Computational identification of prion-like RNA-binding proteins that form liquid phase-separated condensates

Abstract

Motivation Eukaryotic cells contain different membrane-delimited compartments, which are crucial for the biochemical reactions necessary to sustain cell life. Recent studies showed that cells can also trigger the formation of membraneless organelles composed by phase-separated proteins to respond to various stimuli. These condensates provide new ways to control the reactions and phase-separation proteins (PSPs) are thus revolutionizing how cellular organization is conceived. The small number of experimentally validated proteins, and the difficulty in discovering them, remain bottlenecks in PSPs research. Results Here we present PSPer, the first in-silico screening tool for prion-like RNA-binding PSPs. We show that it can prioritize PSPs among proteins containing similar RNA-binding domains, intrinsically disordered regions and prions. PSPer is thus suitable to screen proteomes, identifying the most likely PSPs for further experimental investigation. Moreover, its predictions are fully interpretable in the sense that it assigns specific functional regions to the predicted proteins, providing valuable information for experimental investigation of targeted mutations on these regions. Finally, we show that it can estimate the ability of artificially designed proteins to form condensates (r=−0.87), thus providing an in-silico screening tool for protein design experiments. Availability and implementation PSPer is available at bio2byte.com/psp. Supplementary information Supplementary data are available at Bioinformatics online.

Prediction of liquid-liquid phase separation proteins using machine learning

“…We also calculated the number of overlapped PSPs predicted between any two tools and the overlapping matrix is shown in Figure 3B. For all prediction tools, the proportions of predicted PSPs ranked in the order of tier 1 > tier 2 > tier 3 > tier 4, which is in accordance with the degree of These results re-emphasize that there are only a small proportion of proteins spontaneously involved in the formation of condensates [20,26], even for the scaffolds, a large proportion of which might participate in multi-component LLPS environments.…”

“…We also tested two first generation PSP prediction tools, PScore and CatGranule, which performed best among the 7 first-generation methods [13] and PSPer [20], on dataset F1+. The relationships between percent recall and total percentage of whole proteins accepted at given thresholds, for PScore, CatGranule, PSPer and our Model 1, are shown in Figure 1.…”

“…Each of them is based on specific protein features which are deemed as the driving force of LLPS. Specifically, PScore is based on the expected number of long-range, planar sp2 pi-pi contacts [14]; DDX4-like is based on sequence composition and residue spacing similarity to DDX4 [15]; PLAAC is based on prionlike domains [16]; LARKS is based on low-complexity aromatic-rich kinked segments [17]; R+Y is based on the proportion of arginine and tyrosine and the features of FET family proteins [18]; and CatGranule is based on statistical analysis of amino acids composition responsible for granule forming [19] Recently, Orlando et al [20]. Their tool (PSPer) has been successfully used to predict the phase separation ability for 22 experimentally studied FUS-LIKE proteins [18].…”

Protein Databases Related to Liquid–Liquid Phase Separation

“…Recently, a predictor of LLPS protein (PSPredictor, ) based on machine learning was developed [ 68 ], using the datasets in LLPSDB as a training set. It achieved a fairly high prediction accuracy and outperformed other reported prediction tools so far, which are all based on specific protein sequence features [ 61 , 67 , 69 ]. The well-summarized structural, functional, and detailed experimental information provided in PhaSePro makes it very useful for researchers to find complete and systematic knowledge of LLPS proteins.…”

Prediction of disordered regions in proteins with recurrent Neural Networks and protein dynamics

“…In the case of disorder prediction, such emergent properties can capture the underlying reasons of why a protein region tends to be disordered, and so help to make the disorder predictions more generally applicable. DisoMine has also already been successfully used in a pipeline for the identification of prion-like RNA-binding proteins that form liquid phase-separated condensates [15], defining the disorder content of the regions that typically constitute this class of proteins.…”