ReFOLD3: refinement of 3D protein models with gradual restraints based on predicted local quality and residue contacts

Abstract

ReFOLD3 is unique in its application of gradual restraints, calculated from local model quality estimates and contact predictions, which are used to guide the refinement of theoretical 3D protein models towards the native structures. ReFOLD3 achieves improved performance by using an iterative refinement protocol to fix incorrect residue contacts and local errors, including unusual bonds and angles, which are identified in the submitted models by our leading ModFOLD8 model quality assessment method. Following refinement, the likely resulting improvements to the submitted models are recognized by ModFOLD8, which produces both global and local quality estimates. During the CASP14 prediction season (May-Aug 2020), we used the ReFOLD3 protocol to refine hundreds of 3D models, for both the refinement and the main tertiary structure prediction categories. Our group improved the global and local quality scores for numerous starting models in the refinement category, where we ranked in the top 10 according to the official assessment. The ReFOLD3 protocol was also used for the refinement of the SARS-CoV-2 targets as a part of the CASP Commons COVID-19 initiative, and we provided a significant number of the top 10 models. The ReFOLD3 web server is freely available at https://www.reading.ac.uk/bioinf/ReFOLD/.

Graphical abstract

Refinement of a 3D model using ReFOLD3.

Figure 1.

The ReFOLD3 server results page for an FM CASP14 target (T1031). In CASP14, the ReFOLD3 protocol was used to refine the top selected server models. (A) The results page provides the superposition of the top model and the input 3D model, a summary of the predicted scores and improvements in model quality estimation, along with the confidence intervals and P-value for each generated 3D model. After clicking on images or plots, users can download and view the data in detail (B) The plot of the predicted per-residue accuracy scores produced by ModFOLD8 for the top refined 3D model (green bars) compared with the original 3D model (red bars). Plots can also be downloaded as PDFs. (C) Interactive superposition of the top 3D model and the original 3D model, which are displayed in 3D using JSmol. Models can also be downloaded which including the per-residue error data in the B-factor column.

Figure 2.

The refinement of four CASP14 targets using the ReFOLD3 protocol. In the left panels, the initial structures were coloured by the per-residue accuracy score produced by ModFOLD8. In the middle panels, superposition of the best predicted 3D server model selected by ModFOLD8 or the starting model provided by CASP in the refinement category (cyan), the top 3D model generated by ReFOLD3 (magenta) and native structure (green). In the right panels, GDT_TS plots for the comparison of the original 3D model (cyan) with the top 3D model generated by ReFOLD3 (magenta). (A) CASP14 regular TBM-hard target T1030: RaptorX_TS1 versus McGuffin_TS1, a GDT_TS improvement from 39.84 to 43.77. (B) CASP14 regular FM/TBM target T1035 domain 1: tFOLD-lDT_TS3 versus McGuffin_TS1, a GDT_TS improvement from 48.28 to 50.00. (C) CASP14 refinement FM target R10909: starting model versus McGuffin_TS1, a GDT_TS improvement from 65.61 to 67.06 (D) CASP14 refinement TBM-easy target R1091-D1: starting model versus McGuffin TS1, a GDT_TS improvement from 79.21 to 81.78. Images are created using PyMOL (http://www.pymol.org). GDT_TS plots are from https://www.predictioncenter.org/casp14/.