Waltz, an exciting new move in amyloid prediction

NEWS AND VIEWS

Waltz, an exciting new move in amyloid prediction

2010 Nature America, Inc. All rights reserved.

Despite its messy appearance in the test tube, protein aggregation can be a highly sophisticated and ordered phenomenon. This is particularly true for the ordered assembly of polypeptides into amyloid fibrils whose manifestation accompanies several devastating neurodegenerative diseases. Even so, the mechanism of amyloid formation and the ability to predict which proteins will form amyloids have persistently eluded investigators, largely because of an acute shortage of structural data. In this issue of Nature Methods, Maurer-Stroh et al.1 have targeted this problem by expanding the existing libraries of amyloidogenic hexapeptides to develop a new prediction algorithmWaltzthat selectively recognizes local amyloid propensity.

This development allowed the authors to identify and experimentally validate more than 100 new amyloidogenic sequences hidden in the primary structure of prions, disease-associated proteins and functional amyloids. Interestingly, the authors hint also at a slight bias in the sequence composition between these three categories of fibrillating sequences.

In essence, the phenomenon of amyloid formation is captured by two observations: all proteins can form amyloid fibrils under certain conditions, and their structures at a macroscopic level seem to be very similar; yet, the fibrillation propensity displays a complex dependence on the sequence details. Amyloid assembly seems to be a generic state of proteins, but the

sequence details and native-state integrity determine how easily it is adopted under physiological conditions2.

The conformational landscape of an aggregating protein can be visualized as two competing free-energy funnels3, one pulling toward aggregation and the other toward the native structure. In contrast to the folding funnel4, the topology of the aggregation funnel is determined not only by sequence but also by higher-order parameters like concentration and mass transport. Depending on the conditions, the very same protein may thus end up as various types of amorphous aggregates, proto fibrils, amyloid fibrils or, quite often, an indistinct mixture of all. In some cases, however, there tends to be a bias toward forming ordered amyloids, noted by their distinct cross- structure, especially with more hydrophilic sequence signatures5.

The delicate task of amyloid prediction is therefore twofold: first, one must experimentally distinguish which type of aggregates form, and second, one must derive an algorithm thatgiven a...