The aim of the project is to explore the performance of one of the ExTASY tools – CoCo-MD – for advanced simulation when applied to a biologically-relevant test case, the protein BPTI. Very long and computationally expensive conventional simulations have shown how hard it is to properly explore the shape-space of BPTI – can CoCo-MD do better? Using the ExTASY toolkit, you will run advanced sampling simulations of BPTI on a range of supercomputer resources. Working with biomoleciular simulation scientists in the UK, you will analyse the sampling data produced, and optimise the simulation parameters to identify how best performance can be achieved.

In the folder Papers is a copy of the 2010 Science paper by Shaw et al that describes their 1 millisecond simulation of BPTI.

In the folder Shaw_Data_Analysis is a copy of the trajectory file from that simulation, a copy of a PDB-format file of the sytructure of bpti, and an ipython notebook (and markdown version) showing how I have used PCA analysis to identify the 'rare' conformational state seen in that simulation. Also there is another ipython notebook that shows how you can analyse a slug of trajectory data of the type that this project should produce (in much greater volume!), and see if it is getting anywhere near this rare state.

The tarball gmxcoco.tgz contains a first stab at the CoCo-MD workflow for this project. It has been adapted from the examples used in the Edinburgh workshop last year. It has been tested on Archer and runs OK, but will probably need tweaking for use on another HPC resource, and in any case the real thing will need more replicates and cycles. A couple of README files are included which may help.