The rise of molecular and cellular biology at the end of the twentieth century was accompanied by a proliferation of molecular representations in books and articles. These representations contributed to the visibility and success of the new discipline. I will focus on two different types of representations: protein structures, and cell-signalling pathways. The representation of protein structures became a major challenge at the beginning of the 1960s, not only from the accumulation of new data, but also because these data were at odds with what had been expected. The representation of protein structures acknowledged a dramatic change twenty years later through the lonely work of Jane Richardson. During the next thirty years, the representation has remained remarkably stable. In contrast, the representation of cell-signalling pathways was a shared enterprise, parallel to the description of these pathways, and the characterization of their constituents. The origin of these representations will be traced back: although rooted in chemistry, they have acquired specific characteristics. The nature of these characteristics, their explanatory role in extant biology, but also their limits will be discussed. In particular, the relation between representations and the technologies that generate them will be considered.