On cross-disciplinary ambiguity and conference comfort zones
I recently attended a conference which was unusual for me as most of the speakers come from a computer science culture, rather than a biology one. Somewhat outside my comfort zone. The science that was discussed was quite different from the more biological bioinformatics meetings: the reason being the motivation of the scientists, and what they value in their research culture.
Biology is a discovery science. Earth’s life is out there, and the biologist’s aim is to discover new things about it. Whether it’s a new species, a new cellular mechanism, a new important gene function, a new disease or a new understanding of a known disease. Biology a science of observations and discoveries. It is also a science of history: evolutionary biology aims to find the true relationships among species, which is historical research.
In contrast, in computer science, the goal is the study and development of computational systems. Chiefly the feasibility, efficiency and structure of algorithms. So we have two different drivers here: in biology, we try to discover and/or “fill in the blanks” from what we see in nature. In computer science, we seek to understand and better perform computation.
When shall the twain meet? When the problem in biology is that of information processing, and when computer science can innovate in processing that information. Textbook case in point: a basic statistical model in biology today is that of sequence evolution. It states that, given two DNA sequences descended from a common ancestor, their descent can be depicted as a series of nucleotide deletions, insertions and substitutions. In fact, since historically a deletion event in one sequence can be viewed as an insertion event in another, the model actually narrows down to two types of historical events: the insertion/deletion event (or indel), and the substitution event. The model turns out to be a powerful tool, since it can be used to make predictions. Namely, if two DNA (or protein) sequences are found to have a relatively small number of indel and substitution events between them, they are considered homologous. The “relatively small number” is key here, and understanding when the number of steps is small enough to call the similarity between the sequences homology is a whole field unto itself. Finding homologous sequences is important for understanding evolutionary history, but not only for that. If the sequences are homologous, there is a good chance that the proteins they encode have similar functions, even in different organisms: which is the basis for the use of model organisms throughout biomedical research.
But at this point is where the biologist and the computer scientist may part ways. The biologist (here “biologist” = shorthand for “biological-discovery-oriented-researcher”) will continue to treat the sequence editing model as a tool for discovering things about life, such as finding a human homolog to a gene we know is involved in cancer in mice. The computer scientist (== “computational method investigator and/or developer”) may wish to refine the algorithm or create a new one to make the process faster, or more memory-efficient.
When do we have a problem? When researchers in one field do not understand the purview of the other, and seek a measure of simplicity where there is none. I was once told by a rather prominent virologist that “bioinformatics is all about pipelines”. I asked him what he meant by that and he basically said that all he really needs to have is a tool that will give him a result and a e-value “like BLAST”). When I said that statistical significance is, at best, one of several metrics that can be used to understand results, and that sometimes it does not coincide with biological significance or is simply inappropriate, he replied that “well, it shouldn’t be that way, as a biologist I need to know whether the result means something or not, and have a simple metric that tells me that”.
On the other hand, I had computer scientist claim that, since some proteins are a product fused from different genes (he meant different ORFs actually) , this phenomenon upends the definition of the gene, and that we should actually have a “new biology” which is not “gene centric”. To that, my reply was twofold: first, that biology is not “gene centric” any more than it is “ATP centric” or “photosynthesis centric”; and second, that the best description I can come up with for a gene is a “unit of heredity”. The reply I got was that this definition is not a good one since it is not rigorous, and too open-ended to be workable. (Note that I could not provide a definition for a gene, only a description.)
Both the virologist and the computer scientist were seeking simplicity or unequivocality in the “other’s” field where those are not to be found. The problem stems from a misunderstanding of each other’s fields, which they see only through the interface to their own. Biologists, which think in terms of discovery using hypothesis-driven research, would like to have tools that help test their hypotheses. A computer scientist would like to have a biological phenomenon that is clear-cut and therefore amenable to rigorous modelling. Both are flummoxed when they discover that ambiguity rests in their peers’ fields, even though they can totally accept it in their own.
What is to be done? First, learn more about each other’s fields. If you are a biologist using BLAST (and almost all are), please take care to read up on the statistics behind BLAST results. This will give you an idea of the different metrics BLAST provides you with, and what their meanings are. Do the same for the other software you use, and understand it is not just all about “pipelines”. If you are a computer scientist, and (for example) are interested in genomic annotation, please respect the 150 years* of thought invested in modern biology, that naturally keeps revising. However, understanding basic biological concepts is necessary before you go about arguing against what might be an unintentional strawman.
Also, try to listen more, and attend meetings outside your comfort zone. It seems I learn more from conversations in my “non-regular” meetings than in my “regular” ones. Of course, once the “non-regular” become my “regular” meetings I will learn less, so basically I may have to constantly shift my comfort zone. Then again, to me it seems like science is always poking and prodding outside one’s comfort zone.
———–
(*I picked the publication date of Origin of Species as an arbitrary start date, one might think this is conservative and go back even further).
[…] In On cross-disciplinary ambiguity and conference comfort zones | Byte Size Biology, Iddo Friedberg makes a plea for biologists and computer scientists to pay more attention to each other’s concerns. I reproduce the first and last paragraph here: […]