Journal

Week 2 (9-13 Jun)
Bugs are always annoying, but they are even more so when they take a while to fix. Unfortunately it seems that I chose the wrong occupation, because research is certainly not bug-free. :-) After fixing a memory leak in our code, I began the task looking at the energy/distance distributions of several folding landscapes. Thanks to the people who wrote Matlab, at least the plots are easy to produce.

Week 3 (16-20 Jun)
The biggest event of the week was the arrival of my roommate, Olga. She is also in the DMP program, and is working with Dr. Amato. After I introduced her to the lab, I found myself buried under a pile of papers to read. I guess part of the whole research thing is keeping up on other peoples' work. But paper reading was not the only task for the week. I also spent time on fixing various problems with our code and optimizing certain processes.

Week 4 (23-26 Jun)
Yippy, we're getting close to result time. I have the code all working now, and I just implemented a new filter to select local energy minima from the complete enumeration of the possible RNA conformations. Now I get to see how well just those conformations describe the folding landscape (or roadmap). I'll do this by comparing our results from the complete landscape to the results from the filtered landscape.

Week 5 (1-3 Jul)
It's hard to get much done when I'm gone for a long weekend, and we have the 4th of Jul off. Still, I managed to get an abstract of my research written. I also generated the results I mentioned last week. As is the case with research, the results illuminated some areas of our model that need work. Selecting only the local minima and connecting them with the shortest path to the native state makes for a roadmap with too sparse of connections. Xinyu and I have been discussing ways of 'reconnecting' the selected conformations to each other via paths in the completely enumerated roadmap.

Week 6 (7-11 Jul)
Aside from fixing minor problems with the code, I confirmed that our completely enumerated roadmap was producing comparable results to other models. These results were quite exciting and helped spur a burst of activity. We began working working on the outline to our conference paper. We also continued working on our filter methods. We randomly selected conformations from the completely enumerated roadmap and connect them with the appropriate edges from the same roadmap. Again, we ran into the problem of not having dense enough connections. We noticed this because our eigen-value solver will not converge to a solution on data sets that are too sparse. To confirm that the it was indeed a lack of connections causing the problems, I ran some experiments that each time stepped up the number of conformations that were selected. When I selected less than ~70% of the conformations, we could not get our results to converge. For the roadmaps with at least 70% of the conformations, convergence occurred.

Week 7 (14-18 Jul)
This week was paper week. Now that we have an outline for our paper, it was my task to take the material we had already written and begin putting it into the new outline. This meant copying, pasting, and rewriting. Guang Song, one of the PhD students I worked with last summer, also gave us his protein folding dissertation. Between writing and reading all that, we managed to discover a great probabilistic estimate of the size of a given conformation space. This will be quite useful for our conformation generation methods. Finally, we also began working on better reconnection strategies. This involves implementing both a k-closest and a radius connection strategies.

Week 8 (21-25 Jul)
After giving Guang feedback on his dissertation, Xinyu and I came up with a very ambitious time table for the rest of our summer. The next 4 weeks are going to be packed. We are planning to generate results, write two papers, and give two talks. The first of these talks will be the one that I give at the local Undergraduate Summer Research Symposium. Along with that, I will submit a paper that details our research. The second presentation will be given to the Parasol group. Finally, we hope to submit a paper to RECOMB 2004. That's not all, we are working on coding and debugging our reconnection methods. This process is sometimes frustrating, because just as soon as I thought I have the code debugged, I discovered another bug.

Week 9 (28 Jul - 1 Aug)
Time is getting short. I have only three weeks left in College Station. We are actually keeping to our time table fairly well, but it is tight. This week Xinyu explained the three different edge weight options that we can choose from. This discussion was very helpful, because it illuminated some previously vague areas in my understanding. My reconnection code is now quite debugged. It appears to be working the way we expect it to. I have to give my presentation at the beginning of next week, so I began working hard on both it and the paper. Revising is a pain, but it must be done.

Week 10 (4-8 Aug)
This week was a whirl of paper writing and presenting. Dr. Amato was quite good about requiring that all of us undergraduates practice our presentations in front of the Parasol group before we gave them at the Symposium. This was helpful for improving our explanations and our presentations. Once the presentation was over on Wednesday, I spent the rest of the week frantically rewriting and revising the paper I turned in on Friday. Xinyu was working on generating preliminary results this week while I was writing and presenting. We ended up with some great results that I had just enough time to squeeze into my paper before turning it in. I think I had a few too many late nights this week, but both the presentation and the paper turned out well.

Week 11 (11-15 Aug)
This week is my last, and it certainly is one of my busiest (though I'm not sure it could possibly beat last week). Between getting my oil changed, packing, and saying goodbye, Xinyu and I managed to get quite a bit done on our project. We were stalled a bit at the beginning of the week by an OS upgrade that our code would not run on. Once we got that straightened out, we began to generate and compare lots of results. This involved me writing a Perl script to do the comparisons and create Matlab plots of the data. This was fun, because Perl is my favorite language. It looks like we are actually going to end up meeting our ambitious time table of 4 weeks ago. We have some really exciting results to present and put in our RECOMB paper!

Wow! I can't believe that another summer is gone. I have had another wonderful time in College Station. I can honestly say that this summer was even better than the last. I have met many wonderful people, made great progress on RNA folding, and grown personally. Texas will always hold pleasent memories. Thanks to the Distributed Mentor Program and my advisor Dr. Nancy Amato!