A Wiring of the Human Nucleolus Anders M. Hinsby, Lars Kiemer, E. Olof Karlberg, Kasper Lage, Anders Fausbøll, Agnieszka S. Juncker, Jens S. Andersen, Matthias Mann, Søren Brunak  Molecular Cell  Volume 22, Issue 2, Pages 285-295 (April 2006) DOI: 10.1016/j.molcel.2006.03.012 Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 1 Wiring the Human Nucleolus Flowchart of the steps of the analysis. Proteins are represented as nodes in the network. Blue nodes are from the input set of nucleolar proteins (see Experimental Procedures), white nodes are first-order interaction partners that were not in the input set, and red nodes represent proteins that were identified as nucleolar by reverse proteomics. Edges represent annotated interactions between indicated nodes. (1a) A list of known human nucleolus proteins was used as input to create an interaction map by the “interolog” approach (see text for details); (1b) a neural network was trained to discriminate nucleolus-localized proteins from other proteins; and (2) an algorithm was applied to cluster the large interaction map into subcomplexes of highly connected proteins. The output from the neural network was used to rank the resulting complexes according to their “nucleolus propensity.” The 15 top-ranked complexes were selected for further characterization; (3) putative novel nucleolus-localized proteins identified in the complexes were validated by a targeted search (reverse proteomics) in the raw MS files of the data generated for a previous publication (Andersen et al., 2005). Here, the complex ranked eight, of a total of ∼100 protein complexes, is presented (see Table 2 for full list of top 15 complexes). Molecular Cell 2006 22, 285-295DOI: (10.1016/j.molcel.2006.03.012) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 2 Functional Complexes of the Human Nucleolome The prominent role of the ribosome biogenesis complexes is evident in this sketching of the 15 top-ranked complexes of the human nucleolus. Complexes involved in other functions such as splicing, RNA trimming, and DNA repair are also partially covered in the analysis of the nucleolar proteome. Nodes represent experimentally verified, human nucleolus proteins, and edges signify protein-protein interactions. Dark blue nodes represent the 11 proteins that were predicted to be nucleolar in our analysis and subsequently experimentally confirmed by reverse proteomics (see text for details). Molecular Cell 2006 22, 285-295DOI: (10.1016/j.molcel.2006.03.012) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 3 The Ribosome Biosynthesis Pathway Two stages of the pathway were clearly distinguished in the analysis corresponding to the pre-40S/90S stage and the pre-60S stage. The identified complexes within the two stages were well-supported by independent information on mRNA coexpression (see Supplemental Data) and protein dynamics as a result of transcription inhibition (data from Andersen et al. [2005]) and contained a total of 49 uncharacterized human proteins (circled here, but see Table 3 for details). The two interaction clusters at each stage of the ribosome biosythesis pathway can be attributed to clustering, as it does not seem to be due to any apparent functional difference. Molecular Cell 2006 22, 285-295DOI: (10.1016/j.molcel.2006.03.012) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 4 The Exosome The human exosome is visualized, with node color representing the flux of proteins after inhibition of transcription (nucleolar dynamics data incorporated from Andersen et al. [2005]). The insert confers the nucleolar dynamics information on the 90S and pre-60S ribosome biogenesis complexes. Molecular Cell 2006 22, 285-295DOI: (10.1016/j.molcel.2006.03.012) Copyright © 2006 Elsevier Inc. Terms and Conditions