I am designing diverse types of chimeric antigen receptors (CARs) and I would like to predict their tridimensional structure (that is to say, build PDB models). First, I tried to copy and paste their amino acid sequence in some model prediction software such as Phyre2 and Swiss-Model but, as they use a single template, the resulting model of my CAR consists of an aberrant, non-sense protein. Secondly, I tried to predict the structure of each domain of the CAR alone but including 5 amino acid of the preceding domain and some others of the following domain, in order to “concatenate” all of them by homology, but that failed too. How could I solve this problem without using crystallography?
"First, I tried to copy and paste their nucleotide sequence in some model prediction software such as Phyre2 and Swiss-Model"
You cannot simply submit the nucleotide sequence to any protein modeling program - they expect protein sequences, and interpret the nucleotides as Alanine, Threonine, Glycine and Cysteine!!!
First of all, translate your nucleotide sequence into a protein sequence.
"Secondly, I tried to predict the structure of each domain of the CAR alone but including 5 amino acid of the preceding domain and some others of the following domain, in order to “concatenate” all of them by homology, but that failed too."
The CARs are chimeric proteins (e.g. https://www.creative-biolabs.com/car-t/car-design-construction.htm?gclid=Cj0KCQiA88X_BRDUARIsACVMYD-jg3EGiBtsaVXeiu2pJn44z-975z8dMt7PGUueXobXxx0Aytz2Em4aAnl3EALw_wcB), therefore you are unlikely to find one template to cover the entire protein. Your approach to predict each domain separately is correct. However, depending on the exact construct, the molecule will contain flexible linker regions, which are unlikely to be resolved in an Xray structure. As a consequence, you cannot use overlap to assemble the domains. In addition, the transmembrane region is unlikely to be included in an X-ray structure, although there are some NMR structure of transmembrane helices.
If you use an scFv as antigen recognition unit, the best template may be a Fv or Fab fragment, where VL and VH are on different chains. Automated modeling programs may therefore not realize that the template structure is suitable for both domains as well as their relative orientation. Do a separate homology search of the pdb with the VL and the VH domain, then compare the results to see whether you find a structure that has high-ranked hit for both
Hi,
You can use advanced modelling (Modeller) to build the 3D structure with multiple PDB templates if you have good homology and query coverage. Since you are working on chimeric receptor construction, individually you can search templates and proceed with advanced modelling. when you are going to run the program at that time you can create a single target sequence file.
Hi. Thank for both of your replies.
I believe using advanced modelling (Modeller) is, to date, the best option in this case. But I have some problems when implementing the algorithm. This is what I’ve done:
1. I created a .ali file just with the amino acid sequence of the CAR
2. Then, I searched 2-3 templates (PDB files) per each domain (scFv, hinge…)
3. Run model-multiple.py file
And my problems are these:
1. I don’t know how to align the templates with my CAR sequence (.ali file)
2. The file model-multiple.py requires a code of the target (obviously, it doesn’t exists)
3. What happens with non-aligned amino acid?
What algorithm would you implement instead for the multiple template modelling, Pachineella Lakshmana Rao?
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography