From victor
Revision as of 12:16, 10 November 2014 by Layla (Talk | contribs)

Jump to: navigation, search

In the sample folder are available all input and output files used in this tutorial.
Consider also to visit the Features page for further examples.

Target/Template alignment

Supposing you have already found a template candidate, you need to align it against your target sequence. In this dummy example we take the sequences of two homologous proteins both endowed with 3D structure. That allows us to compare different type of Victor alignments with the "exact" one derived from the structural alignment.

The two proteins are:

  • Target = 2ANL (chain A)
  • Template = 1DP5 (chain A)

The two structure superimpose quite well (RMSD 2.03 A) considering the low level of sequence identity (28.06%). This is the resulting sequence alignment after the 3D alignment done by FATCAT:

3d align.png

Alignment cartoon.png


The subali application let you choose from very different type of algorithms, strategies and parameters. The fist step is to create a file (i.e. pair.fasta, already available in the sample folder) including both the target and template Fasta sequences together like that:

>2ANL:A Target
>1DP5:A Template

Sequence to sequence alignment

Supposing we call the input file with the target and template sequences pair.fasta than by running the following command you obtain a basic alignment with the default parameters (see Features):

subali --in pair.fasta

The resulting alignment is this:

Default align.png

Profile to profile alignment

Most of the time including evolutionary information helps improving the alignment quality. In this example we used PsiBlast to calculate profiles both for the target and the template sequences. The PsiBlast tool is available at:

The profiles have to be generated in a specific format.

When using the online service set the "Formatting options" specifying "Show -> Alignment as -> Plain text" and "Alignment View -> Flat query-anchored with letters for identities".

Instead, if you prefer to generate the input using the command line tool remember to use "-outfmt 4" ("6" in older versions of Blast, see Features). In our case the commands are:

psiblast -num_iterations 3 -db /db/blastdb/nr90 -query 2anl_A.fasta -out 2anl_A.psi -outfmt 4

psiblast -num_iterations 3 -db /db/blastdb/nr90 -query 1dp5_A.fasta -out 1dp5_A.psi -outfmt 4

The output files 2anl_A.psi and 1dp5_A.psi are provided in the samples folder.

Then to generate the alignment simply run:

subali --in pair.fasta --pro1 2anl_A.psi --pro2 1dp5_A.psi

Profile align.png

Evaluate 3D models

Based on the alignment created in the previous section we can easily model the target. In our case, for simplicity, we used the SwissModel online service. In the sample folder you can find two files:

  • model_default.pdb - The model obtained from the default sequence-to-sequence alignment.
  • model_profile.pdb - The model obtained from the profile-to-profile alignment.

Model default.png

By using the following commands for the two models:

frst -v -i model_default.pdb
frst -v -i model_profile.pdb

We obtain the following output (last line):

model_default.pdb	-29822.6749	-6266.6390	 -18.1340	-223.0000	 -46.3974
model_profile.pdb	-1549.1603	-3196.3570	  -1.1504	-236.0000	 -13.6659

Where numbers represent the following energies:

FRST | RAPDF | Solvation | Hydrogen | Torsion

For comparison, the experimental structure of the target obtains the following energies:

2ANL.pdb (Target)	-26691.5315	-8822.3390	 -23.3529	-224.0000	 -11.8836

According to these results the profile alignment is worse than the default alignment. Moreover the model generated from the default alignment appears to be more stable of the native structure of the Target protein (PDB id 2ANL), one explanation could be that SwissModel favours stability when generating structures. The general idea is that with Victor you can easily generate different alignments (changing algorithms and/or parameters) and you can effectively test them by evaluating the quality of the 3D models built from these alignments.

For an extensive discussion about these methods visit the References page.

Build loops

In the last section we will show you how to build a loop. In this example we take the 3DFR and try to model the loop of 5 residues from the position 89 to 93.

The first step is to generate a LUT (see Features) of size 5:

loboLUT_all -c 4

After that you should find in the data folder of the Victor package the following files: aa2.lt, aa3.lt, aa5.lt.

Now to model the loop simply do:

lobo -i 3DFR.pdb -c A -s 89 -e 93