Computer engineers lend a hand to modelling biological systems

    Computer engineering is prying open biological secrets

    October 11, 2012

    Edmonton—Mapping out biological systems is a difficult problem for science to tackle. The complexity and randomness of the human body make it tough to accurately predict how it will react and change over time without biological testing, which isn't always feasible or safe.


    But a new process developed by ECE professor Jie Han and graduate student Jinghang Liang aims to rectify the shortcomings of existing methods of modelling biological systems, by applying some long-standing computer engineering concepts to gene regulatory networks.


    Gene regulatory networks (GRNs) are the systems that allow certain genes to influence others to change states. Being able to model and predict how genes may influence one another is incredibly complex, but by applying a concept known as stochastic Boolean networks, Han and his team have been able to make the process more efficient.


    "We basically look at these gene regulatory networks as if they were a binary logic circuit. So ones or zeros," Han said. "Or in the case of genes, expressed or suppressed."


    Using this model of computing, they can insert randomness into their testing, as well as input certain probabilities from the outset. This allows them to dynamically simulate GRNs over a period time as they evolve, or in space. For instance, they can simulate a system in one way, then add in a logic operator (such as how a gene therapy drug would work) to force a specific gene into a state. They are then able to compare the two models.


    At present, simulating these systems is done mathematically, and each gene caussd an exponential increase in complexity. This makes it difficult to get beyond modelling a system for 10 genes, which would give over a million possible entry points alone for the model. Using the stochastic Boolean method, the process is much more efficient and can even provide an alternative to biological studies.


    “We’re currently at models with 12 genes, and the team is working on 14 right now. Theoretically, the model could scale to very complex networks."


    However, the new method is still limited by just how complex biological systems are. The team needs starting points for models, and that means it needs to know how one gene influences another — information that isn't readily available.


    “A lot of it is ‘pretty sure,’ ” Han said. “ ‘One gene changed and we’re pretty sure this gene influenced it.’ It happens, but it’s difficult to tell in a specific time or place. Again, this comes back to just how complex biological systems are. But using our model, we can link a lot of info that is found from biological experiments that seem random; we can make sense of it.”


    The team’s findings were recently published in the journal BMC Systems Biology, and it has quickly become one of the most accessed articles in its issue, something that caught Han off guard.


    “I guess it was because the topic is broad. Anyone working in genetics might be interested in this,” he said. “Also, it's a methodology paper, not just research. So we're not just applying something that already exists, but we're creating new tools.”


    Some of the surprise also came from the fact that work done by computer engineers was getting attention in biomedical field. Liang admits that when the project started, biology wasn't exactly his field of expertise.


    “Three years ago we didn't know anything about biology. We are computer engineers! So it's been a great learning experience for us.”