Difference between revisions of "Talk:Task 4: Homology-based structure prediction"
From Bioinformatikpedia
(Created page with "Well, I can see you have some bad luck with your protein, but still some comments so far: * You could have tried, what i-tasser does, if you exclude homologous templates (option…") |
|||
Line 5: | Line 5: | ||
[[User:Boidolj|Boidolj]] 17:03, 4 June 2012 (UTC) |
[[User:Boidolj|Boidolj]] 17:03, 4 June 2012 (UTC) |
||
+ | |||
+ | I must say i am rather surprised about your results, in addition to what Jonathan already told you above i have some other issues: |
||
+ | * if you couldnt find anything with Coma and HHPred, why dont you use another searchtool? |
||
+ | * why do you exclude the 99% identity candidate? If we assume the "I dont't know my proteins tructure"-case and you find a hit with 99% identity you would use it wouldnt't you? So why exclude a structure of a slightly changed protein especially if you dont have any other? |
||
+ | *why do you assume that it is a problem that your 1qs0 hit is from bacteria? I thought this Task is about finding conserved regions, because we assume they have the same structural conformation. And that would include cross-organism-hits as well from bacteria as from mammals or anything else |
||
+ | *why do you have doubts which model to choose for the 30% identity part? isn't the choice just simpel using the hit with the best e-Value? |
||
+ | *with your I-Tasser result i am quite surprised, because if i have a look at the models they seem to be very conserved and even the scores (apart from your tm-score) suggest a very good model. Especially the tm-score calculated by I-Tasser (0.69+-0.12) is very good |
||
+ | --[[User:Hollizeck|Hollizeck]] 21:27, 4 June 2012 (UTC) |
Revision as of 22:27, 4 June 2012
Well, I can see you have some bad luck with your protein, but still some comments so far:
- You could have tried, what i-tasser does, if you exclude homologous templates (option II, cut-off).
- Pictures of the models and maybe a table collecting the scores of different methods for the same template would give a better impression
Boidolj 17:03, 4 June 2012 (UTC)
I must say i am rather surprised about your results, in addition to what Jonathan already told you above i have some other issues:
- if you couldnt find anything with Coma and HHPred, why dont you use another searchtool?
- why do you exclude the 99% identity candidate? If we assume the "I dont't know my proteins tructure"-case and you find a hit with 99% identity you would use it wouldnt't you? So why exclude a structure of a slightly changed protein especially if you dont have any other?
- why do you assume that it is a problem that your 1qs0 hit is from bacteria? I thought this Task is about finding conserved regions, because we assume they have the same structural conformation. And that would include cross-organism-hits as well from bacteria as from mammals or anything else
- why do you have doubts which model to choose for the 30% identity part? isn't the choice just simpel using the hit with the best e-Value?
- with your I-Tasser result i am quite surprised, because if i have a look at the models they seem to be very conserved and even the scores (apart from your tm-score) suggest a very good model. Especially the tm-score calculated by I-Tasser (0.69+-0.12) is very good
--Hollizeck 21:27, 4 June 2012 (UTC)