1. How to search proteins?

    There are two ways to search for a protein, text keyword search and sequence similarity search. A protein can be queried by its name, AGI code, NCBI RefSeq accession, or UniProt accession in the Protein Search page. Sequence similarity search can be performed in the BLAST page.

2. How to search protein interactions?

    If you are interested in one protein and want to know what other proteins it may interact with, we suggest that you search for this protein by keyword search or sequence search. The interactions involving this protein will be listed on the Protein Information page.
    If you have a list of proteins and want to know whether they interact between each other, you can paste this list of proteins into the input box in the Interaction Search page, and select "Show only pair-wise interactions between the specified proteins".
    If you have a list of proteins and want to know all the interactions involving these protein, you can paste this list of proteins into the input box in the Interaction Search page, and select "Show all interactions involving specified proteins".

3. How to search GO term interactions?

    If you are interested in one GO term and want to find what other GO terms it significantly links to, you can search the GO term at the "Search GO interactions (Single GO term)" page.
    If you have a list of GO terms and want to know whether they link to each other, you can paste this list of GO terms into the input box in the "Search GO interactions (Multiple GO terms)" page, and select "Show only pairwise interactions between the specified GO" .
    If you have a list of GO terms and want to know whether they link to the other GO terms, you can paste this list of GO terms into the input box in the "Search GO interactions (Multiple GO terms)" page, and select "Show all interactions involving specified GO".
    On the result page, there will be a grid showing all the GO terms that significantly linked to the query GO. In addition to browsing the list, you can search the list by pressing the bottom left search icon other functions.

4. How to search predicted protein functions?

    If you are interested the prediction of one protein, you may search it like searching a protein in the Protein Function Search page.

5. How to perform a gene set linkage analysis?

     If you have one set of Arabidopsis genes and you want to know if it interacts with some GO biological processes. You can submit this gene set in the Gene Set Text format on the Gene Set Linkage Analysis page. Our server will assign a job ID for each submitted gene set. You will receive the gene set linkage analysis results by email. If there is a problem with the email, you can also retrieve your results by the job ID at the Gene Set Linkage Analysis page.

    A Gene Set Text file starts with one optional annotation line, which is marked by a "#". There can be only one annotation line, or none. After the annotation line, there are identifiers separated by tab. The first identifier is the name of the submitted gene set. Following it, the genes in this gene set are represented by their AGI loci.

6. What is the the p Value in the set interaction?

    Significant linkages between gene sets were detected using a method similar to the one described by Li et al. (2008). For each pair of gene sets, the number of inter set protein interactions was first counted. A common gene shared by two gene sets was treated as two distinct delegate genes, with each delegate gene belonging to only one gene set. Any gene that interacted with the shared gene was considered to interact with both delegate genes in both gene sets. The fact that two gene sets sharing a gene was not considered an inter set interaction. Then, the significance P value of the linkage between a pair of gene sets was calculated using the method outline here. The interaction network was randomized maintaining the same topology. Every gene in the network was swapped with a random gene with the same number of interactions. With 10,000 randomizations, the significance P value was the fraction of randomized networks in which the number of inter set interactions was larger than that in the original interaction network.

7. What is the SVM score?

    The SVM score is an indicator of the prediction confidence. It is actually the direct output of the SVM prediction function. According to the theory of SVM, all scores greater than 1 is considered equally confident. In general, the higher the score, the more confident a prediction is. However, this score cannot be directly translated to the probability of true interaction. We are now looking for a robust way to calculate more strait-forward per-interaction accuracy measurement.
    A negative SVM score means a non-interaction prediction. PAIR (V3.3) now includes 5590 experimentally reported interactions collected from the major interaction databases and 145,494 interactions predicted by the PAIR V3 prediction model. The prediction model is not perfect. It was estimated to recognize 24.47% of the real interactions at the time it was constructed. In fact, the predicted interactions cover 26.4% of the 5990 recently compiled experimentally reported interactions. This means, 73.6% of the reported interactions were wrongly predicted as non-interactions with negative SVM scores. So the interactions in PAIR V3.3 with negative scores are experimentally reported interactions that are not recognized by the PAIR V3 prediction model.

8. How to use the embedded network visualizer (Cytoscape Web)?

    Cytoscape Web has been embedded in many PAIR pages to display interaction networks graphically. It is a flash-based network visualizer, which interacts with users via JavaScript API. Therefore, in order to properly operate this network visualizer, it is necessary to enable JavaScript support and install Adobe Flash Player.
    As shown in Figure 1, the network visualizer displays a graph consisting of user-specified interactions. The nodes are proteins and the edges represent interactions between two proteins. Nodes are colored according to their GO functions. The color scheme could be viewd by clicking the "Info/Tips" button in the options control. Mouse-over on nodes and edges will display annotation information about the proteins and interactions. To expand the network, double click a node to include its neighbors. You can change the layout by clicking the "Layout" button under the graph and applying a suitable layout algorithm. Or you can drag the nodes to right positions. A graph control panel is also placed in the right bottom corner, which you can use to zoom or move the graph. Right click on the nodes and edges will bring up context-sensitive menu to let you manipulate this network. One useful right-click menu function for edges is "Add this interaction to My Collection".

9. How to export a network of interactions?

    Click the "Export" button of the graphical interaction network browser will allow you to select a desired output format and export the interactions in the network browser in that format. Currently, we support the exportation in the Microsoft Excel CSV format, Cytoscape SIF format, GraphML format, PSI-MI 2.5 XML format, and our PAIR XML format. The network graph can also be exported as image in the PNG and PDF format.

10. What is "My Collection"?

    "My Collection" is a temporary personalized storage place where you can store your interested interactions and export them as a single data file for future analysis. Interactions may be added to "My Collection" from the "Interaction Information" page, by clicking the "Add to My Collection" button; or be added from the "Protein Information" page, by selecting interactions in the interaction table and clicking the "Add selected interactions to My Collection" button. Interactions can also be added from any network graph, by right clicking an edge and selecting "Add this interaction to My Collection". Please note that interactions in "My Collection" are stored in local browser cookies. This means that you need to enable cookie for the PAIR website, or the "My Collection" function will not work properly. Moreover, the collection of interactions will be erased as soon as you close the browser. The "My Collection" page is here.

11. How to cite us?

PAIR prediction algorithm:

Lin M, Hu B, Chen L, Sun P, Fan Y, Wu P, and Chen X. (2009). Computational identification of potential molecular interactions in arabidopsis. Plant physiology, 151(1):34-46. (PubMed)

PAIR database:

Lin M, Shen X, and Chen X. PAIR: the predicted Arabidopsis interactome resource. (2011) Nucleic Acids Research, 39(Database issue):D1134-1140. (PubMed)

PAIR for systems biology analyses:

Lin M, Zhou X, Shen X, Mao C, Chen X. The Predicted Arabidopsis Interactome Resource and Network Topology Based Systems Biology Analyses (2011) The Plant Cell, 23(3):911-922. (PubMed)