Experimental-Reference Sensitivity of a Fixed Boltz-2 MPC1L–MPC2 Complex Prediction, with an Exploratory AlphaFold Structural Comparison

This Technical Note examines how the choice of experimental reference structure affects the structural agreement of one fixed Boltz-2 prediction of the human MPC1L–MPC2 complex. A Boltz-2 v2.1.1 Human PPI Atlas structure was compared with four cryo-EM structures—apo inward-open 9GIX and inhibitor-bound outward-open 9GIV, 9GIW, and 9GIY—using a common, experimentally defined 73-residue interface Cα mask and rigid-body alignment. The interface Cα RMSD was 3.853976 Å against 9GIX, compared with 0.964115 Å, 1.217374 Å, and 1.077057 Å against 9GIV, 9GIW, and 9GIY, respectively. All six experimental-reference comparisons, predefined mask and chain-alignment sensitivities, and qualified local structural descriptors provide context for this reference dependence. A separately sourced, fixed AlphaFold-derived complex was also evaluated using the same protocol as an exploratory comparison. The accompanying Supplementary Information and reproducibility package provide frozen structural inputs, complete numerical outputs, analysis and verification scripts, source provenance, and explicit methodological limitations. This is a fixed-coordinate case study, not a general benchmark of model accuracy. It does not establish an independent minimum scientific contribution, preferential biological states, inhibitor effects, or a new transport mechanism. Model inference was not rerun.