SABBAC: online Structural Alphabet-based protein BackBone reconstruction from Alpha-Carbon trace

Abstract

SABBAC is an on-line service devoted to protein backbone reconstruction from alpha-carbon trace. It is based on the assembly of fragments taken from a library of reduced size, selected from the encoding of the protein trace in a hidden Markov model-derived structural alphabet. The assembly of the fragments is achieved by a greedy algorithm, using an energy-based scoring. Alpha-carbon coordinates remain unaffected. SABBAC simply positions the missing backbone atoms, no further refinement is performed. From our tests, SABBAC performs equal or better than other similar on-line approach and is robust to deviations on the alpha-carbon coordinates. It can be accessed at http://bioserv.rpbs.jussieu.fr/SABBAC.html.

Figure 1

Peptidic bonds angle distribution. Distribution of the peptide unit planes angular deviations between native and rebuilt structures for perturbed alpha-carbon traces. Each deviation is calculated for locally fitted backbones. Left, average alpha-carbon trace perturbation of 0.2 Å. Right, average perturbation of 1.0 Å.

Figure 2

SABBAC rebuilding example, 1OMD. (A) Native and SABBAC rebuilt structures. (B) Native and SABBAC rebuilt structures from an alpha-carbon trace perturbed by 0.8 Å on average. The native structure is represented in blue.

Figure 3

SABBAC and MaxSprout reconstruction performance for 31 CASP6 targets. For each of target, the best and the rank 5 models have been considered. The fraction of the peptide unit planes deviating by <10° (left) and 40° (right) are plotted as a function of the TM-Scores of the models. Lines correspond to regressions. Red + CASP6 models versus native structure; black x SABBAC reconstruction; blue triangles MaxSprout reconstruction.