Section: New Results

Biodiversity and ecology

Participants : Alain Franc [correspondant] , Jean-Marc Frigerio, Philippe Chaumeil, Razanne Issa, Leyla Mirvakhabova.

Our main activity has been code development in the framework of a research project with ONEMA, and preparing future development. Code development has been fostered with the work of Razanne Issa (CDD ONEMA) in the last three months of 2015, and preparation of further development has been fostered by welcoming Leyla Mirvakhabova (L3, National Research University Higher School of Economics (NRU HSE), Moscow, mathematics) . Declic is a python library providing tools for analysing molecular data for biodiversity stydies. The main object is a distance matrix, from which one can either build a point cloud in a Euclidean space with distances between points as close as possible from distances between reads (Multidimensional scaling), or to build a graph with edges between reads when their distance is smaller than a given threshold. Meanwhile, the team has developed the network around the Galaxy server where an early verison of tools has been installed and made available, especially with SLU at Uppsala (Maria Khalert). Alain Franc has developed a collaboration with Olivier Coulaud and Pierre Blanchard (Hiepacs) for efficient computation of eigenvectors and eigenvalues of large, dense and symmetric matrices, needed for scaling in Multidimensional scaling.

The work of Razanne Issa has made it possible to extend the declic library in the direction of machine learning, by incorporating tools from support vector machines through library sklearn. This development will be pursued in 2016. The work of Leyla Mirvakhabova has permitted a first incursion into topological data analysis as a possible approach for studying the shape of point clouds produced by multidimensional scaling. The collaboration with NRU HSE on this topic will be pursued in 2016.