ABSTRACT
De novo protein structure modeling predicts protein structure from its amino acid sequence when no homologous protein structure is available and can potentially generate membrane protein structure models at low resolution (2-4 Å root mean square deviation [RMSD] from the native structure) for proteins below ~150 residues (1). High-resolution structural models (below 2 Å RMSD from the native structure) have been generated de novo for soluble proteins below ~80-120 residues (2-4). Because of limitations in the modeling accuracy of de novo protein structure prediction methods, homology modeling is still the most useful approach for structure modeling of ion channels. Since determination of the rst x-ray structure of KcsA channel by Roderick MacKinnon’s group in 1998 (5), a number of new ion channel structures have become available. All of these ion channel structures provide valuable templates for modeling
of homologous ion channels. However, we are still limited in a number of conformational states of ion channels captured at high resolution. Even when closed and/or open states of dierent homologous ion channels are available, they may not be representing structures of distantly homologous ion channels. For example, open states of MthK (6), KcsA (7), KvAP (8), and Kv1.2 (9,10) channels show a bend of the pore-lining helix at a dierent position. erefore, experimentally derived constraints have to be used to explore multiple states simply of structurally exible regions of a specic isoform of an ion channel.
