This paper presents a novel approach to the computational representation of cellular lines using transformer-based embeddings. By leveraging state-of-the-art natural language processing techniques, we generate context-aware embeddings from biomedical literature from the PubMed database, offering a more nuanced and biologically relevant representation of cellular lines compared to traditional methods like TF-IDF and SVDD. We applied these embeddings to cluster cellular lines, using the elbow method to identify a set of distinct clusters that reflect biologically meaningful relationships. To evaluate the quality of these clusters, we employed the Topic Coherence metric, achieving a coherence score of 0.395, indicative of moderate consistency across clusters. The results demonstrate the potential of transformer-based models to improve drug discovery by identifying shared characteristics between cellular lines, enabling more accurate drug response predictions and advancing personalized medicine. This method offers an interesting improvement in the precision of cellular line modeling, paving the way for more efficient drug repositioning and targeted therapies in cancer research.