Sequence Versus Composition: What Prescribes IDP Biophysical Properties?

Abstract

Intrinsically disordered proteins (IDPs) represent a distinct class of proteins and are distinguished from globular proteins by conformational plasticity, high evolvability and a broad functional repertoire. Some of their properties are reminiscent of early proteins, but their abundance in eukaryotes, functional properties and compositional bias suggest that IDPs appeared at later evolutionary stages. The spectrum of IDP properties and their determinants are still not well defined. This study compares rudimentary physicochemical properties of IDPs and globular proteins using bioinformatic analysis on the level of their native sequences and random sequence permutations, addressing the contributions of composition versus sequence as determinants of the properties. IDPs have, on average, lower predicted secondary structure contents and aggregation propensities and biased amino acid compositions. However, our study shows that IDPs exhibit a broad range of these properties. Induced fold IDPs exhibit very similar compositions and secondary structure/aggregation propensities to globular proteins, and can be distinguished from unfoldable IDPs based on analysis of these sequence properties. While amino acid composition seems to be a major determinant of aggregation and secondary structure propensities, sequence randomization does not result in dramatic changes to these properties, but for both IDPs and globular proteins seems to fine-tune the tradeoff between folding and aggregation.

Keywords: IDP; IDR; aggregation propensity; secondary structure prediction; sequence randomization.

Figure 1

Scatter plot showing predicted secondary structure content (x-axis) and aggregation propensity (y-axis) of the PDB (blue) and DisProt (orange) datasets. The Disprot dataset is further annotated for classified examples of induced fold (by squares) and unfoldable (by circles) intrinsically disordered proteins (IDPs) in (A), which belong also to the most disordered Disprot predicted proteins by MobiDB-lite score in (B).

Figure 2

Compositional distribution of the PDB and DisProt datasets. Projections to the principal components (PC1 and PC2) are presented in (A), the Disprot dataset has the induced fold (squares) and unfoldable (circles) representatives highlighted. Contributions of individual amino acids to PC1 and PC2 are visualized in (B).

Figure 3

Analysis of properties of the DisProt and PDB datasets, showing the (A) mean values and (B) z-score of aggregation and secondary structure propensity after 1000 sequence permutations, compared to the original sequence.