Using Network Algorithms on Metabolic Pathways
Bioengineers are using computational algorithms in order to study the metabolic networks present in many microbes in order to create new high demand products, such as biofuels and pharmaceuticals. The scale of these network sizes are so large – millions to billions of nodes – that most traditional algorithms become inefficient to use. Biologist Hyun-Seob Song and colleagues are instead using a new computation algorithm, called the alternate integer linear programming (AILP) to analyze these metabolic pathways. What exactly they are looking for are strong pathways that will can produce specific reactions; such edges are incredibly useful in engineering drugs that target certain disease-ridden cells, such as cancerous ones. Song hopes to that this algorithm does not stay confined in the bioengineering field; rather he believes its applications can be used in any large scale network such as road or water systems.
This article stood out to me as it is an application of several of the concepts we learn in the course. It never occurred to me that a metabolic pathway could be illustrated as a network but after reading this article, it makes a lot of sense. One tidbit that stood out to me was that this new algorithm, AILP, can find something call “minimal cut sets”, which seem analogous to the course’s concept of a component. Basically, these are areas where, if cut, can shut the reaction of the entire network down. Similarly, a component of a network is one where all the edges nodes are internally connected, but not externally so. Such cut sets might have reactions within them, but will not continue beyond their component. Song hopes to use these “cut sets” to prevent the growth of tumors, which is a very amazing application of network science.
Another reason this article inspired this blog post was a question I had while reading. Since the network in this article is modeled as microbes being the nodes and their reaction being edges, would the concept of triadic closure be present here? More robust reactions are modeled using stronger edges, so there is a property of an edge being weak or strong. Would two nodes that have strong reactions with a third, also have a reaction with each other? It’s an interesting question, I think, with a lot of consequences if true. One could induce new reactions by introducing new edges between the nodes. I bet that AILP could prove whether the Strong Triadic Closure Property holds and I hope Song and his colleagues are making good use of it if it does.