Weekly Journals

Week 5 ( June 29th ~ July 5th):

 

    The Biology for Modelers course continued this week. We learned more about cell-cell signaling and recombinant DNA. Amarda said she would look into the HMM and discussed further with Lydia about it. Later this week we had a meeting with Professor Cecilia Clementi from the Department of Computational Biology and Chemistry, who is working with Lydia at the project. She suggested a couple papers to me in chemistry that is related to transition path sampling. Even though they are pure physics based chemistry, the papers are elegant and they triggered new ideas in my mind for the protein-folding project.

 

 

Week 6 (July 6th ~ July 12th): 

 

After exploring all the possibilities to automate critical event detection and protein trajectories analysis, it seems that the algorithm for my project finally boils down to a combination of interatomic distance and contact maps. I am going to divide the project into two stages. First, I focus on the frames with critical protein conformation and secondly, extend the contact region detection into the protein trajectory level. My goal is to categorize these trajectories by comparing the contact information of different trajectories.

This week, I completed work on real contact maps. I observed that by finding the appropriate threshold to be around 0.1, the contact region in all the protein conformations under study could be recognized in the map. Some of the results are shown here. However, there is a fair amount of noise caused by adjacent neighbors. Therefore, I needed to set a window of size 4 in order to discard its 4 neighbors when considering if other atoms are in contact with this atom.

 

Week 7 (July 13th ~ June 19th):

 

Early this week, I generated the real contact map for all the critical
protein conformations from different trajectories. Unlike real contact vectors, which can take on any real number between zero and one, binary vectors can only consist of zeros and ones. Consequently, the binary maps contain less noise than the real maps. 

Since as the generated results indicated, the binary map is informative enough to distinguish the contact regions apart from the non-contact regions. After the discussion with Lydia, we decided to use binary contact maps from now on for my project.

Compared to the real contact map, the challenges that arose from using the binary contact map was that it was more difficult to find a threshold that will be able to distinguish all different critical conformations. After numerous experiments and statistical analysis, I figured out that the ideal ratio of cutoff to windowSize for every binary map in any trajectory is approximately 0.1. But for that unify cutoff to windowSize ratio, the value needs to be adjusted lower to 0.6/7. The models under study are simplified such that the interatomic distance
of neighboring atoms is a constant, so that if a model whose interatomic distance varies by an additional parameter to the.cutoff to windowSize formula, then a constant ratio will still be achieved.

 

 

Previous Page                                                                                                 Next Page