In this presentation, I will argue that computers and the data they produce should be approached as historical entities, in order to properly address their interactions with biomedical research. I will present three chronologically consecutive case studies in which the available computers at each time crucially shaped the type of data biomedical scientists obtained and the way they organised their research networks.

In the late 1950s and 60s, researchers working on the genetic code used mainframe computers to test different hypothetical models of interaction between genes and proteins. This led them to conceptualise this interaction as an input-output process inspired by the mechanisms which governed the internal architecture of mainframes at that time. The linear-straightforward model of interaction between DNA and proteins which resulted from that conceptualisation has long survived the use of mainframe apparatus.

Towards the late 1970s biomedical research centres gradually introduced minicomputers, which were located in specific rooms and shared by biologists and personnel in charge of the programming and technical support. This eased cooperation between the two types of workers and strengthened a new category of data: strings of interconnected characters, such as DNA sequences, which were inputted into the computer by the user.

The next decade, in the 1980s, the first DNA sequence databases were created in microcomputers, smaller and cheaper apparatus adapted to individual laboratories or offices. This led to the recruitment, by biomedical centres, of independent teams of systems engineers in charge of database design. The systems engineers modelled the database architecture on networks which connected the DNA sequences with various multilayered features. They also faced adaptation problems, due to their lack of interaction with the user biologists working at the same research institutions.