I recently attended the ISMB (Intelligent Systems for Molecular Biology) conference in Chicago. This conference is one of the biggest bioinformatics conferences, with scientists from different corners of the world studying a whole range of organisms, systems, models…In fact, the ISMB conference seems to be typically split up into different sections, termed “communities of interest” (COSI in short).
For instance, I had the great opportunity to give a talk in the Function COSI, a community focussed on the problem of gene and gene product function prediction. One of my interests is the accurate annotation of metabolic enzymes, so, I found this section of the conference to be particularly enlightening to me. In particular, I was rather intrigued by the CAFA experiment (Critical Assessment of protein Function Annotation). The experiment is set up as a competition (with a deadline) to find the better predictors of function given proteins unannotated at the time. Several months past the deadline, the annotations predicted for these proteins are assessed using experimental verification after the deadline. Neat, right? This experiment makes me think of the Netflix Prize, the goal of which was to predict user ratings. In the latter case, however, the test data was actually withheld from the competitors.
BioVis (Biological Data Visualizations) was another COSI that I simply could not miss. I have spent many hours either staring at the drawings of metabolic networks, or trying to make them myself. On the one hand, I salute those who have painstakingly illustrated those networks, finding the best way to show the flow of metabolites from one pathway to another…At the same time, I wonder if there might not be an original solution for network visualization, a solution staring at us right now, chuckling at our inability to even catch a glimpse of it…a completely new way to think of metabolic pathways? It was with this mentality that I went into BioVis. I suppose I came out of this section of the conference, mind full of ideas…yet still hilariously puzzled. See, the visualization solutions presented at BioVis—while all varied—had to consider the issues specific to the problems being addressed. So, perhaps one should not seek the optimal way to view “things” in a metabolic network, but to ask what those “things” might be to begin with. However, shouldn’t we also beware of those blinders that we wear in the process, those simplifications that we need to make so that we may draw conclusions from our impervious data?
Is this too obvious? Am I being too vague?
In any case, BioVis remains as much art as it is science to me.
There is so much more I want to tell you. There was a fantastic keynote presentation given by Dr Madan Babu on protein disorder. I attended an awesome tutorial on deep learning for network biology. Personally, I myself loved being in Chicago, trying all kinds of food, seeing all kinds of sights, hearing the heartbeats of a city so alike and yet so different from Toronto. Alas, I find myself running out of time. Let us have this conversation another day.