Difference between revisions of "Gaucher Disease - Task 06 - Lab Journal"

From Bioinformatikpedia
(Multiple sequence alignment)
(1. Multiple sequence alignment)
Line 18: Line 18:
   
 
<!--
 
<!--
For our protein, the Pfam alignment of the only family found, [http://pfam.sanger.ac.uk/family/PF02055.11#tabview=tab3 Glyco_hydro_30 (PF02055)], contained only 1151 sequences, which is not enough for <code>freecontact</code>. Therefore, we used own alignments from task 2, as HHblits found 17,538 hit with default E-value cutoff and 3,189 hits with E-value cutoff of 10E-10 after 10 iterations (actually after 5). We took both alignment to compare the results, as the search with default E-value on the one hand has generated more alignments, which is an advantage for <code>freecontact</code>, but on the other hand some of those hits could have been false positives, as discussed.
+
For our protein, the Pfam alignment of the only family found, [http://pfam.sanger.ac.uk/family/PF02055.11#tabview=tab3 Glyco_hydro_30 (PF02055)], contained only 1151 sequences, which is not enough for <code>freecontact</code>. Therefore, we used own alignments from task 2, as HHblits found 17,538 hits with default E-value cutoff and 3,189 hits with E-value cutoff of 10E-10 after 10 iterations (actually after 5). We took both alignment to compare the results, as the search with default E-value on the one hand has generated more alignments, which is an advantage for <code>freecontact</code>, but on the other hand some of those hits could have been false positives, as discussed.
   
 
The alignments can be found in: <code>/mnt/home/student/kalemanovm/master_practical/Assignment6_Evolutionary_sequence_variation/P04062_ali/</code>
 
The alignments can be found in: <code>/mnt/home/student/kalemanovm/master_practical/Assignment6_Evolutionary_sequence_variation/P04062_ali/</code>

Revision as of 00:45, 28 August 2013

1. Multiple sequence alignment

For HRas, we downloaded the full MSA in FASTA format of the Pfam family Ras (PF00071). It contains 21,243 sequences. For the calculation of correlated mutations using freecontact, the MSA (ras.txt) had to be reformatted into the 'aln' format (ras.aln) with /usr/share/freecontact/a2m2aln like this:

/usr/share/freecontact/a2m2aln --query '^RASH_HUMAN/(\d+)' --quiet < ras.txt > ras.aln

The alignment is located here on the student server: /mnt/home/student/kalemanovm/master_practical/Assignment6_Evolutionary_sequence_variation/pfam_Ras_ali/ras.aln

For our protein, we also used the full MSA in FASTA format of the Pfam family Glyco_hydro_30 (PF02055), which contains 1151 sequences. The MSA was reformatted analogously:

/usr/share/freecontact/a2m2aln --query '^GLCM_HUMAN/(\d+)' --quiet < PF02055_full.txt > PF02055_full.aln

The alignment is located here on the student server: /mnt/home/student/kalemanovm/master_practical/Assignment6_Evolutionary_sequence_variation/pfam_Glyco_hydro_30_ali/PF02055_full.aln


2. Calculate and analyze correlated mutations

The following steps below are all described on HRas, but were done for Glucocerbrosidase the same. All used programs can be found in this directory /mnt/home/student/gerkej/gaucher/task6/ . The results for each protein in the corresponding subdirectories pfam_Ras/ and P04062/.

1. With the reformatted alignments the residue contacts where predicted with freecontact:

 freecontact --parprof evfold < ras.aln >  ras_contacts.out

2. All pairs with an smaller distance than 5 residues to its sequence neighbours were removed. The remaining pairs were ranked to its CN values.

python rank_contacts.py ras_contacts.out filtered_ras_contact.out

3. An Analysed of its distribution and range of scores was done by R.

4. All pairs of predicted and filtered contacts with a CN>1 were taken as high scoring pairs. These high scoring pairs were checked against the real contacts of the pdb file (HRas: 121p.pdb [1], Glucocerebrosidase: 1OGS.pdb).

python distance_check.py filtered_ras_contact.out 121p.pdb

The program stores the coordinates of all atoms documented in the pdb file. Then it calculates the distance of all high scoring pairs by using the euclidean distance. In case any atoms of two amino acids have a distance less than 5A the contact is right predicted. Otherwise, it is classified as FP. The resulting file contains all information about the high scoring pairs including its state (TP or FP).

5. To identifiy a possible correlation a Pearson Correlation was calculated between the scores and their state (TP/FP)in R.

values=data.frame(c(CN,state))
cor(values,method="pearson")

6. All high scoring pairs were visualised in a contact map with R (contact_map.R). First of all, the distances of the pdb file that are lower than 5Å were determined (by /pdb_distance_check.py). The pdb contacts which were not predicted as high scoring pairs (FN) are coloured lightblue in the contact map. The calculated high scoring pairs and their state can be seen in darkblue (TP) and red (FP).

7. After that the filtered_ras_contact.out file was parsed with calc_hotspot.py to calculate the hotspot residues. The program gets L (sequence length) high scoring pairs with the highest scores as input. From these pairs the program sums the scores of each residue and normalizes it with the average of all top L scores. All normalized residue scores are summarized in hot_spots.txt.

The EVcoupling server was run for both proteins to campare the 50 hot spot. The number of residues predicted as hotspots by both programs were counted to get the overlap of the predictions.