In this report, we discuss the following topic from cancer genetics and epigenetics: intra tumor heterogeneity deconvolution from bulk RNA-seq data. A diverse collection of cells with specific expression profiles are present in a tumor. Being able to evaluate the different cell states, their proportions and expression profiles from a bulk RNA-seq raw counts would help in the creation of tailored treatments at reduced cost and in a short amount of time. This work is part of a semester project. We concentrate here on the core of the project and leave aside the additional tasks performed during the same time. We laid the groundwork for future study on intra tumor heterogeneity deconvolution from bulk RNA-seq data by working on both the simulation of bulk RNA-seq data from single-cell RNA-seq data and the classification of melanoma cells using a semi-supervised method. In this project, we use TuPro dataset and therefore focus our work on melanoma cancers.