ProQ3D: improved model quality assessments using deep learning

By Karolis Uziela, David Ménendez Hurtado, Nanjiang Shu, Björn Wallner, Arne Elofsson. 2016

159 citations recorded by the source corpus.

Summary: Protein quality assessment is a long‐standing problem in bioinformatics. For more than a decade we have developed state‐of‐art predictors by carefully selecting and optimising inputs to a machine learning method. The correlation has increased from 0.60 in ProQ to 0.81 in ProQ2 and 0.85 in ProQ3 mainly by adding a large set of carefully tuned descriptions of a protein. Here, we show that a substantial improvement can be obtained using exactly the same inputs as in ProQ2 or ProQ3 but replacing the support vector machine by a deep neural network. This improves the Pearson correlation to 0.90 (0.85 using ProQ2 input features). Availability and Implementation: ProQ3D is freely available both as a webserver and a stand‐alone program at http://proq3.bioinfo.se/ Contact: arne@bioinfo.se Supplementary information: Supplementary data are available at Bioinformatics online.

Subfields: Molecular Biology, Spectroscopy, Evaluation & Benchmarks, Computational Biology & Bioinformatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

In Spectroscopy

In Evaluation & Benchmarks

In Computational Biology & Bioinformatics

Citations in the corpus

IntFOLD: an integrated web resource for high performance protein structure and function prediction (2019, positive): “Each IntFOLD5 model was considered individually using 6 pure-single model methods (CDA (20), SSA (20), ProQ2 (4), ProQ2D (39), ProQ3D (39) and VoroMQA (40)), and four alternative quasi-single model methods (DBA (20), MF5s (20), MFcQs (20) and ResQ (41)).”
Learning from Protein Structure with Geometric Vector Perceptrons (2020, methodology): “FaeNNz 8 (Studer et al., 2020), ProQ3D (Uziela et al., 2017), and VoroMQA 9 (Olechnoviˇc & Venclovas, 2017) learn a multi-layer perceptron or statistical potential on top of such structural features.”
Methods for estimation of model accuracy in CASP12 (2017, background): “However, it can be noted that we have recently developed an improved version of ProQ3, ProQ3D that uses a deep-learning approach but identical inputs as ProQ3.(18) The final version was not ready for CASP12 and the preliminary version used did not perform better than ProQ3.”
ModFOLD6: an accurate web server for the global and local quality estimation of 3D protein models (2017, unknown): “The ModFOLD6 server performed particularly well in terms of differentiating between good and bad models (Table 2), local scoring (Supplementary Tables S2–5) and assigning absolute global accuracy values (Supplementary Tables S6–9).”
GraphQA: protein model quality assessment using graph convolutional networks (2020, unknown): “Inthis work, we tackle QA with graph convolutional networks (GCNs), which offer several desirable properties over previous methods.”
ATOM3D: Tasks On Molecules in Three Dimensions (2020, methodology): “, 2019], 3DCNN voxel-based methods trained on structural information, and Proq3D [Uziela et al., 2017], a deep-learning based method which employs structural information, Rosetta energy terms [Leaver-Fay et al.”
Large-scale structure prediction by improved contact predictions and model quality assessment (2017, methodology): “PcombC is a linear combination of three separate assessment scores: Pcons (Lundström et al., 2001), ProQ3D (Uziela et al., 2017), and the agreement between predicted and observed contacts in the model.”
Deep convolutional networks for quality assessment of protein folds (2018, methodology): “Table III shows a comparison of our model (3DCNN) with a number of state-of-the-art MQA methods: ProQ2D, ProQ3D [11], VoroMQA [44], and RWplus [45].”