Bioinformatics, 39(1), 2023, btac847 https://doi.org/10.1093/bioinformatics/btac847 Advance Access Publication Date: 2 January 2023 Applications Note Databases and ontologies The PRALINE database: protein and Rna humAn singLe nucleotIde variaNts in condEnsates Andrea Vandelli1,2, Magdalena Arnal Segura3,4, Michele Monti3, Jonathan Fiorentino 3, Laura Broglia3, Alessio Colantoni4, Natalia Sanchez de Groot1, Marc Torrent Burgas1, Alexandros Armaos3,* and Gian Gaetano Tartaglia 3,4,* 1 Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Barcelona 08193, Spain, 2Universitat Pompeu Fabra (UPF), Barcelona 08003, Spain, 3Center for Human Technologies (CHT), Istituto Italiano di Tecnologia (IIT), Genova 16152, Italy and 4Department of Biology and Biotechnologies, University Sapienza Rome, Roma 00185, Italy *To whom correspondence should be addressed. Associate Editor: Christina Kendziorski Received on July 23, 2022; revised on November 16, 2022; editorial decision on December 20, 2022 Abstract Summary: Biological condensates are membraneless organelles with different material properties. Proteins and RNAs are the main components, but most of their interactions are still unknown. Here, we introduce PRALINE, a database for the interrogation of proteins and RNAs contained in stress granules, processing bodies and other assemblies including droplets and amyloids. PRALINE provides information about the predicted and experimentally validated protein–protein, protein–RNA and RNA–RNA interactions. For proteins, it reports the liquid–liquid phase separation and liquid–solid phase separation propensities. For RNAs, it provides information on predicted second- ary structure content. PRALINE shows detailed information on human single-nucleotide variants, their clinical sig- nificance and presence in protein and RNA binding sites, and how they can affect condensates’ physical properties. Availability and implementation: PRALINE is freely accessible on the web at http://praline.tartaglialab.com. Contact: alexandros.armaos@iit.it or gian.tartaglia@iit.it 1 Introduction important, since they play a role in the process of condensation. Highly structured RNAs attract large amounts of proteins thanks Although the exact composition and functions of the different to their intrinsic ability to establish stable interactions (Sanchez condensates are unknown, they are enriched in protein and RNA de Groot et al., 2019). Moreover, RNAs can act as scaffolding ele- molecules that interact through protein–protein, protein–RNA and ments (Armaos et al., 2021), whereas a polypeptide of 100 amino RNA–RNA networks. Solid-like condensates, and in particular amy- acids can interact with one or two proteins, a chain of 100 nucleo- loids, are generally considered to be inherently irreversible aberrant tides is able to bind to 5–20 proteins (Vandelli et al., 2022). clumps (Dobson, 2017), while liquid-like condensates are dynamic Poorly structured transcripts also induce condensation, as they entities that exchange components with the surrounding environ- base-pair with other RNAs establishing a dense network of con- ment and grow, collapse and fuse in the nucleus and cytoplasm tacts (Treeck et al., 2018). All these data are gathered in (Marchese et al., 2016). Liquid-like condensates perform different PRALINE, a database that provides information on different con- functions on RNA molecules, such as storage in the germline, local- densates’ components, their interaction networks and disease- ization in neurons and protection from harmful conditions. The related variants. most known liquid-like condensates are processing bodies (PBs) and stress granules (SGs), both enriched in RNAs, which allow them to form and dissolve rapidly (Lorenzo Gotor et al., 2020). Yet, subtle changes in the composition or concentration of condensates’ constit- uents can induce the formation of solid-like assemblies (Cid-Samper et al., 2018). This is the case of Amyotrophic Lateral Sclerosis 2 Organization and content of the database (ALS), where single-nucleotide variants (SNVs) in FUS trigger a PRALINE can be accessed using protein or RNA names provided as liquid-to-solid phase transition (LSPT) (Patel et al., 2015). Gene Name, Ensembl Gene/Transcript ID (https://www.ensembl. Structural properties of the RNA and changes upon mutations are org/) and UniprotKB ID (https://www.uniprot.org/; Fig. 1A). C The Author(s) 2023. Published by Oxford University Press. V 1 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unre- stricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 2 A.Vandelli et al. Fig. 1. PRALINE database. (A) Search bar. The input can be a protein or an RNA in different ID formats. (B) Liquid–liquid phase separation (LLPS) and liquid–solid phase transition (LSPT) propensity profiles of a protein are predicted using catGRANULE and Zyggregator algorithms. (C) Protein SNVs description table: the difference in LLPS and LSPT compared to the WT is provided. (D) CROSS secondary structure propensity profile image of a RNA sequence. (E) RNA SNVs description table: the difference in CROSS secondary structure propensity compared to the WT, corresponding to a 11-nt window around the mutation, is provided, as well as proteins and RNAs interacting with the query transcript containing the SNV. (F) Example of an RNA–RNA interaction table. The information about RNAs’ binding sites, condensates localization and SNVs falling inside at least one of the binding sites are reported. The examples B–F relate to TARDBP • Searching for a specific protein, the user can retrieve informa- https://www.encodeproject.org/eclip/) that involve the SNV con- tion on the condensate state (droplet/liquid-like or amyloid/ taining region (Fig. 1F). solid-like) and the organelle in which it has been found (SG/PB). The predicted liquid–liquid phase separation (LLPS) and LSPT For most genes, information is available at both the protein and propensities and profiles of the wild-type sequence are provided, RNA levels, so it is possible to navigate from one molecule to the calculated with catGRANULE (Bolognesi et al., 2016) and other, revealing the links between them. Zyggregator (Tartaglia et al., 2008) methods, respectively (>0.80 accuracy in predicting regions of the proteins involved in 3 Applications protein condensation; Fig. 1B). Experimentally validated pro- PRALINE is a database that provides a comprehensive view of pro- tein–protein interactions are available through links to tein and RNA interactions and SNVs in human liquid-like and solid- BioGRID (https://thebiogrid.org/), while experimental and pre- like condensates. Information about experimentally validated and dicted protein–RNA interactions can be retrieved from RNAct predicted molecular interactions, including protein–protein, pro- (https://rnact.crg.eu/). Protein–RNA interactions are calculated tein–RNA and RNA–RNA, is provided, as well as the predicted using catRAPID, an algorithm trained on NMR and X-ray RNA secondary structure content and both LLPS and LSPT propen- structures (area under the curve (AUC), of 0.77 on eCLIP inter- sities of proteins. For each SNV, we provide a description of the associated actions) (Lang et al., 2019). The number of SNVs is shown for diseases, the binding sites and the change in RNA secondary struc- the protein of interest and, for each SNV, it is possible to inter- ture, LLPS and LSPT propensities. Combining physico-chemical rogate the amino acid position, the difference in LSPT and LLPS properties of molecules and disease-related annotations, PRALINE propensities compared to the reference (i.e. wild-type protein) helps to unravel macromolecular connections that sustain different and to retrieve information related to disease (Landrum et al., types of condensates and how variants can affect their equilibrium. 2014; Pi~ nero et al., 2020). LSPT and LLPS scores and profiles PRALINE is the first database providing LLPS and LSPT predictions for SNVs, and we envisage that it would greatly facilitate the design are provided (Fig. 1C). of experiments to study condensates’ formation and implication in • Searching for a specific RNA, the user can retrieve information human diseases. Although tested extensively and validated experi- on the condensate state (SG/PB), the RNA secondary structure mentally, catGRANULE predictions could not be benchmarked content (table and profile predicted using CROSS, http://s.tarta against a database of individual SNVs causing LLPS, due to a lack glialab.com/page/cross_group), the experimentally validated of adequate published resources. The availability of such databases RNA interactions (RISE database, http://rise.life.tsinghua.edu. will lead to a more precise understanding of the relationship cn/) and the predicted or experimentally validated protein inter- between SNVs, structural conformations, protein–RNA assembly and diseases. actions reported in RNAct (https://rnact.crg.eu/) for both the ref- erence sequence and SNVs (Landrum et al., 2014; Pi~ nero et al., 2020) (Fig. 1D and E). The RNA–RNA interactions table reports Acknowledgements information on different binding partners, if the interactors be- The authors would like to thank Adriano Setti for the RNA–RNA interactions long to a condensate, binding sites location in the transcripts and section and Leila Mansouri for the database name. related SNVs (Fig. 1F). The SNV section reports the position in the transcript, the difference in secondary structure compared to the reference (a numerical value and a profile image are pro- Funding vided) (Delli Ponti et al., 2017), associated diseases and interac- This work was supported by the ERC [ASTRA_855923] and H2020 projects tions with RNAs (RISE database) as well as proteins (eCLIP [IASIS_727658 and INFORE_825080]. The PRALINE database 3 Conflict of Interest: none declared. Dobson,C.M. (2017) The amyloid phenomenon and its links with human dis- ease. Cold Spring Harb. Perspect. Biol., 9, a023648. Landrum,M.J. et al. (2014) ClinVar: public archive of relationships among se- Data availability quence variation and human phenotype. Nucleic Acids Res., 42, D980–D985. Lang,B. et al. (2019) RNAct: protein-RNA interaction predictions General information is at https://praline.tartaglialab.com/about, for model organisms with supporting experimental data. Nucleic Acids where we provide a detailed description of the datasets and the tools Res., 47, D601–D606. employed in the database. Data provided in PRALINE are available Lorenzo Gotor,N. et al. (2020) RNA-binding and prion domains: the Yin and at https://praline.tartaglialab.com/downloads. The tutorial is at Yang of phase separation. Nucleic Acids Res., 48, 9491–9504. https://praline.tartaglialab.com/tutorial. Marchese,D. et al. (2016) Advances in the characterization of RNA-binding proteins. Wiley Interdiscip. Rev. RNA, 7, 793–810. Patel,A. et al. (2015) A liquid-to-solid phase transition of the ALS protein FUS accelerated by disease mutation. Cell, 162, 1066–1077. References Pi~ nero,J. et al. (2020) The DisGeNET knowledge platform for disease genom- Armaos,A. et al. (2021) RNA-protein interactions: central players in coordin- ics: 2019 update. Nucleic Acids Res., 48, D845–D855. ation of regulatory networks. BioEssays, 43, 2000118. Sanchez de Groot,N. et al. (2019) RNA structure drives interaction with pro- Bolognesi,B. et al. (2016) A concentration-dependent liquid phase separation teins. Nat. Commun., 10, 3246. can cause toxicity upon increased protein expression. Cell Rep., 16, Tartaglia,G.G. et al. (2008) Prediction of aggregation-prone regions in struc- 222–231. tured proteins. J. Mol. Biol., 380, 425–436. Cid-Samper,F. et al. (2018) An integrative study of Protein-RNA condensates Treeck,B.V. et al. (2018) RNA self-assembly contributes to stress granule forma- identifies scaffolding RNAs and reveals players in fragile X-associated trem- tion and defining the stress granule transcriptome. PNAS, 115, 2734–2739. or/ataxia syndrome. Cell Rep., 25, 3422–3434.e7. Vandelli,A. et al. (2022) The interplay between disordered regions in RNAs Delli Ponti,R. et al. (2017) A high-throughput approach to profile RNA struc- and proteins modulates interactions within stress granules and processing ture. Nucleic Acids Res, 45, e35. bodies. J. Mol. Biol., 434, 167159.