3DRobot: automated generation of diverse and well-packed protein structure decoys

Abstract

Motivation: Computationally generated non-native protein structure conformations (or decoys) are often used for designing protein folding simulation methods and force fields. However, almost all the decoy sets currently used in literature suffer from uneven root mean square deviation (RMSD) distribution with bias to non-protein like hydrogen-bonding and compactness patterns. Meanwhile, most protein decoy sets are pre-calculated and there is a lack of methods for automated generation of high-quality decoys for any target proteins.

Results: We developed a new algorithm, 3DRobot, to create protein structure decoys by free fragment assembly with enhanced hydrogen-bonding and compactness interactions. The method was benchmarked with three widely used decoy sets from ab initio folding and comparative modeling simulations. The decoys generated by 3DRobot are shown to have significantly enhanced diversity and evenness with a continuous distribution in the RMSD space. The new energy terms introduced in 3DRobot improve the hydrogen-bonding network and compactness of decoys, which eliminates the possibility of native structure recognition by trivial potentials. Algorithms that can automatically create such diverse and well-packed non-native conformations from any protein structure should have a broad impact on the development of advanced protein force field and folding simulation methods. AVAILIABLITY AND IMPLEMENTATION: http://zhanglab.ccmb.med.umich.edu/3DRobot/

Contact: jiay@phy.ccnu.edu.cn; zhng@umich.edu

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

3DRobot protocol for protein structure decoy construction

Fig. 2.

Comparison of 3DRobot and control decoy sets on Evenness score. (A) I-TASSER_set; (B) Modeller_set; (C) Rosetta_set

Fig. 3.

Comparison of 3DRobot and control decoy sets on normalized pair-wise RMSD (npwRMSD). (A) I-TASSER_set; (B) Modeller_set; (C) Rosetta_set

Fig. 4.

Comparison of 3DRobot and control decoy sets on secondary structure distribution (P_sec) relative to the native structure. Open circles are P_set value for native structure and solid squares are the average P_sec for decoy conformations. The solid error bar denotes the standard deviation of P_sec for each decoy set. The dotted line with bar shows P_sec range of each decoy set. (A) I-TASSER_set; (B) Modeller_set; (C) Rosetta_set

Fig. 5.

Typical examples of hydrogen bonding network construction for decoys with high structure deviations. (A) Comparison of decoy conformations of 1tig by I-TASSER and 3DRobot, respectively. (B) Comparison of decoy conformations for 1mdc by Modeller and 3DRobot, respectively

Fig. 6.

Illustrative examples showing correlation of RMSD versus secondary structure (P_sec) or radius of gyrations (R_G). The gray solid circles denote the original decoys and the dark hollow diamond are the decoys by 3DRobot. (A) P_sec versus RMSD correlation from Modeller_set on 2pna; (B) P_sec versus RMSD correlation from I-TASSER_set on 1di2A; (C) R_G versus RMSD correlation from Rosetta_set on 1acf; (D) R_G versuss RMSD correlation from I-TASSER_set on 1gyvA

Fig. 7.

An illustrative example showing Dope_REF energies of the structure decoys by I-TASSER and 3DRobot. (A–C) X- and Y-axes are the residue order number. Energy for residue pair is calculated as the sum of energies of all atom pairs between the two residues. The upper triangle of each plot shows the energetically favorable residue pairs (energy<−3.0), and the lower triangle of each plot shows the energetically unfavorable residue pairs (energy > 3.0). Residue pairs with energy between −3.0 and 3.0 are not shown. Different colors are used to illustrate energy variation. (D–E) structure decoys by I-TASSER_set and 3DRobot (red) superposed on the native structure of 1jnuA (green)