GPCR structure analysis

This directory contains the data and scripts used for the structural analysis in the context of the paper Multiplexed selectivity screening of anti-GPCR antibodies.

METHODS AND MATERIALS

1. The raw input data is stored under data/221107_raw_data.csv. We replace the entry

253,delta,HCAR2;HCAR3,Q8TDS4;P49019,ENSG00000182782;ENSG00000255398,TRUE,TRUE,FALSE,,51,NRCLQRKMTGEPDNNRSTSVELTGDPNKTRGAPEALMANSGEPWSPSYLGP

with

253,delta,HCAR2,Q8TDS4,ENSG00000182782,TRUE,TRUE,FALSE,,51,NRCLQRKMTGEPDNNRSTSVELTGDPNKTRGAPEALMANSGEPWSPSYLGP

253,delta,HCAR3,P49019,ENSG00000255398,TRUE,TRUE,FALSE,,51,NRCLQRKMTGEPDNNRSTSVELTGDPNKTRGAPEALMANSGEPWSPSYLGP

to ensure that we have a unique uniprot id in every row. We store the resulting data as data.csv which will be used in the subsequent analysis. The processed data is stored under analysis/data_results.csv.

2. From https://alphafold.ebi.ac.uk/download we download all structures under human proteom (--2022-11-11 15:08:40-- https://ftp.ebi.ac.uk/pub/databases/alphafold/latest/UP000005640_9606_HUMAN_v4.tar). We then select the relevant models by uniprot id and store them under data/model_pdb.
3. Next we compute an alignment of the antigen sequence (Antigen_Sequence) with the protein sequence from the uniprot id (Uniprot_ID). When computing the alignment, identical characters are given 1 points, 0.5 points are deducted for each non-identical character, 1 points are deducted when opening a gap, and 0.5 points are deducted when extending it. We normalize this score by the length of the antigen sequence and save it as normalized_alignment_score. In particular, in case of a perfect match the score is 1. We also save the aligned subsequence of the protein as aligned_subsequence.
4. Next we analyze the region in the alphafold pdb file defined by the aligned_subsequence computed in the previous step. We compute:
   • the average plDDT score (average_plDDT_score) (that is, we sum up the plDDT scores for all residues in the subsequence and divide the result by the number of residues in the subsequence)
   • the number of plDDT scores below 50.0 divided by the total number of residues in the aligned subsequence (plDDT_below_50)
   • the average relative ASA (accessible solvent area) score (average_relative_ASA) (that is, we sum up the relative ASA scores for all residues in the subsequence and divide the result by the number of residues in the subsequence)
   • the number of relative ASA scores below 0.25 divided by the total number of residues in the aligned subsequence (relative_ASA_below_025)
   • the secondary structure (by 3-state characters, see DSSP below) (secondary_structure)
   • the number of states H in the secodary structure divided by the the total number of residues in the aligned subsequence (relative_H-dssp) (and similar scores for the states E and C)
5. For the final analysis and for producing figures we only use entries with a normalized_alignment_score greater or equal to 0.90 (note that a score of 1 means a perfect match)

DSSP

• we choose accessible surface area (ASA) values according to Sander and Rost, 1994, https://doi.org/10.1002/prot.340200303
• see also https://biopython.org/docs/1.75/api/Bio.PDB.DSSP.html
• for installation of dssp see: https://ssbio.readthedocs.io/en/latest/instructions/dssp.html

8-letter character	3-state character	Structure
H	H	Alpha helix (3-12)
B	E	Isolated beta-bridge residue
E	E	Strand
G	H	3-10 helix
I	H	Pi helix
T	C	Turn
S	C	Bend
-	C	None