Next-generation high throughput sequencing technologies provide efficient, rapid, and low cost access to sequencing. Its application to transcriptomes, called RNA-seq, enables the study of both the sequence and the expression of the transcripts. Many bio-informatics methods are still developed for RNA-seq data processing, trying to get the maximum out of it. Assembly methods allow us to study non-model species (no reference genome available) as well as model species. The work presented here is mostly related to RNA-seq data on non-model species.In the first study, to understand the initiation of hybrid incompatibility, we performed a genome-wide transcriptomic analysis on ovaries from parental lines and on hybrids from reciprocal crosses of \emph{D. mojavensis} and \emph{D. arizonae}. We didn't see a global deregerulation of genes or transposable element. Instead, we show that reciprocal hybrids presented specific gene categories and few transposable element families misexpressed relative to the parental lines. The analytical workflow developed for this project will be used to analyze transcriptomic data from the testis, but also to study the reciprocal crosses from other lines of D. mojavensis with D. arizonae leading to variable levels of sterility in hybrids. A second project tacked here is the identification and quantification of SNPs from RNA-seq data without a reference genome with KisSplice. Kissplice was developed to identified several type of variants (splicing events, indels) directly from the de Bruijn graph, build from the sequenced reads. We also developed other KisSplice-tools, for downstream analyses of the SNPs, including the prediction of their impact on the protein sequence