Learning sparse models for a dynamic Bayesian network classifier of protein secondary structure

Abstract

BackgroundProtein secondary structure prediction provides insight into protein function and is a valuable preliminary step for predicting the 3D structure of a protein. Dynamic Bayesian networks (DBNs) and support vector machines (SVMs) have been shown to provide state-of-the-art performance in secondary structure prediction. As the size of the protein database grows, it becomes feasible to use a richer model in an effort to capture subtle correlations among the amino acids and the predicted labels. In this context, it is beneficial to derive sparse models that discourage over-fitting and provide biological insight.ResultsIn this paper, we first show that we are able to obtain accurate secondary structure predictions. Our per-residue accuracy on a well established and difficult benchmark (CB513) is 80.3%, which is comparable to the state-of-the-art evaluated on this dataset. We then introduce an algorithm for sparsifying the parameters of a DBN. Using this algorithm, we can automatically remove up to 70-95% of the parameters of a DBN while maintaining the same level of predictive accuracy on the SD576 set. At 90% sparsity, we are able to compute predictions three times faster than a fully dense model evaluated on the SD576 set. We also demonstrate, using simulated data, that the algorithm is able to recover true sparse structures with high accuracy, and using real data, that the sparse model identifies known correlation structure (local and non-local) related to different classes of secondary structure elements.ConclusionsWe present a secondary structure prediction method that employs dynamic Bayesian networks and support vector machines. We also introduce an algorithm for sparsifying the parameters of the dynamic Bayesian network. The sparsification approach yields a significant speed-up in generating predictions, and we demonstrate that the amino acid correlations identified by the algorithm correspond to several known features of protein secondary structure. Datasets and source code used in this study are available at http://noble.gs.washington.edu/proj/pssp.

HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment

“…Although PSIPRED had been trained on PSI-BLAST MSAs, HHblits MSAs improved the Q3 score (fraction of correctly predicted secondary structure states) for proteins from the PDBselect 2007 dataset (Online Methods) from 80.4% to 81.3% and the secondary structure segment overlap (SOV) score from 77.5% to 78.6% (Supplementary Table 1). These results, obtained without training a large parameter set, are among the best achieved at present 14 .…”

A fundamental protein property, thermodynamic stability, revealed solely from large-scale measurements of protein function

“…For example, the prediction of protein structure might benefit from large-scale protein functional data that reveal amino acid preferences within particular structural elements (e.g., the paucity of proline residues in β-strands) and the functional effects of mutations that occur at spatially proximal positions. The feasibility of this approach is illustrated by existing structural prediction methods that are founded on these concepts but require extensive existing sequence alignment or structural training data (34,35). Another example relates to the understanding of enzyme mechanism, which might be uncovered by an analysis of the pattern of mutations that increase or decrease catalytic activity in large-scale protein functional data.…”

Deep mutational scanning: a new style of protein science

“…Typically, algorithms base predictions on the amino acid preferences in each type of secondary structure (α-helix, β-sheet or loop) in a training set of proteins with known structures 18,19 . As an alternative, large-scale mutational data on proteins with known structures could also reveal amino acid preferences within structural elements, and the resulting preferences used to enhance structure prediction algorithms.…”