Deep convolutional networks for quality assessment of protein folds

TL;DR

This paper demonstrates that deep convolutional networks can effectively assess protein structure quality directly from raw 3D atomic densities, matching state-of-the-art methods without feature engineering.

Contribution

It introduces a deep neural network that learns to evaluate protein models from raw data, eliminating the need for handcrafted structural features.

Findings

Achieves comparable performance to top existing methods.

Performs consistently across multiple datasets.

Implicitly identifies regions deviating from native structures.

Abstract

The computational prediction of a protein structure from its sequence generally relies on a method to assess the quality of protein models. Most assessment methods rank candidate models using heavily engineered structural features, defined as complex functions of the atomic coordinates. However, very few methods have attempted to learn these features directly from the data. We show that deep convolutional networks can be used to predict the ranking of model structures solely on the basis of their raw three-dimensional atomic densities, without any feature tuning. We develop a deep neural network that performs on par with state-of-the-art algorithms from the literature. The network is trained on decoys from the CASP7 to CASP10 datasets and its performance is tested on the CASP11 dataset. On the CASP11 stage 2 dataset, it achieves a loss of 0.064, whereas the best performing method achieves a loss of 0.063. Additional testing on decoys from the CASP12, CAMEO, and 3DRobot datasets confirms that the network performs consistently well across a variety of protein structures. While the network learns to assess structural decoys globally and does not rely on any predefined features, it can be analyzed to show that it implicitly identifies regions that deviate from the native structure.