I have a group of orthologous proteins as A A' A" B C D D' E F G H where all alphabets belong to different species; A A' A" and D D' are inparalogs. I want to remove the inparalogs from this group and only want keep only one of the inparalog that is much closer (similar) to other group members. The final group should be like A* B C D* E F G H (where A* and D* is much closer to other species sequences). I can do it manually be creating a phylogenetic tree subsequently reducing the inparalogs but the problem is that I have around 23-33 species and >1000 groups. Strictly, occurrence of one species is allowed only once per group.
Hi,
First of all, I am not sure what kinds of information you have right now. I just assumed that you have all the pairwise distance (i.e., number of difference) among these proteins.
Say for A, A' and A'', you could just pick the one with minimum sum of d(AB)+d(AC)+d(AD^). Where d(xy) is the distance between protein x and y. And D^ should be the one among D and D' that gives you the minimal d(AD^).
It is actually same as creating the phylogenetic tree, though. Because pairwise distance is really what you need to create the tree. But it is also what you need to solve this kind of question.
HTH
Dear Shishir,
I have read different answer here and on LinkedId post. I agreed with Jianrong Yang on the comment that what you needed to resolve your question is actually a phylogenetic tree, let me represent your example with a tree to give a better answer.
/-A
/-|
/-| \-A'
| |
/-| \-A"
| |
/-| \-B
| |
/-| \-C
| |
| | /-D
/-| \-|
| | \-D'
/-| |
| | \-E
/-| |
| | \-F
--| |
| \-G
|
\-H
EDIT: I include now the newick of the example since it's not well visualize: (((((((((A,A\'),A\"),B),C),(D,D\')),E),F),G),H);
Then, you can apply a metric like the one propose before to select among the specific duplications for A & D but, and here an example why I think you should use trees, the situation is different
/-A
/-|
/-| \-A'
| |
| \-D'
|
| /-A"
| /-|
| /-| \-B
| | |
--| /-| \-C
| | |
| /-| \-D
| | |
| /-| \-E
| | |
| /-| \-F
| | |
\-| \-G
|
\-H
EDIT: I include now the newick of the example since it's not well visualize: (((A,A\'),D'),(((((((A\",B),C),D),E),F),G),H));
On this case, you group or rorthologous sequences are the ones from the bottom subtree. How often could you face this situation? That depends on the scope of species you are working with but it's important to be aware of that. Remember that most accurate framework (e.g. Max. Likelihood & Bayesian) are not solely based on distances so it may be possibel on the second example that A & A' are quite similar to A" and you can not distinguish them without using a tree.
Hope this helps,
S
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology