometric data handling tool, distances between shapes calculation, superimposition residuals (GPA), simmetrization, evolutionary process simulation, and plot functions.An evolutionary simulation was performed using geometric morphometric data of freshwater crab Aegla singularis. Different chart types (grids and deformation vectors, graphic distances) were used to show the variation of the forms generated during the simulation with respect to the initial form.ResultsBefore the simulation, the data was pre-processed using the superimposition function of the package, reducing its dimension and reducing the computational requirements.Simulations were performed for both evolutionary models (random walk and directional evolution) over one million years (one million steps). The coefficients for the simulations were sampled from a normal distribution with different mean and deviation. Overestimated rate variation (mean μ = 1) was used to observe the effect of different evolutionary models of shape variation. The results were stored as images files showing shape variation over time (using different kind of plot styles).ConclusionThe tool managed to represent adequately the evolutionary mechanism simulated. The simulation performs very well on a low-end desktop computer, although it could be optimized for bigger data sets. In addition, statistical tools could be added to see if the simulation parameters (selection coefficients) fit the data [3], improving the reliability of the simulation.