I’m currently reading my way through literary texts on art and science while I gather pace for potentially writing my own book proposal. Interdisciplinary works such as these are always a great source of rage, just in case this is something you feel is missing from your world. I used to attend systems biology lectures and conferences (combining mathematics and biology disciplines), and these were just the same, so much so that I had to stop attending in the end. The talks were invariably given by mathematicians who were trying to use their tools to describe systems in biology, such as cellular signalling pathways. They almost always consisted of a couple of slides of poorly understood biological theory, followed by many slides detailing complex differential equations. One of my lecturers once told me that some mathematicians believe that you can cut a cell in half and it’ll still operate as usual.
I should know better than to read interdisciplinary texts before bed, only the previous book I was reading  was doing quite well in its last few chapters, so I’d forgotten what a problem they could be. In fact, looking back, the last book has so many scrawled notes in the margins of the first chapter that they continue into a separate notebook. It seems that for these types of works, it is the opening chapters that present the most difficulty: the author must bring together diverse concepts from seemingly disparate fields to provide an introduction to the book. In later chapters, the author can safely write in the discipline they know (e.g. art), and can get away with such sweeping statements as “metaphor for the discipline of science”. In the next book , the offending article was in the preface.
A. S. Byatt is asking why human beings make art at all.
The answer, often enough, seems to be, for the same reason that they make metaphors. I don’t understand that either – except that metaphor-making is a fundamental part of the way the human mind makes connections. And art explores connections like those in ways very different from science’s orderings – even though scientists are aided by flashes of intuition.
In this last sentence it appears that the author believes that all that scientists do is discover things, then categorise them. This immediately puts one in mind of the avid collectors of the Victorian period, such as the oologists who collected eggs from wild birds such as the Osprey (Pandion haliaetus), pushing them to the brink of extinction. I doubt whether many scientists today are actively engaged in this kind of process. With the rise of bioinformatics, there was a fashion for using computer algorithms to classify, for example, genes, proteins and protein structures, with the view of being able to use these databases to predict the structure or function of an unknown protein, just from its sequence. Whereas structure could to some extent be predicted, especially if one already had a good template to work from, the prediction of function from primary sequence alone remains near impossible. These bioinformatics groups are now working with experimental biologists with high throughput systems, and are becoming increasingly frustrated that protocols in the lab work on a much longer timescale than those in a computer.
In fact the ability to make connections between seemingly abstract concepts is key to making ground-breaking discoveries and being a ‘good’ scientist. My PhD project began life with some basic searches of online databases in which bioinformaticians had processed and classified large datasets from high throughput screens. My starting point was a protein for which we had a sequence but no structure or function (RdgBβ, PITPNC1). I was able to use computer modelling techniques to predict its structure (although this has only partially been confirmed by experimentation) . I gained clues as to its function from databases reporting protein-protein interactions, often from yeast two-hybrid screens.
To carry out a yeast two-hybrid screen, protein sequences are engineered to include either a DNA-binding domain or an activation domain within a plasmid. One of each type of plasmid is put into a mutant yeast strain lacking the ability to make a particular nutrient for itself (e.g. an amino acid). The fusion proteins (protein-DNA-binding domain or protein-activation domain) are synthesised from the plasmids by the yeast, and if the two proteins interact, the DNA-binding and activation domains will be brought into close proximity, together forming a transcription factor that will drive expression of a gene allowing the yeast to make its missing amino acid. In brief, if the mutant yeast is able to grow on the media lacking the essential amino acid, the two proteins can be said to interact with one another.
Some research groups specialise in operating these kind of high throughput screens using every protein in the human genome with the view of creating a human ‘interactome’, and it is hoped that this type of data will provide clues as to the function of unknown proteins. Interactome data gets uploaded to sites such as that of The European Bioinformatics Institute. Other bioinformatics tools, such as GeneMANIA can then be used to visualise the interactions. Interactions predicted in this way must be confirmed by experiment, usually going further than a simple “yes, they interact” to defining the site of interaction between the two proteins and the conditions under which they interact. More often than not, an interaction predicted by a yeast two-hybrid screen will not be physiologically relevant because the two proteins would never actually be in the same place at the same time within the same cell to meet one another. Often, the interaction will only take place under such specific conditions that the researcher may conclude that they do not interact, whereas really they have not stumbled upon the correct conditions.
Since many non-significant interactions are returned by high throughput screens, the eye of a human researcher is required to decide which interactions might be physiologically relevant and worth actively pursuing. This cannot be predicted by computer algorithm and requires the researcher to draw upon existing knowledge and experience in the first instance. The scientist is required to make abstract connections between proteins that might have been missed by a different eye. This is the way that scientific discoveries are made: it takes a human mind to say, “ok, if I know x about y, and p about q, I predict that if I do f, then g will happen.” Byatt was too quick to dismiss the importance of scientists’ “flashes of intuition”.
In the quote above, Byatt also questions why human beings make metaphors. Metaphors serve as a communication tool to aid in the understanding of foreign concepts. They are essential for the teaching of scientific theories and processes, which otherwise might be difficult to get to grips with. An example is the ‘lock and key’ metaphor for describing enzyme substrate recognition. Like a lock and key, it was once hypothesised that an enzyme’s substrate would exactly fit into a matching, yet opposing, shape in the enzyme. Since everyone is familiar with a lock and a key, this concept requires no further explanation. A later hypothesis proposed an ‘induced fit’ model, in which, upon coming into contact with its substrate, the enzyme alters its shape to accommodate the substrate more tightly. A mixture of the two hypotheses is now the accepted model.
I have also recently written about the value of metaphor in conceptual art, and indeed I believe it is a vital tool in the communication of science to a lay audience. In explaining abstract processes to a diverse audience, the use of a universally familiar concept saves a thousand words and provides a huge step up for the audience’s understanding.
1. Ede, S. (2005) ‘Art and Science (Art and Series)’, I B Tauris.
2. Ede, S. (ed.), Byatt, A. S. (preface) (2000) ‘Strange and Charmed: Science and the Contemporary Visual Arts’, Calouste Gulbenkian Foundation.
3. Garner, K., Li, M., Ugwuanya, N., and Cockcroft, S. (2011) ‘The phosphatidylinositol transfer protein RdgBβ binds 14-3-3 via its unstructured C-terminus, whereas its lipid-binding domain interacts with the integral membrane protein ATRAP (angiotensin II type I receptor-associated protein)’ Biochemical Journal 439: 97-111.