Weekly Journals

[ Main Page | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | final ]


Week 1 (May 23 - May 29)

After Amarda, the grad student I will be working with this summer, gave me a quick run-down of her work (determining protein structure using robotic techniques), I started reading up on some background, focusing on material from one of Dr. Kavraki's classes. I have a decent background in biology, so I concentrated mostly on robotic methods, learning about inverse kinematics (IK), Denavit-Hartenberg (DH) local frames, and cyclic coordinate descent (CCD). I also wrote a simple program in MATLAB to rotate dihedral bonds and played around with Visual Molecular Dynamics (VMD), which was really fun. I tried implementing the CCD algorithm but while I understand the theory and math behind it, I had some trouble with the difference between the math in theory and in practice. On Friday, I picked up the Introduction to Protein Structure book by Branden and Tooze to read over the weekend.


Week 2 (May 30 - June 5)

Tiring week. I spent most of Tuesday evening moving all my junk from my apartment to my house, and the Houston heat did not help. I finished reading the first part of the protein structure book, which dealt with different motifs. I wished I had picked up one that had more on protein folding, but I did find out lots more about how alpha helices and beta sheets join together. I also read through Amarda's thesis, which cleared up most the questions I had last week. I also read about other IK methods, particularly random tweak, so I could decide which is better for protein chain determination. I saw more of the math behind the methods and rather than mostly theory like last week, I saw some of the proteins generated through IK and how they compared to the native conformation. I am trying to find out more about conjugate gradient descent.


Week 3 (June 6 - June 12)

I spent most of the week researching a bunch of different techniques and tools: steepest descent, simulated annealing, conjugate gradient descent, Monte Carlo with the Metropolis criterion, Ramachandran histograms, rotamer libraries, energy forcefields, and Molecular Dynamic (MD) simulations. I also found out that my first project will be integrating rotamer libraries into Amarda's protein structure determination code. With all the different libraries out there, I am looking into the history and trying to figure out which one will work best given our objective.

Also, I got my own office Friday! Well, technically, I am borrowing a graduate student's office for the summer. In any case, it's great to have that kind of private space, and it's very close to Amarda's office.

Finally, some of the other undergraduates in my lab are participating in the W.M. Keck Center Undergraduate Research Training Program, which focuses on computational and structural biology. While I am not an official part of the program, Dr. Kavraki and Lisa Blinn (the program director) were happy that I decided to join them on the site visits they have every Friday to tour research and clinical facilities. We visited UH (University of Houston) today and heard about the research of eight different professors! Interesting stuff too - protein structure, human physiology, signal processing, etc.


Week 4 (June 13 - June 19)

I did some more research on rotamer libraries, including their background, the different types, the factors used to develop them, their uses, etc. Given my project, I also researched specific libraries, mostly that of Lovell et. al. and that of Dunbrack et. al.. I also spent some time preparing a couple of write-ups on rotamer libraries in general and on how they are used in side-chain modeling on a given backbone, developing some algorithms. From my software engineering class this past spring, I am glad I have finally learned the importance of the design phase =) For my first step in summer coding, I have downloaded the docking code from the group repository to look over, but it is in CVS. I would rather use SVN, but it should be okay. Hopefully, we can transfer the code to SVN before I have to start committing; otherwise, I can just see myself accidentally locking it or messing up the commits...

On a side note, Dr. Kavraki asked me to create a wiki for the group so we could have a common place for everyone to post announcements. I created one using mediawiki (with lots of help from Allison, another graduate student in the lab) and put up some basic security. I had never installed programs on a web server before, and I have only very rarely used wiki. In the end, though, it was a lot easier than I expected, but again, that's thanks to Allison, who installed all the system requirements beforehand and was willing to put up with my bombardment of e-mail questions. If we start using it for more, I might put up something more personalized, but for now, I am pretty satisfied with mediawiki's standard look.

The Keck group went to MD Anderson Cancer Center today. We got to see where our research goes "down the line", i.e. pharmacophore development and clinical trials. The trip was not as exciting as last week despite the biological implications, but it was interesting to see a different (non-academia) type of research setting.


Week 5 (June 20 - June 26)

I went over the code most of the week and wrote a small routine to familiarize myself with the molecule package and ccd. Also, we have decided to focus on motion prediction instead. That is, what conformations does a protein go through from a given starting structure to a given ending structure. Towards this end, I have studied up more on principal component analysis (PCA), which basically captures the major motions of a protein through eigenvectors and singular value decomposition.

We headed to the Human Genome Sequencing Center for Keck this week. I love genetics, so it was really interesting. It brought back some fond memories of high school biology. Also, they use robots (very awesome) and have a decent-sized cluster to run computational analysis of the sequences they find.


Week 6 (June 27 - July 3)

Over the weekend Amarda and I ran some tests and discovered that the starting and ending conformations we wanted to use were too similar to run any sort of motion analysis. Rather than using xtal structures, we decided to use PCA analysis of the protein's conformational space. Other group members have already run PCA, and using the eigenvectors they have generated, I have been able to model the closing motion of the flaps in HIV-protease and view this in VMD. Other than the code, I spent some time picking the features I wanted to use for ccd closure and figuring out how to keep the rest of the protein native-like. The assumption is that since most of the motion is along the flaps, keeping the rest of the protein rigid while moving the flaps along the PCA should still produce physical conformations.


Week 7 (July 4 - July 10)

I spent most of this week fixing up the code (more modular) and analyzing the output. Just moving along the PCA could produce conformations that are energetically unfeasible, so some randomization of the loops must be performed to generate a "cloud" of conformations. Also, we tried to see what would be a reasonable energy cutoff and PCA step size.


Week 8 (July 11 - July 17)

We tried to more quantitative this week rather than relying too much on VMD. Looking at rmsd outputs, we could tell that most of the motion is indeed in the flaps. We also tried different damping constants and integrated Amarda's minimization code. At first, it took about seven minutes to generate one conformation, though, and since we hope to generate hundreds (if not a couple thousand) conformations, I also had to go in and see where the bottleneck was. In the end, I shaved off two minutes.

On another note, I really hate rainy Houston...


Week 9 (July 18 - July 24)

As I mentioned before, just moving along the PCA will eventually cause the conformational energy to be above a given energy cutoff. Randomization has not been helping as much as I hoped because the protein keeps trying to move against the same energy barrier. Instead, we are going to try and backtrack to a previous physical conformation and try again from there. Also, I have started multiple runs of the program and will hopefully be able to identify some trends next week. (Where do conformations tend to gather? How much does redoing CCD vs. randomizing the protein help?)

We finally had another Keck outing this week, this time to attend a Pharmocoinformatics Overview at the UT Health Science Center Graduate School of Biomedical Sciences. There were tons of presentations (12 in all), some directly related to what I have been doing. I liked the exposure to the drug-development side of bioinformatics, and this was definitely one of the best presentations of URTP so far.


Week 10 (July 25 - July 31)

Final week before I head off to Seattle and Victoria for family vacation next week! Spent most of the week finishing up some experimentation and tying up any loose ends with Amarda. I also spent some time with MATLAB making graphs so I could visualize my output in something other than numbers.

For Keck this week, we went to see the supercomputers at the UT Health Science Center for Health Information Sciences. Pretty cool stuff, especially the virtual reality demonstration. We were able to manipulate protein structures in 3D VMD and simulate docking with a haptic device.


after Week 10

Wow, vacation was awesome! The mountain atmosphere really does something for you - very refreshing. I got back on Sunday morning (Aug 7) and spent my final free days making a poster and powerpoint presentation for the Keck URTP mini-symposium Aug 12. It was nice getting to see all the URTP people and their summer work. On a side note, I moved back to Sid Rich on Aug 10 to prepare for Rice O-Week and have had my hands pretty full with Peer Academic Advisor responsibilities. I also finished up my final report during the week, and it is now Aug 21, which means school starts tomorrow! It has been a great summer, though, and I am glad I get to continue my work this coming semester.