Structural Analysis of the v4 eVLP Gag-3xNES-Cas9 Fusion Protein by AlphaFold3: Per-Residue Confidence, Domain-Level Stability, and Predicted Aligned Error Matrices

This dataset contains all AlphaFold3 (AF3) structural prediction data underlying the structural analysis reported in Section 3.5 and Supplementary Figures S1-S5 of the manuscript. Four protein constructs were predicted using AF3: 1. SpCas9 (Streptococcus pyogenes Cas9, UniProt Q99ZW2, 1,368 aa) — isolated structure 2. MMLV Gag (Moloney murine leukemia virus Gag polyprotein, UniProt P03332, 538 aa) — isolated structure 3. FMLV Gag (Friend murine leukemia virus Gag polyprotein, UniProt P26806, 538 aa) — isolated structure 4. v4 eVLP Fusion (Gag-3xNES-MMLV_PR-Cas9, 1,977 aa) — full-length fusion construct For each construct, AF3 Model 0 (ranked highest by AF3) was used for all analyses. The dataset includes: - AF3 input sequences (FASTA format) - Domain boundary definitions (16 functional domains spanning the 1,977-aa fusion protein) - Per-residue pLDDT confidence scores (Model 0 and all 5 models) - Domain-level summary statistics (mean/min/max pLDDT per domain) - Predicted Aligned Error (PAE) matrix for the fusion protein (1,977 x 1,977, symmetrized) - Domain-averaged PAE matrix (8 x 8 functional domain pairs) - Per-residue delta pLDDT (fusion minus isolated) for SpCas9, MMLV Gag, and FMLV Gag - C-alpha backbone coordinates with pLDDT coloring for SpCas9 and the fusion protein - AF3 Model 0 CIF structure files for all four constructs - AF3 summary confidence metrics (ptm, iptm, ranking_score, etc.) - Rendered supplementary figures (S1-S5) in PNG and PDF format All predictions were generated using AlphaFold3 (Google DeepMind, 2024) via the AlphaFold Server. Per-residue confidence scores (pLDDT) were extracted from the B_iso_or_equiv field of CA atoms in the CIF files. PAE matrices were extracted from the AF3 full_data JSON output and symmetrized as (PAE + PAE^T) / 2.