Dr Rachel Cavill

Research Interests

• Metabonomics data analysis: Analysing data generated by metabonomics, mainly NMR data in my case.
• Chemometrics: Using and developing the use of algorithms for the analysis of such data.
• Integration of multiple types of "omics" data: How do we build a coherent picture of the system when we have noisy and often seemingly conflicting data from different technologies?
• Data mining: How do we extract useful knowledge from large datasets, such as those generated by biologists?
• Machine Learning: Exploring the methods which computers can use to automatically answer questions about or find patterns in large datasets.
• Evolutionary Computation (especially Genetic Algorithms and Genetic Programming): EC is a machine learning method which starts with a population of candidate solutions then 'evolves' better solutions through a simulation of evolutionary processes.
• Evolvability: It is clear that some representations (and their associated operators) are more evolvable than others; yet defining precisely what makes them evolvable and learning how to design evolvable representations remains tricky.

Current Role

I am a research associate working on the carcinoGENOMICS project in biomolecular medicine.  The principal aim of the carcinoGENOMICS project is to develop in vitro assays as an alternative to the current chronic rodent bioassay for assessing the genotoxic and carcinogenic potential of chemicals. As part of this work, we have been generating metabolic profile data describing sample sets of treated and untreated cells (liver kidney and lung) provided by our partners across Europe.  In parrallel, other partners have been generating transcriptomic and other cytomic measurements on matching samples.

My role within this project is to perform data analysis of the data we generate at Imperial and where possible to integrate this with the transcriptomic data produced elsewhere.   To achieve this we employ a whole range of methods including machine learning techniques in addition to the more standard chemometric techniques.

In March 2009 I spent 1 month as a visiting researcher at the Max Planck Institute for Molecular Genetics in Berlin, where I worked alongside other researchers from our project in developing methods for data integration at the pathway level based around the ConsensusPathDB. 

Background

Prior to this post, I obtained an MMATH in Maths and Computer Science and then a PhD within the Intelligent Systems group in the Electronics Department, both at York University.

My thesis "Multi-Chromosomal Genetic Programming", focused on developing a polyploid genetic programming system and then characterising the problems on which this system was advantageous.  Much of my PhD focused on exploring the biology behind evolutionary systems, in order to be able to apply ideas from biology (such as polyploidy) in a "bio-inspired" manner to the field of evolutioanry computation. 

I have also worked as a research assistant for 6 months at Robert Gordon University in Aberdeen, where I looked at using Inductive Logic Programming to classify G-Protein Coupled Receptors; and for 3 months in the Maths Department at York, performing simulations of introducing congestion charging into city centre traffic networks.

Matlab Code

Genetic Algorithm code.