BETASCAN: probable beta-amyloids identified by pairwise probabilistic analysis

Abstract

Amyloids and prion proteins are clinically and biologically important beta-structures, whose supersecondary structures are difficult to determine by standard experimental or computational means. In addition, significant conformational heterogeneity is known or suspected to exist in many amyloid fibrils. Recent work has indicated the utility of pairwise probabilistic statistics in beta-structure prediction. We develop here a new strategy for beta-structure prediction, emphasizing the determination of beta-strands and pairs of beta-strands as fundamental units of beta-structure. Our program, BETASCAN, calculates likelihood scores for potential beta-strands and strand-pairs based on correlations observed in parallel beta-sheets. The program then determines the strands and pairs with the greatest local likelihood for all of the sequence's potential beta-structures. BETASCAN suggests multiple alternate folding patterns and assigns relative a priori probabilities based solely on amino acid sequence, probability tables, and pre-chosen parameters. The algorithm compares favorably with the results of previous algorithms (BETAPRO, PASTA, SALSA, TANGO, and Zyggregator) in beta-structure prediction and amyloid propensity prediction. Accurate prediction is demonstrated for experimentally determined amyloid beta-structures, for a set of known beta-aggregates, and for the parallel beta-strands of beta-helices, amyloid-like globular proteins. BETASCAN is able both to detect beta-strands with higher sensitivity and to detect the edges of beta-strands in a richly beta-like sequence. For two proteins (Abeta and Het-s), there exist multiple sets of experimental data implying contradictory structures; BETASCAN is able to detect each competing structure as a potential structure variant. The ability to correlate multiple alternate beta-structures to experiment opens the possibility of computational investigation of prion strains and structural heterogeneity of amyloid. BETASCAN is publicly accessible on the Web at http://betascan.csail.mit.edu.

Figure 1. Sample output of the BETASCAN algorithm.

The results for the β-helix domain of pectate lyase C are shown. (A) heat-map of all β-strand probabilities. The horizontal axis indicates the N-terminal residue of potential β-strands, while the vertical axis indicates strand length. The upper and lower boxes display results for the two orientations of the strand. Colors indicate propensity of strand formation. Red indicates above-background probability, while blue indicates below-background probability. (B) predicted most likely β-strands based on single strand probabilities. BETASCAN predictions are marked as horizontal lines, shading from red (maximum predicted score) to yellow (zero score, i.e., probability equal to background). The horizontal axis indicates the N-terminal residue of potential β-strands, while the vertical axis indicates the log-odds propensity. Overlapping strands represent alternate folding patterns with indicated likelihoods. Purple brackets indicate experimentally determined β-strands as derived from the PDB structure. (C) predicted most likely strand-pairs based on pairwise probabilities. Purple dots indicate the N-terminii of experimentally determined strand-pairs as derived from the PDB structure.

Figure 2. Statistics on BETASCAN accuracy in the set of β-helices.

(A) effect of maximum β-strand length on sensitivity and positive predictive value of BETASCAN, measured by strand and by residue. (B) effect of maximum β-strand length on average absolute value difference in predicted and crystal-structure observed β-strand edges. (C–E) effectiveness of BETASCAN singleton scores in β-structure prediction; (C) ROC curve calculated residue-by-residue; (D) graph of sensitivity against (1-PPV) calculated strand-by-strand; (E) ROC curve calculated strand-by-strand. (F–H) effectiveness of BETASCAN pairwise scores in β-structure prediction; (F) ROC curve calculated residue-by-residue; (G) graph of sensitivity against (1-PPV) calculated strand-by-strand; (H) ROC curve calculated strand-by-strand.

Figure 3. BETASCAN output for amyloid and prion proteins with experimentally determined β-structures.

Green vertical brackets indicate experimentally derived locations of β-strands; blue brackets indicate locations determined by a separate method. In the same manner as Figure 1b, BETASCAN predictions are marked as horizontal lines, shading from red (maximum predicted score) to yellow (zero score, i.e., probability equal to background). Overlapping lines indicate alternate folding patterns for the β-strands, with indicated probability. Two graphs are included to display the results for each orientation of the strand. For purposes of comparison, the set of highest-scoring non-overlapping strands in the BETASCAN single-strand prediction was taken as the predicted structure. Corresponding outputs of PASTA ,, TANGO , and Zyggregator are displayed below the BETASCAN results. Refer to Table 1 for a summary of the correspondences of these predictions. (A) amyloid-β structure as determined by Luehrs et al. (green) and Petkova et al. (blue); (B) het-S structure as determined by Ritter et al. (green) and Wasmer et al. (blue); (C) α-synuclein structure as determined by Heise et al. ; (D) amylin structure as determined by Kajava et al. ; (E) tau protein fragment PHF43 structure as determined by von Bargen et al. .

Figure 4. Sensitivtiy/specificity curve of BETASCAN highest singleton result for 120 non-redundant sequences experimentally observed for aggregation potential, collated by and clustered by CD-HIT .

Figure 5. Relationships between physical features of β-sheet components and the definition of the computational search spaces.

(A) variable definitions as used in Materials and Methods . The two vertical beta-strands form a single strand-pair, with odd residues labeled in white and even residues in black. The strands share the same orientation o and extend from p to p+l and from q to q+l. (B) structure of the lattice of the β-strand search space defined by the variables p (location), l (length), and o (orientation). Changes in the parameters of a β-strand are physically possible in a single step only along the paths marked by arrows. The arrowheads therefore define the relative locations queried by the maxima-finding algorithm at each point. (C) structure of the lattice of the strand-pair search space defined by the variables p (first strand location), q (second strand location), l (length) and o (orientation, not shown). In addition to the physically possible changes in B, shifts of one or two residues in the relative strand positions are possible. Arrowheads indicate the relative locations queried by the strand-pair maxima-finding algorithm for each point.