Comparing grid computing solutions for reverse-engineering gene regulatory networks

Models of gene regulatory networks encapsulate important features of cell behaviour, and understanding gene regulatory networks is important for a wide range of biomedical applications. Network models may be constructed using reverse-engineering techniques based on evolutionary algorithms. This optimisation process can be very computationally intensive, however its computational requirements can be met using grid computing techniques. In this paper we compare two grid infrastructures. First we implement our reverse-engineering software on an opportunistic grid computing platform. We discuss the advantages and disadvantages of this approach, and then go on to describe an improved implementation using the QosCosCrid, a quasi-opportunistic supercomputing framework (Qos) for complex systems applications (Cos). The QosCosCrid is able to provide advanced support for parallelised applications, across different administrative domains and this allows more sophisticated reverse-engineering approaches to be explored.