In-vivo automatic model inference of gene networks via Optimal Experimental Design in microfluidics

Enabling the quantitative analysis and synthesis of biological networks, over the past decade, Systems and Synthetic Biology played key role in extending our understanding of biological systems. Mathematical models, at the heart of both the disciplines, are therefore enabling a “silent revolution” in life sciences. Unfortunately, obtaining such models still remains an exceedingly laborious and expensive activity. This is mainly due to how model identification experiments have been carried out so far in biology: mainly limited by technology, we have used “step” or “pulse” stimuli, even though it has long been known these led to lowly-informative experiments. How can we then maximize the information associated to an experiment in Systems and Synthetic biology?

Optimal Experimental Design (OED), a discipline at the intersection of Statistics and Control Theory allows us to accomplish this by continuously estimating the best stimulus to apply in real-time to the cells to extract the largest amount of information from an experiment. In doing so OED allows to minimize time required to identify a mathematical models and reduce the cost associated to the experiments, e.g. reagents – usually a significant fraction of research expenses in this field. We have now conceived a new experimental setup that combines microfluidics and fluorescence microscopy and allows us, for the first time, to implement arbitrarily complex stimuli, like the ones required to exploit OED in biology. Yet a challenge remains to be solved: the search of the optimal stimulus OED carries out is extremely computationally expensive which poses a severe issue on the real-time nature of the problem. General Purpose GPU computing, the most advanced form of computation available, will be exploited to solve this problem as it allows a seamless parallelization of the optimization task (the successful candidate will be provided with appropriate training).

As part of this project you will design an algorithm that will automatically infer Ordinary Differential Equations models of biological systems using Bayesian statistics (i.e. particle filtering) and work on the first implementation of an Optimal Experimental Design algorithm on our CUDA enabled computerized microscope.

Further information can be found at: http://menolascina.mit.edu

Closing Date: 

Wednesday, November 1, 2017
School of Engineering logo
School of Engineering

Principal Supervisor: 

Eligibility: 

Applications from students with a background in Biology/Biotechnology, Mathematics, Physics and Engineering are encouraged.

Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possibly supported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.

Funding: 

Informal Enquiries: