Computational characterization of the uncharacterized human protein C17orf58 identifies a PCOLCE-like NTR domain and predicts a C-terminal mode of metalloproteinase binding

Abstract

A substantial fraction of human proteins have no described function. C17orf58 (UniProt Q2M2W7; "UPF0450 protein C17orf58") is one of them: it has no experimental structure and no functional literature. Here I report a computational characterization of C17orf58 that combines structure prediction, structural homology search, evolutionary conservation, predicted protein-protein interactions and human population variation. Only the C-terminal half of the protein (residues 193-339) is predicted to fold, forming a netrin-like (NTR) domain stabilized by three disulfide bonds; the N-terminal half is predicted to be disordered. Three independent structure-comparison methods identify the NTR domain of procollagen C-endopeptidase enhancer 1 (PCOLCE1) as the closest structural relative, ahead of tissue inhibitors of metalloproteinases (TIMPs), while sequence-based searches detect no relatives outside C17orf58 orthologs. C17orf58 orthologs are found from cartilaginous fish to mammals; the six disulfide cysteines are 95-100% conserved, and conserved surface residues cluster into a patch that is a candidate binding site. AlphaFold 3 confidently predicts complexes of C17orf58 with the catalytic domains of MMP-3 and MMP-14 (ipTM 0.82-0.86), in which the C-terminal carboxylate of the final residue, Ile339, sits 2.2-2.4 angstroms from the catalytic zinc. The prediction is lost when the C-terminus is extended by six residues (ipTM 0.48), persists in the full-length mature protein, and is not reproduced by the NTR domain of PCOLCE1. A second NTR domain, from SFRP1, is also predicted to bind MMP-3 confidently, although without entering the active site, indicating that some of the predicted affinity may reflect a family-level bias of the model. Human variation data show no detectable constraint on the gene. These results provide a first structural and evolutionary description of C17orf58 and a specific, experimentally testable hypothesis: that C17orf58 binds matrix metalloproteinases through its free C-terminus. All conclusions are computational predictions and require experimental validation.