Design Ligand Induced Protein Oligomer

Small molecule-regulated protein oligomerization provides a powerful mechanism for manipulating biological processes by controlling protein proximity with high temporal precision. However, such systems only rarely exist in nature and remain a significant challenge for de novo design. This repo contains the necessory codes and scripts to for designing protein homooligomers whose assembly is regulated by small molecule ligands with matching symmetry. The C3-symmetric drug amantadine (AMA) is used as the target small molecule here.

Key innovations include: Homooligomer scaffolds generation: Build symmetric (homomeric) protein architectures around small-molecule ligands with matching symmetric, thus endowing ligand responsiveness to the protein complex. Heterodimer protein generation: Integration of RFdiffusion, Biased-ProteinMPNN to generate heterodimers from homotrimeric complex. Sequence-structure co-optimization: Combine Rosetta FastDesign, ProteinMPNN, and AlphaFold2 for high-confidence designs. Validation-ready workflows: Scripts to assess protein-ligand interaction, protein-protein interaction, hydrogen-bond network check, interface shape complementarity.

Design Ligand Induced Homooligomer

Scaffold Generation

ProBuilder can generate symmetrical protein scaffolds around small molecule symmetric axis (e.g., C3-symmetric homotrimers). Example: To design an AMA-induced homotrimer:

./inducible_homooligomer_generator -use_ss -motif_pos 1 -num_repeats 1 -rpx_db rpx_cart2.0_ang26.0_ss_ALLAA.phmap -rpx_cart_resl 2.0 -rpx_ang_resl 26.0 -symmetry C3 -rif_table AMA_DNEQ_rotated_rif_polar_bonus_-3_20220529_cart1.0_ang16.0_bound512.phmap -rif_cart_resl 1.0 -rif_ang_resl 16.0 -context_pdb AMA.pdb -len 65 -output_prefix WU65aa3H01173_ -nstruct 100 -ss HHHHHHHHHHHHHHHHHHHHHHLLLLHHHHHHHHHHHHHHLLLLLHHHHHHHHHHHHHHHHHHHH -motif_pdb ./scaffolds/WU65aa3H01173.pdb -sampling_stage_cutoff_tolerance_factor 0.001 -gzip -output_dir output_folder -mcmc_inner_cycles 10000 -mcmc_outer_cycles 10 -extra_ligand -1_-12_20 -sidechain_neighbor_cutoff 0.5 -context_clash_radius 1.5 -CB_swelling_factor 1.25 -rif_score_cutoff -240 -rif_weight 20 -inter_chain_score_cutoff -2.0 -ligand_neighbors_cutoff 1.4 -ligand2_neighbors_cutoff 1.4

The meaning of each parameter can be referred to: ./AMA_induced_homotrimer_design/README.md Adjust *_cutoff parameters to balance scaffold diversity and quality (lower values increase output but may compromise stability).

Interface Sequence Design by Rosetta FastDesign

Rosetta FastDesign optimizes side chains at small-molecule-binding and protein-protein interfaces. The codes in 2_fdes folder identifies designable residues (e.g., those within 5 Å of the ligand or partner chains) and optimizes protein sequences using a energy function favoring polar contacts and interface shape complementarity.

Hydrogen-Bond Validation

The python scripts in 3_find_hbonds screens designs for critical small-molecule interactions (ligand-protein H-bonds). Designs with favorable with the amine group of AMA are selected for further optimization.

Sequence Optimization

The deep learning based protein sequence optimization method ProteinMPNN (4_mpnn) is used to redesign amino acid sequences of residues distal from AMA. This step is used to further improve the folding potential and expressibility while preserving small-molecule contacts.

Structural Validation

The structure prediction network AlphaFold2 (5_af2) is used to access the sequence-structure compatibility of the designs. We only predict the structure of the monomeric chain with AF2.

Design Ligand Induced Heterooligomer

Heterooligomer designs were generated from the AMA-responsive homotrimers. This highlights the high extensibility of our AMA-responsive homotrimer. This broadens the chemogenetic toolkit for manipulating complex biological processes with small molecule drugs.

The heterodimer is generated by fusing two neighboring chains of the homotrimer. Biased sequence optimization was applied via ProteinMPNN to promote heterodimer formation, while disrupt homotrimer formation. We introduce tied residues at a monomer unit both in heterodimer and homotrimer and add positive and negative bias to design proteins more likely to form heterodimer instead of homotrimer when adding AMA. Pos_factor of 1.0 and neg_factor of -0.35 is recommanded, though these parameters can be adjusted as needed.

Scoring scripts

We provided the scoring scripts for computing the Rosetta based scores (eg. contact molecular surface, ddg) and AF3 metrics. These scripts can be found in the subfolder. You need to have Rosetta, AF2 and AF3 installed, and the environment properly configurated to run these scripts.

Future Works

At present, our pipeline can be used to generate protein complexes whose assembly is regulated by symmetric small molecule ligands. We are working to generalize this 'docking-while-binding' concept to asymmetric small molecule ligands.

Acknowledgements

We used libraries or source code from the following packages:

• xbin module of rpxdock
• getpy