You have to first ensure that all parts are in the correct relative position and orientation. Depending on the exact information you have, you can visually position them in a molecular graphics program, fit them against additional templates that may not have enough resolution for atomic models, but may yield a good positioning especially of helices, or use docking. If you have overlapping models that share terminal secondary structure elements, you can also use these to fit one segment on the next by a superposition of the overlapping parts.

You then either merge the different parts into a single object in your graphic modelling program or export the parts in the correct relative orientation as PDB files and cut-and-paste the parts into a single file in correct order in a text editor.

You then need to correct chain and residue labels to ensure each residue has an unique identifier and all residues belonging to the same protein have the same chain label.

Since main-chain bonds are normally recognized by distance criteria, your raw model will probably contain strand breaks - you can mend these by introducing an explicit bond in a modelling program and use local energy minimization to regularize bond lengths and angles (keeping all atoms fixed except for the residues flanking the splice site), or , if the splice site is in a flexible region, use loop modelling to close up the gap.

Once your raw model is a s good as you can make it (you may produce multiple models to test different hypotheses how things might fit together), you can start with careful packing optimization/energy minimization. Watch carefully where things move - this can highlight flaws in your raw model that may be better resolved by remodelling parts of the structure than by brute force minimization!

I agree with Annemarie Honegger.

I have faced same problem for my transmembrane protein. As per your description (Joel) I think you don't have suitable tamplate for your protein and blast results also poor. It's a good idea to generate multiple models and stich them together, but I think in this process you may miss some correct folding and arrangement which occurs due to the interdomain interactions. As modeling of protein not always gives you correct answers if you don't have correct templates for your protein. I don't know much about your protein. Still I think you may use threading approach (itasser) to build complete protein. Then use itasser model as a template and take other relevant pdb structures and do multitemplate modeling using modeller. If you like you can see our publication where we took similar approach to model one of a ABC transporter.

https://www.tandfonline.com/doi/abs/10.1080/07391102.2017.1372311

With modeller is possible to use one or more plus PDB file as template to generate the final structure of the pdb. but i have read that you don't want this option. ok in this case you can use modeller to attach single portion to your final model. An example

1. first pdb model N term

2. second pdb model TM helix

3. third pdb C-term model

you can create file .PIR to align these portion to get the final coordinate.

I'm attached you an example.

let me know it is useful

obviuosly if necessary , you make a refinment of final protein

best

andrea

Hello Anne and Krish,

thanks for your kind replies and I have been successful, and yes my templates that I used had to be hand threaded due to their very very

distant homology and therefore very low identity. Accordingly I stitched

the models together via Pymol visualization and Modeller using a multi-

template approach where my PDBs and alignment were my of my own.

Thanks again for your input:)

Joel

Hello Andrea and thank you for your update and I have been successful (my last email overlapped yours):)

good

best

Thanks;)