Methanotrophic bacteria play a critical role in biological methane mitigation, However their functional potential detection in engineered systems remains challenging by the high sequence similarity between particulate methane monooxygenase (pMMO) and ammonia monooxygenase (AMO). In wastewater treatment, where both methanotrophs and ammonia oxidizers co-occur, this ambiguity undermines efforts to quantify methane oxidation capacity from metagenomic data alone. We have recovered metagenome-assembled genomes (MAGs) from nitrification and anammox reactors (Klaghamn and Gryaab WWTPs, Sweden), with confirmed pmoA/amoA hits across multiple bins. Sequence identity and phylogenetic placement alone are insufficient to resolve substrate specificity for these sequences, representing a core unresolved question in our study of the "methane paradox" in engineered biofilms. Here we propose to apply GPU-accelerated protein structure prediction (ColabFold/AlphaFold2-Multimer) on Alvis to generate structural models of MAG-derived pMMO and AMO subunit complexes, with specific focus on active site copper-coordinating residues in the pmoB/amoB subunit. Predicted structures will be benchmarked against experimentally resolved pMMO structures (PDB: 3RGB, 7RPZ) and the best available AMO reference models, using TM-score and active site RMSD as functional discrimination metrics.