Additional file 8: Selected structural and docking models

Contents
1. Sequence_Inputs
   - T3-I1_consensus_sequences.fasta: T3 and I1 mitochondrial consensus protein sequences used for AF3 prediction.
   - nuclear_encoded_reference_sequences.fasta: nuclear-encoded reference protein sequences used in all T3 and I1 comparisons.
2. AF3_Selected_Models
   - 12 selected mitochondrially encoded AF3 monomer models (six proteins for each haplogroup).
   - 9 selected nuclear-encoded AF3 monomer models shared across haplogroups.
   - Each selected model is supplied as mmCIF together with its AF3 summary-confidence JSON file.
3. HADDOCK_Selected_Results
   - Exact prepared input PDB structures, ambiguous interaction restraint files, and HADDOCK configuration files for 22 docking runs (11 interfaces x 2 haplogroups).
   - Cluster-level and model-level CAPRI/HADDOCK evaluation tables.
   - The top four representative PDB models from the highest-ranked FCC cluster in each run.
4. AF3_model_selection.csv and HADDOCK_selection_summary.csv
   - Explicit model and cluster selections corresponding to the manuscript methods and tables.
5. MANIFEST_SHA256.csv
   - File paths, byte sizes, and SHA-256 checksums for archive integrity verification.

Selection rules
- AF3: for each protein, the model with the highest ranking_score was selected, as reported in Tables 2 and S3.
- HADDOCK: FCC clustering used a 5.0-A contact-distance cutoff. The cluster ranked first by HADDOCK score was selected, and its first four representative models were retained.

Related supporting files
- Additional file 5 contains AF3 quality metrics for all predicted models.
- Additional file 7 contains the complete contact-map and PLIP results.
- Additional file 11 contains model-level HADDOCK source data and statistics for Fig. 4.

Model-path correspondence
- In Additional file 11 (Source Data 1), the Model entries are exact paths relative to the root of this archive. They point to the retained files under HADDOCK_Selected_Results/<interface>/<haplogroup>/Representative_Models/ and use each run's actual selected cluster ID (for example, cluster_2, cluster_3, cluster_5, cluster_8, cluster_10, or cluster_18), rather than a generic cluster_1 label.

These files are computational models and analysis outputs, not experimentally determined structures or independent biological replicates.
