Figure 2

Impact of a misalignment in homology modeling. The target protein T0295 from CASP7 is modeled using the crystal structure of human dimethyladenosine transferase (PDB 1ZQ9) as a template. In panel (A), we show the structural alignment of the template 1ZQ9 with the best (yellow) and worst (blue) models generated by MODELLER 9v5 for T0295. The structural diversity between these models is low; Thr31 for example superposes very well in the three structures. In panel (B), we show the effect of an error in the sequence alignment that serves as input to the modelling process. We shifted the alignment by one residue at the level of Thr31, and generated new models for T0295. The superposition of the template 1ZQ9 with both the best and the worst models shows locally a structural heterogeneity in the loop that contains Thr31. Thr31 is being shift by 3.4 Å due to one single error in the sequence alignment.