Thesis Information

This page is for information relating to my MPhil thesis project, titled "A Comparison of Parallel Global Optimisation Algorithms for Reverse Engineering Gene Networks".

The project involved compared the performance of optimisation algorithms (the Parallel Lam Simulated Annealing algorithm, and a Parallel Evolutionary Strategy that I developed). The algorithms were used to fit a model of a gap gene network in the early Drosophila embryo to high-resolution spatio-temporal gene expression data. The Parallel ES was found to be more reliable and faster than the Parallel SA.

My project supervisor was Yogi Jaeger.

A PDF copy is available here

Source Code

There are two separate programs, one for the Parallel ES and one for Lam SA. The code I used is modified from the code used by Jaeger (2004) and Fomekong-Nanfack (2007) for SA and ES respectively. Follow the instructions below to install both programs.

Parallel Evolutionary Strategy

1. Download and unpack the serial ES code from here into a directory DIR
2. Replace DIR/src/fly_ILESDS.cpp with this file
3. Replace DIR/src/fly/fly_opt.c with this file
4. Replace DIR/src/myUtil/ga_opt.h with this file
5. Install as instructed in DIR/Doc/README, but when running ./configure use the command
   
   $ CC=mpicc CCFLAGS="-DUSING_MPI" CXX=mpiCC CXXFLAGS="-DUSING_MPI" ./configure
   
   where mpicc is your MPI C compiler, and mpiCC is your MPI C++ compiler
6. The MPI binary for the Parallel ES is DIR/src/fly_ILESDS, to be run with mpiexec. It behaves just as in the serial case, except it uses a number of islands equal to the number of nodes passed to mpiexec.

Parallel Lam Simulated Annealing

1. Download and unpack the parallel fly code from here into a directory DIR
2. Replace DIR/lam/lsa.c with this file
3. Install as instructed in DIR/README

Note that the extra SA code only adds time-stamping to the logs, and is used to produce descent curves.

Data

The data-set used in the thesis is available here, in the file format that both programs accept.

References

J. Jaeger et al (2004) Dynamic control of positional information in the early Drosophila embryo. Nature, 430(6997):368-371 PDF

Y. Fomekong-Nanfack et al (2007) Efficient parameter estimation for spatio-temporal models of pattern formation: case study of Drosophila melanogaster. Bioinformatics, 23(24):3356-3363 Link

Back

Save the cheerleader, save the world.