Optimising intrinsic protein disorder prediction for short linear motif discovery

Kirsti M G Paulsen, Norman E Davey, Sobia Idrees, Åsa Pérez-Bercoff & Richard J Edwards.

This work was presented at the Sydney Bioinformatics Research Symposium 2018. (Abstract below.) Click on thumbnail for full resolution PDF.

Abstract

Short linear motifs (SLiMs) are short stretches of proteins that are directly involved in protein-protein interactions (PPI). Identifying SLiMs is important for understanding fundamental processes involved in normal cellular function. SLiMs are commonly only 3 - 10 amino acids in length and form low affinity interactions. This makes them ideal for fast cellular processes, such as cell signalling or response to stimuli, but also difficult to predict experimentally. As a result, many computational SLiM prediction methods have been developed. In order to increase the signal to noise ratio of SLiM predictions, different sequence masking techniques have been developed. These attempt to screen out areas that are unlikely to contain SLiMs and thereby preferentially eliminate the random nonfunctional sequences. One widely implemented masking strategy is to remove protein regions that form stable three-dimensional structures; SLiMs are typically found in regions of intrinsic disorder that are natively unstructured in their unbound form. To date, there has been no systematic study of how best to predict intrinsic disordered protein regions for SLiM discovery. Poor quality predictions will not have the desired noise-removal, while over-stringent masking will remove too many true positives. The aim of this study is to compare how ten different disorder prediction methods affect SLiM occurrence prediction and to identify the best method and settings for this purpose. The disorder prediction scores for each residue in the human proteome was obtained from the MobiDB database. Further, this study aims to investigate whether the optimal disorder masking settings for occurrence SLiM prediction are the same for de novo SLiM prediction and for identification of SLiM mediated PPIs.

ELM--the database of eukaryotic linear motifs

Dinkel H, Michael S, Weatheritt RJ, Davey NE, Van Roey K, Altenberg B, Toedt G, Uyar B, Seiler M, Budd A, Jödicke L, Dammert MA, Schroeter C, Hammer M, Schmidt T, Jehl P, McGuigan C, Dymecka M, Chica C, Luck K, Via A, Chatr-Aryamontri A, Haslam N, Grebnev G, Edwards RJ, Steinmetz MO, Meiselbach H, Diella F & Gibson TJ (2012): ELM—the database of eukaryotic linear motifs. Nucleic Acids Research 40(D1): D242-D251.

Abstract

Linear motifs are short, evolutionarily plastic components of regulatory proteins and provide low-affinity interaction interfaces. These compact modules play central roles in mediating every aspect of the regulatory functionality of the cell. They are particularly prominent in mediating cell signaling, controlling protein turnover and directing protein localization. Given their importance, our understanding of motifs is surprisingly limited, largely as a result of the difficulty of discovery, both experimentally and computationally. The Eukaryotic Linear Motif (ELM) resource at http://elm.eu.org provides the biological community with a comprehensive database of known experimentally validated motifs, and an exploratory tool to discover putative linear motifs in user-submitted protein sequences. The current update of the ELM database comprises 1800 annotated motif instances representing 170 distinct functional classes, including approximately 500 novel instances and 24 novel classes. Several older motif class entries have been also revisited, improving annotation and adding novel instances. Furthermore, addition of full-text search capabilities, an enhanced interface and simplified batch download has improved the overall accessibility of the ELM data. The motif discovery portion of the ELM resource has added conservation, and structural attributes have been incorporated to aid users to discriminate biologically relevant motifs from stochastically occurring non-functional instances.

PMID: 22110040