Analysis of Protein Phosphorylation and Its Functional Impact on Protein-Protein Interactions via Text Mining of the Scientific Literature

Abstract

Post-translational modifications (PTMs) are one of the main contributors to the diversity of proteoforms in the proteomic landscape. In particular, protein phosphorylation represents an essential regulatory mechanism that plays a role in many biological processes. Protein kinases, the enzymes catalyzing this reaction, are key participants in metabolic and signaling pathways. Their activation or inactivation dictate downstream events: what substrates are modified and their subsequent impact (e.g., activation state, localization, protein-protein interactions (PPIs)). The biomedical literature continues to be the main source of evidence for experimental information about protein phosphorylation. Automatic methods to bring together phosphorylation events and phosphorylation-dependent PPIs can help to summarize the current knowledge and to expose hidden connections. In this chapter, we demonstrate two text mining tools, RLIMS-P and eFIP, for the retrieval and extraction of kinase-substrate-site data and phosphorylation-dependent PPIs from the literature. These tools offer several advantages over a literature search in PubMed as their results are specific for phosphorylation. RLIMS-P and eFIP results can be sorted, organized, and viewed in multiple ways to answer relevant biological questions, and the protein mentions are linked to UniProt identifiers.

1. Introduction

Post-translational modifications (PTMs) are an important contributor to protein diversity. PTMs play a pivotal role in protein function, regulating activity, localization, and protein-protein interactions (PPIs), and therefore disruptions in PTMs can lead to disease [1]. In particular, protein phosphorylation is an essential regulatory mechanism in many biological processes. Proteins can be phosphorylated at different and/or multiple positions, most commonly on serine, threonine, and tyrosine residues. Protein kinases, the enzymes catalyzing the phosphorylation reaction, play a key role in regulating these events and have become therapeutic targets for drug design in multiple diseases [2–4]. However, few drugs targeting kinases have been completely successful in the clinic mainly due to the conserved nature of kinases. Consequently, many of the available inhibitors lack sufficient selectivity for effective clinical application. The identification and characterization of kinase-substrate interactions are keys to improve the approaches to targeted drug development [5].

The scientific literature contains a wealth of protein phosphorylation data derived both from traditional experiments that focus on a small number of proteins and from high-throughput experiments that attempt to assess the phosphorylation state of the whole proteome [6]. Researchers frequently query PubMed or specialized databases to gain access to this information. Similarly, database biocurators collect literature, and read and extract the most salient information relevant to their domain. Given the continuing increase of the size of the PubMed database, finding or collecting information that is spread across this vast knowledge pool remains challenging. Automatic methods to bring this data together can help to summarize the current knowledge and to expose hidden connections. For example, one article might describe that phosphorylation of a protein at a given site is implicated in a particular disease, and another article might describe a kinase that phosphorylates the site, leading to the connection of the kinase to the disease, which could be investigated further. Text mining tools have evolved considerably in number and quality and are being used to address a variety of research questions in the biomedical domain; for recent reviews see [7–9].

iProLINK (integrated Protein Literature, INformation and Knowledge) [10] offers a portfolio of text mining tools and annotated corpora developed by our group. Some of these are intended for developers to serve as modules in specific steps of their text mining pipelines (e.g., iXtractR [11] for relation extraction, and iSimp for sentence simplification [12]). Others are applications for biomedical researchers and biocurators to facilitate the exploration of the literature about proteins (pGenN [13], eFIP [14,15], eGIFT [16], and RLIMS-P [17,18]) and microRNAs (miRTex [19]) (Table 1).

Table 1.

Text mining tools available in iProLINK.

Tool	Description	Bioentities/Relations	Standard Used
pGenN	Identifies plant gene name mentions in Medline abstracts	Protein/gene	UniProt identifier EntrezGene
eGIFT	Identifies informative terms (iTerms) and documents relevant to a gene/protein (abstract level)	Protein/gene Informative term (iTerm)	GO term UniProt Keyword
miRTex	Identifies miRNA-target relations as well as miRNA-gene and gene-miRNA regulation relations in Medline abstracts	miRNA-target gene-miRNA miRNA-gene
RLIMS-P	Identifies information relevant to protein phosphorylation: kinase, substrate and sites. (abstract and full-length PMC open access articles)	Kinase-substrate Substrate-site Kinase-substrate-site	UniProt Identifier
eFIP	Identifies phosphorylation-dependent protein-protein interactions (abstract and full-length PMC open access articles)	Phosphorylation-dependent PPI Impact on PPI (promote or inhibit)	UniProt Identifier

Among these applications, RLIMS-P and eFIP facilitate the extraction of phosphorylation information from the literature and therefore are the focus of this book chapter.

RLIMS-P is a rule-based information extraction system that identifies kinase, substrate, and site relations in the scientific literature (including PubMed abstracts and PMC open access (OA) full-length articles). For example, the tuple <Akt, CHK1, Ser280> is extracted by RLIMS-P from the following sentence:

“CHK1 is directly phosphorylated by Akt at Ser280, a modification that results in cytoplasmic sequestration” [20].

Since these three entities (kinase, substrate and site) are rarely co-mentioned in the same sentence, RLIMS-P employs techniques that combine information found in different sentences. The kinase or substrate names detected could correspond to individual proteins (e.g., Crm1), protein complexes (e.g., CDK1-cyclin-B), or a group of related proteins (e.g., Src kinases), whereas a site could be a residue type (e.g., serine, threonine, and tyrosine), a specific residue (e.g., Ser-391), or a protein region or domain (e.g., C-terminal domain) [18]. RLIMS-P has been benchmarked with multiple corpora [17]. The F-scores (harmonic mean between precision and recall), based on a collection of sections derived from 100 full-text articles, have previously been reported to be 0.88, 0.91, and 0.92 for kinases, substrates, and sites, respectively [17]. In addition, RLIMS-P integrates GNormPlus [21] to link the detected kinase and substrate names to UniProt identifiers whenever possible.

eFIP builds on RLIMS-P by first detecting mentions of protein phosphorylation (kinase, substrate, and site), but adds detection of protein-protein interactions (PPIs) involving the phosphorylated protein. The types of PPIs captured include interactions between two proteins, or interactions between a protein and a protein complex, protein region, or protein class. Once the phosphorylation and PPI mentions are detected, the second step is to identify a possible relation between the two events. The evaluation of eFIP on full-length articles achieved an F-measure of 0.84 on 100 article sections [14]. Selected data from RLIMS-P and eFIP has been integrated in iPTMnet (http://proteininformationresource.org/iPTMnet/) and is actively used in the curation of proteoforms in the Protein Ontology [22].

This chapter demonstrates how to use RLIMS-P and eFIP to uncover information about protein phosphorylation and phosphorylation-dependent PPIs from the literature.

2. Materials

2.1. Web Sites

iProLINK: http://proteininformationresource.org/iprolink

RLIMS-P: http://proteininformationresource.org/rlimsp

eFIP: http://proteininformationresource.org/efip

2.2. General aspects of the RLIMS-P and eFIP interfaces

Input

Both the RLIMS-P and eFIP web sites allow the input of keywords or phrases that can be combined with Boolean operators (AND, OR, NOT) in the same way as building a PubMed query. Similarly, MeSH terms (controlled vocabulary used to index Medline abstracts) can be included in the search (e.g., “Alzheimer Disease” [Mesh]). The input is sent to the PubMed web site and relevant PMIDs are retrieved. The PMIDs are then used to query a backend database that hosts pre-processed results for PubMed abstracts and full-length PMC OA documents by RLIMS-P or eFIP. In both systems, you have the option to restrict the search to a particular organism of interest (Figure 1A 3, Figure 2A). You can also select to exclude review articles if you are only interested in research articles, and/or query only abstracts (Figure 1A 4). eFIP also supports searches based on protein roles (kinases, substrates, interacting partners) for protein names. Alternatively, a list of PMIDs or PMCIDs, delimited by comma, space or listed in new lines, can be entered (Figure 1A 5, Figure 2A).

Fig. 1.

RLIMS-P for extraction of kinase-substrate-site information about CHK1. A. RLIMS-P homepage, showing the different functionalities: login capability (1), input query options such as keywords (2) or PubMed IDs (5) and search options (3, organism restriction or 4, exclusion of review articles or only abstracts). B. Partial display of RLIMS-P results for CHK1 search with summary statistics (1), and tables with “View by Summary” (2), and “View by Substrate” views (3). The “Text Evidence” (4) column provides links to the text evidence page. Text mining results can be downloaded in CSV format (5)

Fig. 2.

eFIP for extraction of phosphorylation-dependent PPI information about CHK1. A. eFIP homepage, showing the different functionalities: input query options such as keywords (1), protein names (2), or PubMed/PMC IDs (4) and search options for organism restriction and protein type including substrate, kinase or interactant (3). B. Partial display of eFIP results for CHK1 search with summary statistics (1), and table with “View by Summary” (2). The columns “No. of Sentences” (3) and “Text Evidence” (4) provide links to the text evidence pages. Text mining results can be downloaded in CSV format and be viewed in Cytoscape (5)

Results

The RLIMS-P result page presents summary statistics of the retrieved results (Figure 1B 1), listing separately the number of documents with potential phosphorylation information (i.e., those with the word “phosphorylation” or similar ones) and those with phosphorylation information according to RLIMS-P (i.e., there is at least one substrate identified). In addition, eFIP shows the summary statistics for interactants detected (Figure 2B 1).

Editing capabilities

To unlock editing capabilities, user registration and login are required (Figure 1A 1, Figure 2A, see Note ¹). Edited results can be downloaded.

Cytoscape

eFIP offers a graphical view of the text mining results, displaying the protein entities as nodes and their relations as edges. The node names correspond to the protein entities in the result table, with some of the longer names abbreviated. The graph can be saved in PNG and XGMML-beta (Cytoscape compatible) format (Figure 6 2). Substrates, kinases, and interactants are represented as nodes with red circles, green pentagons, and orange circles, respectively. Interactions that are enabled or enhanced by phosphorylation are depicted as edges using solid orange lines with pointed arrowheads, whereas those that are decreased or inhibited are depicted by dashed orange lines with T-type arrowheads (Figure 6).

Fig. 6.

The Cytoscape representation for the phosphorylation and PPI events extracted from articles with PMID 12676962, 17380128, and 20639859. Events extracted are illustrated with nodes and edges; see legend for details (1). The graph can be exported (2)

3. Methods

For illustration purposes, we will showcase RLIMS-P and eFIP tool usage with examples from the Checkpoint kinase-1 protein, commonly referred to as CHK1 or CHEK1. This protein is a serine/threonine-specific protein kinase. It coordinates the DNA damage response (DDR) and cell cycle checkpoint response [23]. Activation of CHK1 results in the initiation of cell cycle checkpoints, cell cycle arrest, DNA repair and cell death to prevent damaged cells from progressing through the cell cycle [24]. A recent review article by Goto et al. [25] describes the regulation of CHK1 via phosphorylation, its substrates and the functional impact. To validate the approach, we compare the output of our text mining tools with the knowledge in the review article when applicable. We illustrate in the following text a variety of examples of RLIMS-P and eFIP usage via specific biological questions.

3.1 How to find kinases acting on a given substrate. What sites are phosphorylated?

Is CHK1 phosphorylated? If so, which sites? By what kinases? To answer these questions, we will use the RLIMS-P website (http://proteininformationresource.org/rlimsp, Figure 1A). The goal in this case is to find the articles mentioning CHK1 as a substrate, as we are interested in its phosphorylation sites. To achieve the most comprehensive result, it is recommended to include the different names by which CHK1 is known (e.g., CHEK1, Checkpoint kinase-1). If you are not familiar with the variety of names that are used for your protein of interest, you can check in a reference curated source, such as UniProt [26] or Entrez [27]. For this case, we will use the query (Figure 1A 2):

CHK1 OR CHEK1 OR “checkpoint kinase-1”

1. Go to RLIMS-P website and enter this query in the box and submit. Results are returned as shown in Figure 1B. Information on the top of the page summarizes the general statistics for the search results (Figure 1B 1), including the number of articles with potential protein phosphorylation mentions and the number of kinase, substrate, and site mentions (see Note ²).

2. Display results by “Substrate.” The results from the search in RLIMS-P include articles where the keywords are mentioned and which are about protein phosphorylation. The default table view is a summary listing the kinase and substrate mentions for each PMID. To obtain the subset where CHK1 is the phosphorylated protein, choose the option “View by Substrate” from the pull-down menu (Figure 1B 2) (see Note ³).

3. Find CHK1 as substrate. The table in Figure 1B 3 is now substrate centric. Next, we have to find CHK1 in the substrate column. As shown in this table, there are many articles describing phosphorylation of CHK1 (where CHK1 acts as a substrate). In addition, the kinases that phosphorylate CHK1 and the phosphorylation sites can now be easily identified in the columns “PTM enzyme” and “Phosphorylation Site,” respectively.

4. Validate and summarize the information. When the results are viewed by substrate (as shown in Figure 1B 3), all the phosphorylation sites on a substrate are shown. Now continue with our example by looking for CHK1 as substrate. The “No. of Sentences” column provides quick access to evidence sentences with color-coded highlighting of kinase (green), substrate (blue), and site (red) mentions (see Figure 4 bottom panel). This page is almost the same as the page linked out through icons in the “Text Evidence” column (Figure 1B 4), except that it restricts its sentence display to those where the information tuples are directly derived. To validate the information, the evidence can also be viewed by clicking on the icon in the “Text Evidence” column (Figure 1B 4), which will take you to the evidence page (Figure 3A). The evidence page presents a table summarizing the data extracted from the article with links to the source sentences (Figure 3A 2), a block showing the relevant sentences from the text (abstract or full text) with color-coding highlighting (Figure 3A 3), and the normalization table, which suggests UniProt identifiers for the kinases and substrates detected (Figure 3A 3–4). Results can be filtered by specific sections of the article (e.g., figure legends, result section, abstract, etc., see Figure 3A 1). If a user is logged in, he or she can validate individual information tuples by clicking on the check or “X” next to the annotation to agree or disagree, respectively (Figure 3B 1). The example shown in Figure 3B demonstrates the agreement on data extracted for phosphorylation of Ser-280 on Chk1 by PIM kinases. User can add additional information in the comment box, in this case, the more specific kinase PIM1 (Figure 3B). In addition, the “Add Annotation” (Figure 3B 2) allows addition of manually curated information tuples. Furthermore, the normalization table becomes editable after user logs in (Figure 3B 3–4).

Fig. 4.

RLIMS-P “Kinase view” partial results for CHK1 search. The “No. of sentences” column provides a quick link to evidence sentences for the specific annotation

Fig. 3.

Analysis of CHK1 phosphorylation text evidence for PMID:23748345. A. RLIMS-P text evidence view. The information can be filtered by the different sections of the article when applicable (1). The table shows kinase-substrate-site data, with each row displaying a unique information tuple with the sentence number and section source (2). The text panel on the right (3) contains the evidence text with the sentence numbers. The kinase, substrate, and site are color coded. The gene normalization table (4) shows possible UniProt identifiers for the kinases and substrates mentioned in the table. B. RLIMS-P table view when editing capability is unlocked. New columns appear: “Comment” for adding notes, and “Validation” for accepting/rejecting the annotation (1). Missing annotations can be added (2). Normalization data can be validated as well (3). If needed an auto-filled UniProt search can be triggered by clicking on the search icon with “UniProt” link in the “Name” column (this icon appears after hovering over anywhere in a given row) (4). Clicking on the link leads to the corresponding UniProt entries [26]. C. Partial view of the downloaded CSV format file. The file includes PMID, substrate, kinase, site, and evidence sentence

Another way to review the RLIMS-P results is to download them in CSV format, which could be done on a single article or on the selected collective result by clicking the Download button in the right corner of the Results page (Figure 1B 5). The file can be opened in Excel (Figure 3C) where you can filter or sort the information as needed. For example, you can download all results and then filter to i) show those where CHK1 is the substrate and ii) hide rows with “Blank” information in the “PTM enzyme” and “Site” columns (Figure 3C). The file contains the evidence sentences to assist you in validating the results (see Note ³).

The results can be summarized as in Table 2. RLIMS-P found all the sites and kinases cited in the review by Goto et al. [25], and in addition, RLIMS-P found an article describing a kinase not listed in that review, namely PIM1 (bold in Table 2).

Table 2.

CHK1 phosphorylation sites with kinases validated from RLIMS-P results (species non-specific)

Site	Kinase	PMIDs
Ser-280	P90 RSK AKT PIM1	19406993, 15710331, 22481935, 15107605, 12062056, 22357623, 23748345
Ser-286	CDK1 CDK2	20798862, 19837665, 22686412, 18983824, 16629900
Ser-296	CHK1	20639859, 22357623, 22686412, 23068608, 20053762
Ser-301	CDK1 CDK2	20798862, 19837665, 22686412, 18983824, 16629900
Ser-317	ATR ATM	21730979, 19625493, 20798862, 18723495, 20062519, 16629900, 16547171
Ser-345	ATR ATM	21289283, 20798862, 20976184, 15107605, 15159397, 22357623, 16629900, 22357623, 23383325, 16547171, 23422000, 11687578, 20053762, 17210576, 20609246

3.2 How to find all substrates for a given kinase

Because CHK1 is itself a kinase, we can easily identify all substrates of CHK1 by choosing the “View by Kinase” option (Figure 4). A variety of substrates are identified here under the column “Phosphorylated Protein (Substrate).” This column can be sorted using the arrow next to the title “Phosphorylation Protein (Substrate)” so that the information regarding the same substrate is brought together. Table 3 shows the summary of substrates of CHK1 and detected phosphorylation sites. Based on the number of articles linked to the substrates, CDC25 proteins seem to be the most widely studied CHK1 substrates.

Table 3.

Substrates of CHK1 and phosphorylation sites. These are manually validated results from RLIMS-P output. n/a indicates that the phosphorylated site is not described in the RLIMS-P output

CHK1 Substrates	Site	PMIDs
AURKB	Ser-331	22024163, 23321637
BLM	Ser-646	20719863
BRCA2	Thr-3387	24627786, 18317453
CDC10	n/a	24006488
CDC2	Tyr-15	24996846, 11479224
CDC25	Ser-287	9744884
	Ser-99	10198041
	n/a	9923681, 11133168, 15272308, 9774107, 10469601, 17912454
CDC25A	Ser-123	12399544, 12759351
	Ser-178,Thr-507	14559997
	Ser-73	12110582
	Ser-75	12759351
	Ser-76	14681206, 20348946, 18480045, 21252624, 20798862
	Thr-504	15272308
	n/a	12110582, 12399544, 20609246, 18414041, 24022480, 19244340, 21851590, 15272308, 19638579, 23272087, 21347609, 9278511, 18480045
CDC25B	Ser-230,Ser-563	17003105
	n/a	9278511, 10713667, 20798862
CDC25C	Ser-216	14681223, 9278511, 10676638, 11027648, 24922656, 15282313, 10557092, 22623962, 23874958, 20700484
	n/a	18272544, 10090724, 9278511, 11479224, 11925443, 15220526, 10681541, 22941630, 20798862, 21347609, 11278490, 10068474, 24038466
CDK1	n/a	20798862
CDKN1A	n/a	21791608
CDKN1C	n/a	21791608
	Ser-296	23068608, 20053762, 21289283, 24996846
CHK1	Ser-317,Ser-345	21851590
	n/a	14681223, 15371427, 23548269, 23593009, 19421147
CK1D	n/a	23861943
CK2	n/a	15225637
CLP1	n/a	22918952
CLSPN	Thr-916	16963448
	Ser-80	22792081
CRB2	Thr-73	22792081
CSNK1D	Ser-328, Ser-331,Thr-397	23861943
	Ser-328,Ser-331,Ser-370,Thr-397	23861943
	Ser-328,Thr-329,Ser-331,Ser-361,Ser-382	23861943
E2F6	n/a	23954429
ENOS3	Ser-1179	22001744
ERRFI1	Ser-251	22505024
FANCD2	n/a	21926477
FANCE	Thr-346,Ser-374	17296736
H2AFX	Ser-345	24913641
H2AX	Thr-16	20639511
KAP1	Ser-473	21851590
LATS2	Ser-408	21118956
	Ser-835	23886938
MAD2	n/a	23454898
MDMX	Ser-367	16511572
p33 (ING1b)	Ser-126	17585055
p50	n/a	22152481
PDS1	n/a	11390356, 17671432
RAD51	n/a	18317453
RAD9	n/a	24376897
RASSF1	Ser-184	24197116
RB1	n/a	17380128
RELA	Ser-612	15970704
	Thr-505	17962807
RPA1	n/a	16412704
SETMAR	n/a	25024738
	Ser-495	22231448
SYK	Ser-295	22585575
TAU	n/a	23550703
TLK1	n/a	12660173
	Ser-695	24376897, 12955071
TP53	Ser-20	15467443, 17339337
	Ser-23	23152407
	n/a	15659650, 11599922, 23272087
TP73	Ser-47	14585975
WEE1	Ser-549	11251070

3.3 How to find the interacting partners of phosphorylated proteins

In our examples in the following text, we address i) how phosphorylation on CHK1 affects its interaction with other proteins, ii) how PPIs are affected by proteins phosphorylated by CHK1, and iii) how phosphorylation of other proteins affect their interaction with CHK1. eFIP is capable of identifying the impact of phosphorylation, e.g., whether the phosphorylation enables the binding to a partner or inhibits the binding.

1. Go to the eFIP homepage (http://proteininformationresource.org/efip)

2. Enter the following protein names in the “Enter Protein Names and Type” query box: CHK1 OR CHEK1 OR “checkpoint kinase-1” and click Submit (Figure 2A 2). Note that the search can be restricted to retrieve results with CHK1 as a substrate, a kinase or an interactant (Figure 2A 3).

3. Select “Substrate View”. After submission, the result page (Figure 2B) displays the data in a summary view (as a list of entities detected that are grouped by PMID). Similar to RLIMS-P, by selecting “Substrate view” the information can be grouped by phosphorylated substrate, so that we can check the PPIs for phosphorylated CHK1.

4. Select “Kinase view” to investigate phosphorylation-dependent PPIs for CHK1 substrates. Review results for CHK1 as kinase, and check the information for interactant with its associated text evidence. Figure 5A depicts the text evidence for PMID: 14559997. In this particular case, the phosphorylation of CDC25A on Ser-178 and Thr-507 by CHK1 promotes the binding to 14-3-3 proteins. In addition to highlighting kinases, substrates, and sites using the same color scheme as RLIMS-P, interactants are highlighted in orange.

   You can also check for information about CHK1 as interactant, using the “Interactant view.”

5. Download the result table. Similar to RLIMS-P, eFIP results can be downloaded in CSV format by using the “Download Table” link in the upper left corner of the results table (Figure 2B). Table 4 provides a summary of results where CHK1 participated in a phosphorylation-dependent PPI either as the phosphorylated substrate or as the interactant. Table 5 provides a summary of phosphorylation-dependent PPIs where CHK1 acts as the kinase.

Fig. 5.

Analysis of CHK1 eFIP text evidence for PMID:14559997. A. The results can be filtered by section of the article (1). The table with extracted results shows the substrate, kinase (if known), phosphorylation site, the impact (inhibit, enhance, unknown association/dissociation), the interactant, the section of the article where the information comes from and the sentence number in that section (2). The gene normalization table (3) shows possible UniProt identifiers for the kinase, substrate, and interactant mentioned in the table. The text evidence panel (4) contains the evidence sentences with section title and sentence numbers. B. The Cytoscape representation of the relations extracted in the article

Table 4.

Summary of eFIP results for phosphorylation-dependent PPI involving CHK1

Substrate	PMID	Kinase	Site	Impact	PPI	Interactant
	12676962		Ser-345	enables	association	14-3-3
	12415000		Ser-345	enables	association	RAD24 (14-3-3 homolog)
	15585577			enables	association	RAD25 (14-3-3 homolog)
CHK1	20639859	CHK1	Ser-296	enables	association	CDC25A
	12676962		Ser-345	enables	association	chromatin
	16360315			enables	dissociation	chromatin
	23593009		Thr-125	inhibits	association	RAD9
	23593009		Thr-143	enables	association	RAD9
CRB2	22792081			increases	association	CHK1
	15707391		Thr-916, Ser-945	enables	association	CHK1
CLSPN	22792081			increases	association	CHK1
	12766152			unknown	association	CHK1
	12545175			unknown	association	CHK1

Table 5.

Summary of eFIP results for CHK1 substrates with phosphorylation-dependent PPIs

Substrate	PMID	Kinase	Site	Impact	PPI	Interactant
BRCA2	18317453	CHK1		enables	association	RAD51
CDC25	23166842	CHK1		increases	association	SCFBETATr CP
	22806395			enables	association	RAD24 (14-3-3 homolog)
CDC25A	14559997	CHK1	Thr-507,Ser-178	enables	association	14-3-3
	15272308		Thr-504	enables	association	14-3-3
	15272308		Thr-504	inhibits	association	Cdk1-cyclin A
						Cdk1-cyclin B
						Cdk2-cyclin E
CDC25B	20798862	CHK1	Ser-309, Ser-323	enables	association	14-3-3
CDC25C	23874958	CHK1	Ser-216	increases	association	14-3-3beta
	20798862		Ser-216	enables	association	14-3-3
RB1	17380128	CHK1	Ser-612	enables	association	E2F1
RAD51	18317453	CHK1		enables	association	BRCA2

3.4 Visualization of phosphorylation and interaction events in Cytoscape

eFIP also supports visual exploration of phosphorylation interaction networks using Cytoscape [28], which depicts in one graph a network of kinase-substrate relations, as well as PPI relations, including both the enhancement and inhibition of an interaction. Therefore, the phosphorylation-dependent interactions described in Subheading 3.3 can be displayed in Cytoscape.

Cytoscape view from text mining PMID evidence page

• 1Go to the eFIP homepage.
• 2Search for PMID:14559997. Enter the PMID in the search box (Figure 2A 4) and submit.
• 3Open the “Text Evidence” page. Click on the “hand” icon in the last column (Figure 2B 4) to see the text evidence (Figure 5A).
• 4Click on “See Cytoscape View”. The link to the Cytoscape view is at the top right of the evidence table (Figure 5A 5). The Cytoscape view for this example is shown in Figure 5B. CHK1 phosphorylates CDC25A at two residues. The phosphorylated residues enable interaction with 14-3-3.

Cytoscape view for multiple articles (see Note ⁴)

• 5Go to the eFIP homepage.
• 6Conduct query. For this case, enter the PMIDs 12676962, 17380128, and 20639859 separated by commas in the search box (Figure 2A 4) and then submit.
• 7Open “Cytoscape View.” The link to Cytoscape is on the top left of the result table (Figure 2B 5). The Cytoscape view for this example is shown in Figure 6.