Vaccine design, Epitope Atlasses

Claus Lundegaard (lunde@cbs.dtu.dk) and Morten Nielsen (mniel@cbs.dtu.dk)

Overview

During this exercise you will:

1. Make atlasses showing epitopes of two genomes
2. Zoom in on interesting areas
3. Make proteasome predictions for a short peptide contstruct
4. Try to optimise the construct manually and evaluate it using the predictions

Purpose of exercise, description of data

The purpose of this exercise is to show a visualization of the prediction tools you have seen during the course and let you use them in a small optimization problem. We vil use two organisms to visualize the predictions: The bacteria Richettsia prowazekii (typhus) and Hepatitis C virus.

Change to this afternoons working directory:

cd exercise8

Richettsia prowazeki

You will use the program genewiz to make the atlasses and a number of configuration files. Start with an atlas of the complete genome of Richettsia prowazeki, 1111523 bp. The file rpxx.epitopeatlas.cf is the configuration file that GeneWiz will use to generate a nice looking epitope atlas. We will not go into details with this file just be sure you have it.

ls -l rpxx.epitopeatlas.cf

genewiz -p rpxx_whole.ps rpxx.epitopeatlas.cf

pdfatlas rpxx_whole

1. Predicted strong binding subpeptides (Affinity<50 nM).
2. Predicted moderate binding subpeptides (Affinity<500 nM).
3. Predicted Proteasomal Cleavage

genewiz -p rpxx_zoom.ps rpxx.lin.epitopeatlas.cf

pdfatlas rpxx_zoom

nedit rpxx.lin.epitopeatlas.cf &

genewiz -p rpxx_zoom_new.ps rpxx.lin.epitopeatlas_new.cf

pdfatlas rpxx_zoom_new

Hepatitis C Virus (HCV)

Make an HCV atlas and take a look, this atlas is much faster to make only 9609 bp.

genewiz -p HCV_lin.ps HCV.lin.epitopeatlas.cf

pdfatlas HCV_lin 

Now we have only one coding strand, but we have included the variability, and a combined epitope score. The combined score is an attempt to integrate the predictions and the conservation of the sequence into one score. It should give a better idea of which epitopes that are the best cadidates.
Now zoom in on the first 2kb part of the genome:

nedit HCV.lin.zoom.epitopeatlas.cf

genewiz -p HCV.zoom.ps HCV.lin.zoom.epitopeatlas.cf

pdfatlas HCV.zoom

Polytope Construction

Take a look at the file

less predictions.tab

The format is a little unfriendly but the columns are:

1 Position in (poly)protein 
2 Amino acid at this position
3 Epitope (9-mer)
4 Name of protein in original genbank file
5 Predicted HLA-A2 affinity (log50k)
6 Predicted Proteasome cleavage (raw NetChop output)
7 Predicted signal peptide (raw SignalP output)
8 Geometric average of NetChop outputs within the 9mer
9 = 1
10  Number of sequenced in variability alignment
11 = 2
12 Variability expressed as Shannon entropy (Low conserved, High variable)

We now want to select potential epitopes from these predictions.

We are using the following thresholds:

MHC-affinity:    <50 nM  (>0.638)
Cleavage output  >0.8
Variability  < 1

The following command selects epitopes according to these criteria

gawk '{ if($5>0.638 && $6>0.8 && $12<1) print($3)}'  predictions.tab 

and this command will create a file with the three 9-mers separated by tabs:

gawk '{ if($5>0.638 && $6>0.8 && $12<1) printf("%s\t",$3)}END{printf("\n")}'  predictions.tab > epitopes.out

Now we want to create af FASTA file with the three epitopes with linker regions consisting of three alanines between the epitopes, and in the ends.

gawk '{print(">polytope\nAAA"$1"AAA"$2"AAA"$3"AAA\n")}'  epitopes.out > polytope.fasta

Look at the file:

cat polytope.fasta

NetCTL

A recently new server is the NetCTL server. This tool will evaluate both affinity, proteasomal cleavage, and TAP binding, resulting in a single score that can be sorted by.
Go to NetCTL.

Paste in your construct in the sequence window. Select the A2 supertype, and sort by combined score.

Evaluate the results:

Q6: Will there be cleavage at the C-terminals of the epitopes?
Q7: Is there any new fusion epitopes?

Push the back button and edit in the window the linker regions to optimize the polytope.

Do a new prediction and evaluate.

Continue untill a satisfactory result is obtained (this might not be possible).

Another tool is a polytope optimizer, that will generate the optimal linkers between epitopes. However, polytope is an older tool not considering TAP binding, and to minimize the calculation time only one NN is used for each of the cleavage and the binding predictions. Furthermore these NNs are only trained using sparse encoding.

We have made two files in the format used by the polytope program. poly1.list and poly2.list.

cat poly1.list

cat poly2.list

As you can see the epitopes are the same, but the order and linkers are different. Write down the sequences of the three epitopes for later use.

First we want to see which of these polytope that have the best construction. Right click on NetCTL to open a fresh server window. To get the polytope in copy-paste ready FASTA format use the following command:

cat poly1.list | grep -v # | gawk 'BEGIN {print">polytope_1"} {printf("%s",$2)} END {printf("\n")}'

Paste the polytope_1 header and sequence into the server window. Set the supertype to A2. Set the sort by score to sort by combined score. Set the weight on TAP to 0 (as the polytope program we use later do not take TAP into account). Run the prediction.
Identify your initial epitopes in the output.

Q8: What are the rankings and combined scores for your epitopes?
Do the same with the poly2.list.

Q9: What are the rankings and combined scores for your epitopes?
Q10: Why are the rankings different?

Now we would like to try to optimize one of the polytopes.

polytope_cont3 poly1.list   > poly1.out      

cat poly1.out

As you see, a lot of middle calculations are going to the screen, but finally it will end with a suggested new optimized polytope. If you scroll up a little you will see the polytope in a column format were the first column is the number of the residue, and the second column is the actual residue. Linker residues will be in lowercase. At the C-terminal of the epitopes is indicated the predicted strength of cleavage, and if there is any predicted cleavage sites inside the epitopes, it is indicated with a CWITHIN marker. Note that the actual prediction values for MHC binding and cleavage differ from those of NetCTL, since the predictions methods used in the polytope program are reduced versions of the NetCTL methods.

The sequence of the final polytope is also given on a line starting with:

"Best Energy" We can now make Atlases of our initial polytope and our final polytope to compare the two states.
For the initial polytope do:

plotatlas -b -el poly1.list poly1.out

pdfatlas poly

mv poly.pdf poly.init.pdf

plotatlas -el poly1.list poly1.out

pdfatlas poly

mv poly.pdf poly.final.pdf