absplit

(c) 2021 UCL, Andrew C.R. Martin

Code to take an antibody PDB file and split it into separate Fvs

Known issues

1. The code needs to retain the PDB header for downstream steps
2. Code needs to apply symmetry operators (e.g. 1b0w, 2rhe)
   • decided to do this first with pdbsymm
3. Some truncated sequences won't score highly enough to be flagged as antibodies - just score over the aligned region.
4. Crystal packing sometimes identified as a complex. e.g. 1shm_1

Algorithm

1. Read the PDB file and split into chains
2. For each chain, scan with a library of VH and VL domains to locate these. This is done by alignment and the standard interface positions are recorded.
3. For each VH/VL domain, find the CoG
4. For each VH/VL domain measure the CofG distance to all others. If within a cutoff, then this is a potential VH/VL pair.
5. For each potential VH/VL pair (note that these can be VL/VL or VH/VL and may be within the same chain) check the standard interface residues (assigned from the alignment) and ensure that they are close together
6. Add back HETATMs (other than water)
7. We now have assigned VH/VL pairs, so check those against other chains to look for antibody/antigen interactions. Note that an 'antigen' may be a different antibody.
8. Now check the HETATMs for HET antigens
9. Output the VH/VL domains together with any contacted antigen