Domain identification by iterative analysis of error-scaled difference distance matrices.

Iterative interpretation of error-scaled difference distance matrices is suggested as a means of dividing a protein into structural domains on the basis of conformational differences between different models. Two conformers of Src kinase [PDB codes 1fmk [Xu et al. (1997). Nature (London), 385, 595-602] and 2src [Xu et al. (1999). Mol. Cell, 3, 629-638]] in the inactive state with and without a substrate analogue bound are analysed in order to demonstrate the approach. SH3, SH2 and the N- and C-terminal lobes of the kinase domain are detected as structural modules that move with respect to each other. Notably, a relative movement between the SH3 and SH2 domains is detected although both structures of Src kinase are in the 'assembled' state. Detailed analysis shows that Arg318, a residue topologically located in the N-terminal lobe of the kinase domain, structurally belongs to the C-terminal lobe. The movement of this residue together with the C-terminal lobe upon substrate binding leads to the loss of a salt bridge between Arg318 and Asp117, a residue in the SH3 domain, providing an explanation for the increased mobility of the SH3 domain.