Mar 20 2012

Fruity Loops and Arrowheads

One of the features of any network is the appearance of motifs, patterns (sub-graphs) that recur within a network much more often than expected at random. These small circuits can be considered as simple building blocks from which the network is composed. This analogy is quite useful, since many of these motifs would appear to have their corollaries in electronic circuitry. Motifs appear to play an important role in transcription factor regulation, which is pretty significant, because transcription factors regulate the expression of most genes.

Electronic circuits spend a lot of their time filtering out noise, and many motifs in molecular biology do this same function. Most of this noise is stochastic: a charmingly ancient Greek word that more-or-less equates to random, but not exactly, being derived from the Greek stochastikos, ‘skillful in aiming.’ To me, stochastic embodies the old saw that ‘if something can happen, it will.’

Stochastic occurrences mess up our pretty, deterministic, view of how things happen. For example, the response of our bodies to radiation therapy is stochastic since not every cell receiving the radiation energy is at a point in its life cycle when the radiation therapy can effect it: Some cells are (those currently reproducing) whilst other cells are not (those currently at rest). This is a problem when one uses methods of measurement that are Gaussian (i.e they ‘average things out’) because these types of measurement will yield indications of a more gradated response than really occurred.

Stochastic music was pioneered by the composer Iannis Xenakis. Unfortunately, while undoubtedly making for interesting math, I agree with this author that the music sounds more or less like an hour-long extended visit to a junk yard. I do like some of Xenakis’ other music though, such as Metastaseis..

Stochastic versus graded response.

Most network motifs try to soften up the responsiveness of transcription factors to stimulus with a high signal to noise ratio, sort of like how an experienced parent can tell the difference between a crying child who just needs to nap versus a child in true distress.

Since developing the Quodlibet module in Datapunk I’ve been increasingly aware of the need to incorporate these types of motifs in my network calculations, when and where they show up. Quodlibet currently does quite a bit of graph/network/combinatorial calculation already (click on the Analytics link in any molecular map to see it at work) so looking for network motifs seemed the logical next step.

I mostly work in Perl, so I often take advantage of CPAN (The Comprehensive Perl Archive Network) which hosts a wealth of Perl modules, including a terrific collection of bioscience libraries, such as Bioperl. These modules are what makes Perl so great: you do not have to reinvent the wheel and in many instances you don’t even have to know exactly how the wheel works. Just plug it in, feed it good stuff and get the output. CPAN has been a great asset with Quodlibet, but contained no modules for network motif detection. Fortunately an online homework assignment provided me with the means to get started.

However, before I launch into any of that stuff, in my next blog I’ll take a look at the most common network motifs. These have been identified largely through the work of Uri Alon and his lab at the Weizmann Institute of Science in Israel, and if your interest runs in the direction of computational biology, I recommend that you take a look at his book Introduction to Systems Biology: Design Principles of Biologic Circuits (Chapman and Hall/CRC). The loop is a basic network motif, so next blog we’ll take take a look a a few variations to get an idea about how these things work things.

Like the digital circuitry in your computer, clusters of network motifs are capable of computational processes. Think about this: Humans share about 98% of their genome (at least the sequences) with apes. This of course begs the question ‘why are we not more similar?’ The answer is that while we share much of the base sequences, there are tremendous differences in the ‘computational knowledge’ that acts upon these sequences, in particular the networks involved in transcription factor regulation: operons, regulons and modulons. It is the ‘combinatoric wisdom’ that seems to differentiate between the classes of life forms.

No responses yet

Leave a Reply