Abstract : Any living organism is the result of complex interactions between its genome and its environment, interactions characterizedby transfers of matter and energy required for the survival of the organism and the transmission of its genome. Since the discovery in the years 1910 that the chromosome is the mechanical basis of the genetic information, the biologists study genomes in order to decipher the mecanisms and processes operating in the development of organisms and the evolution of populations. Thanks to the technological improvements of the last decades, several genomes were fully sequenced, their number increasing quickly, but they are far from being deciphered. Indeed, some of their components, the transposable elements, are still not well understood, although they were detected in almost every species studied so far, and they can account for up to 90% of their genome. Transposable elements are DNA sequences that can move and duplicate within genomes. They hence have a major impact on genome structure but also on the expression of neighbouring genes, notably via epigenetic mechanisms. Their evolution is also peculiar as they have a non-mendelian vertical transmission and as numerous cases of horizontal transfers were highlighted. However, except for some model organisms for which reference quences are available, the annotation of transposable elements often corresponds to a bottleneck in the analysis of genomic sequences. Moreover, comparative genomics studies have shown that genomes are much more dynamic than previously expected, particularly in plants, thus making even more difficult the precise annotation of transposable elements. During my PhD work, I started by comparing existing computer programs used in de novo approaches of transposable element identification. In this aim, I designed a test protocol on the genomes of Drosophila melanogaster and Arabidopsis thaliana. As a result, I proposed a de novo approach combining several tools, thus enabling the automatic recovery of a great numberof reference sequences. Moreover, I showed that our approach highlighted the structural variations present within well-known families, notably by distinguishing structural variants belonging to a same family of transposable elements, thus reflecting the diversification of such families during their evolution. This approach was implemented in a package (REPET) making possible the analysis of transposable elements in numerous genomes from plants, insects and fungi among others. This work lead to a roadmap describing, from a practical point of view, how to annotate the transposable element content of any newly sequenced genome. As a consequence, many questions about the impact of these elements on the evolution of genome structure can now be tackled using several genomes more or less related withe ach other. I also propose several perspectives, notably the simulation of the data required for the improvement of the tools, a way complementary to the modeling of transposable element dynamics.