My research topic addresses the so-called decoy selection problem in protein structure prediction. We have algorithms spewing out hundreds of thousands of three-dimensional structures computed for a given protein sequence in a few days on one CPU. These structures are known as decoys. The challenge is how to determine which of them is the native structure. This is a crucial part of the de novo structure prediction problem, where for a given protein sequence, we want to know what its native structure is as the first step to understanding anything about the function of that protein. I became interested in the topic because the protein structure problem has a lot of useful applications. Once we know the relationship between the protein structure and their function, we are able produce our own protein to used in the field of medical science. My long-term goal is to do more research about bioinformatics, and probably apply for the graduate program in this field. I think this research experience is a very good practice chance for me to get familiar with it and also the way of doing research. I think it wil also be useful for me when I want to apply for graduate school. What I do on a weekly basis is achieve the alogrithm I have with computer code. Run the code to get the distance and do the cluster for it using WEKA, we want to see which algorithm give a best and quick approach to solve this problem. So after I got the data I perform analysis on it to compare them with each other. One thing I have discovered this week, is I found the Reducing the dimensionality of the protein-folding search algorithm provided George D. Chellapa and George D. Rose give us significantly reduce the running time than do the cluster for each phi and psi angle.