Several antibody-antigen docking problems

HADDOCK
1

Hello, I am a rookie in antibody-antigen docking, I would appreciate it if you could give some suggestions;
HADDOCK is an excellent job that helps me a lot, and I was troubled by some problems when referring to the manual(HADDOCK2.4 Antibody - Antigen tutorial using PDB-tools webserver – Bonvin Lab).

  1. In the step of " Using ProABC-2 to identify the paratope": when using full length of antibody sequence as input, ProABC-2 offers the Fv domain of heavy chain and light chain separately, which residues should be part of the paratope? Is there a standard for reaching a consensus?
  2. There are different ways to prepare the structure for HADDOCK input; Using pdb-tools web server and run.py script from " [haddocking]HADDOCK-antibody-antigen**" in local could get different results: pdb-tools web server and local also get some difference, so I want to know which one to choose.
  3. Is there a difference between using the structure of full-length and Fv parts as input for paratope prediction？ Will it affect the docking reslut？

Thank you in advance!

2

Dear Yi

  1. In the step of " Using ProABC-2 to identify the paratope": when using full length of antibody sequence as input, ProABC-2 offers the Fv domain of heavy chain and light chain separately, which residues should be part of the paratope? Is there a standard for reaching a consensus?

ProABC should identify the 6 hypervariable loops. Make sure that the residue number match the numbering of your processed PDB file (step 2 below).

  1. There are different ways to prepare the structure for HADDOCK input; Using pdb-tools web server and run.py script from " [haddocking]HADDOCK-antibody-antigen**" in local could get different results: pdb-tools web server and local also get some difference, so I want to know which one to choose.

Those tools will not change the coordinates. So in principle both should be fine.
What kind of differences are you speaking about.

  1. Is there a difference between using the structure of full-length and Fv parts as input for paratope prediction？ Will it affect the docking reslut？

Waste of CPU resources if using the full length - you target anyway the hypervariable loops.

3

Thank you so much for your clear reply.
The difference is the length of chain, 3 residues at the end may be lost in my case. But I suspect it might be my problem. I’ll test it again.

Another question is “From the results of ProABC-2 we see that the variable domains of both light and heavy chain represent roughly the first 110-120 residues of each chain”. I learned that “At the protein sequence level, the light (25kDa) and heavy (50kDa) chains are comprised of two and four sequence homology regions of ~110 amino acid residues, respectively”, so how should I extract the accurate Fv seq from the full sequence? first 110 or first 120 or case by case?

4

Look at the structures. It should become pretty clear where to cut (at the domain boundaries)

5

If the structure is not available, what should I do?

6

How can you model a complex without the structure of the antibody…? Haddock does require an input 3D model

7

There is a situation in which the structure is not available, so I need to model it. The method of simulating structure is limited to the Fv domain. I have learned that there do have methods to extract the Fv seq from the full-length sequence.
Thank you all the same.

1 Like
8

Hello, it’s me again. :grin:

  1. As we mentioned in the email, we calculate all contacts within a given cutoff to determine the contact residues, so what is the empirical cutoff generally speaking? 9Å is a loose definition? which cutoff is solid?

  2. There are the results from contact-chainID, does each column represent “residues num in first chain/chain id/atom/residues number in the second chain/chain id/atom/distance(Å)”? The question is the length of my chain A is 115 amino acids. If not, how should I get the position of the residues in the interaction surface from a PDB file so that I can use them as active residues? Could you give me any suggestions?

18 A N 33 B N 28.035966
18 A N 33 B CA 27.980521
18 A N 33 B C 26.735026
18 A N 33 B O 25.950225
18 A N 33 B CB 28.123526
18 A CA 33 B N 26.776857
18 A CA 33 B CA 26.745557
...
132 A CB 146 B N 48.483703
132 A CB 146 B CA 49.626985
132 A CB 146 B C 50.998842
132 A CB 146 B O 51.154802
132 A CB 146 B CB 49.664195

Thank you so much!

9

Often a 3.9A cutoff is used to detect contacts between heavy atoms. (e.g. this is what is used in LIGPLOT/DIMPLOT).

In CAPRI the fraction of native contacts is calculated using a 5A cutoff (heavy atoms)

10

Got it; Thank you.