To run a search in HubMed, enter the search terms into the text box found on either HubMed's front page, after running the bookmarklet, in Firefox's search box, or at the top of a page of search results.

When a search is run from the front page, you can choose whether the search results should be ordered by date (starting with the newest articles, as in PubMed) or by relevance (starting with articles that contain the search terms most frequently in the title and/or abstract). This ordering choice will be remembered for subsequent searches, but can be changed at any time by selecting the appropriate link from within a page of search results, or from the front page.

To use the bookmarklet, first drag the 'HubMed' bookmarklet link from the front page to your browser's bookmarks/favourites toolbar - this makes it available at any time. When needed, press this bookmarklet to open the search box, enter your search terms and press OK; this will launch a search in HubMed as normal.

If you're using Firefox, installing HubMed's search plugin from the front page will add HubMed to the list of available search engines.

There are two ways to modify existing search terms: the first is to click on the bold search terms listed at the start of a results page, which will turn them into a modifiable text box. Make changes as necessary, then press 'Search'. Alternatively, the 'Refine search' section at the end of a results page will show all the possible queries that were fitted to the search terms by PubMed (and the number of articles matched by each query). Uncheck any unwanted queries, add any extra terms--if neccesary--into the text box, then press 'Search'.

If you click the little 'expand search terms' icon next to the serch terms, this will look at the first 500 abstracts matching the search query and determine which words occur most frequently in those abstracts compared to all abstracts in the PubMed database. This can take a while, but when it's finished the most closely related terms will be displayed below the original search terms and can be added by clicking on each word. A boolean option allows search terms to be separated by 'AND' (thus narrowing the search) or 'OR' (thus broadening the search).

HubMed's search results are presented as a list of 20 items per page. To navigate the results, use the 'Next 20' and 'Prev 20' links to move a page at a time, or the drop-down list to move directly to another page.

To toggle the display of sentences from the abstracts that contain the search terms, press 'Show Extracts'. Extracts will remain visible until the 'Hide Extracts' button is pressed.

For each article of interest, the 'Abstract' link will load the abstract for that article inline. Alternatively, check the boxes next to multiple articles and press 'Display Checked Abstracts' to show the full details for all the selected articles.

While browsing through search results, pressing the 'Clip' link will add an article to the clipboard, which can be accessed at any time using the 'clipboard' button in the menu bar. The number next to the 'clipboard' button indicates the number of articles currently stored in the clipboard, and can be used to clear the clipboard (it turns into an 'x' when the mouse hovers over it).

There are three different ways to get to the full text of an article: the 'Fulltext' link uses PubMed's linkout service to try and locate the article; if this doesn't work, you'll end up at PubMed with just an abstract. The second option uses SFX to locate the article - this may be a direct link to the publisher's site, or may use a DOI (Digital Object Identifier). The final option, if all else fails, is a 'GS; link to Google Scholar, which can sometimes turn up copies of the article on other websites.

The 'Related' link attached to each item will show a list of articles that PubMed has determined to be closely related to the original article, based on keyword analysis. This list can also be viewed as a graph, which uses a TouchGraph Java applet to display connections between related articles.

To show a list of articles that have either cited or been cited by the current article, use the 'Cites' link. This list will probably be incomplete, unfortunately, as it only uses data provided from articles deposited in PubMed Central. The Cites page also has links to Google Scholar, which will probably have a bigger list of citations. If the article has a DOI, there will also be a link to Scopus, which has a large citation tracking database but requires a subscription for access.

To export citation metadata for an article, click the 'Export' link, then choose an export format from the list presented. If your software is able to handle accented characters, then tick the UTF-8 box. To use the chosen export format automatically in the future, tick the box to 'Remember this choice'. For exporting citation metadata for multiple articles, check the boxes next to each article then use the 'Export Citations' button at the bottom of the page and choose an export format as above.

The 'Terms' link will create links from phrases and words in the abstract to appropriate Gene Ontology, UniProt and Wikipedia pages.

To store articles that you find interesting, press the 'Tag' link. This will open a form in which you can add tags and comments for the article, then post it to HubMed's Tag Storage (in order to store articles, you need to register with the site to get a username and password). If anyone has already tagged the article, the number of people that have added tags will be shown alongside the 'Tag' link.

The TouchGraph applet requires that your computer has a version of Java (JRE) greater than version 1.4, and that the Java plugin for your web browser is installed. Once the applet is launched, you should see up to 20 articles displayed as nodes surrounding one central node - the original article. The small 'info' button attached to each node wil open a small window containing the title, authors and abstract for that article when clicked, as well as a link back to show that article in HubMed. Double-clicking on any of the nodes will fetch the 'related articles' data for that node and expand the graph accordingly. If no 'related articles' data is available (as is sometimes the case for very new articles), the node will display a small red circle.

Use the options at the top of the window to control the graph display, including zooming and increasing the number of displayed articles. The graph can be panned by dragging the mouse around after clicking on an empty section of the window.

Assuming you already have a feed reader in which to read feeds (if not, I recommend NetNewsWire for Mac OS X), you can subscribe to a feed after running a regular search. If your feed reader supports autodiscovery of feeds, then just give it the address of the search results page and it should find the feed automatically (you may see a feed icon show up in the address bar of your web browser, which is another quick way to subscribe). If not, then press the orange 'feeds' button in HubMed's menu bar, copy the address for either the Atom (preferred) or RSS (for older feed readers) feed and use this to subscribe.

The feed will contain the latest 20 items that match the search query. If there are ever more than 20 matching items, follow the feed's link back to HubMed to see all the results for that query. The journal title in each item is linked to the fulltext on the publisher's web site, where available, whereas the article title is linked back to HubMed's page for that article, from where other tasks can be performed.

NOTE: Due the large volume of requests from Bloglines and Google Reader to HubMed's server, often for feeds that no-one is reading any more, feeds are currently not being served to these sites. I suggest that you use a desktop feed reader, or a web-based feed reader that only fetches updates on-demand, i.e. when the feeds are actually read.

If you're using Endnote, then first you need to download the HubMed import filter from HubMed's front page and unzip it into Endnote's import filters folder. Then, from a page of search results, either choose 'Export' for a single article or 'Export Citations' for multiple checked articles. Check UTF-8 if you have version 9 or greater of Endnote, then choose the RIS export format. This will start the download of a RIS file, which you can save to disk and then import into Endnote.

For reference managers such as BibDesk or JabRef, choose the BibTeX export format. Depending on the document formatting tool you're using, you may want to check the UTF-8 option to preserve accented characters. The BibTeX output will load in a frame at the bottom of the window, and your browser should scroll there automatically; you can then select the BibTeX text and copy and paste it into your reference manager. If you prefer to download the citation metadata as a text file, use the 'Download BibTeX file' link found below the BibTeX output window

Citation data in RDF/XML (particularly useful for tools such as PiggyBank) and MODS formats is output as for RIS above, and will either be displayed in your browser or downloaded to disk.

To export citation metadata to RefWorks, either download citations in RIS format as above then upload the RIS file to RefWorks (for multiple articles), or choose the RefWorks export format to send a single article directly into RefWorks.

HubMed's 'Rank Relations' feature is a powerful way to recursively refine a collection of articles to get a focussed coverage of a particular subject area. Start with a normal HubMed search, base on some keywords and using either date- or relevance-ranked search results. Select those papers that are the most interesting and add them to the clipboard. Browse around related articles using the 'Related' links and have a look at other papers published by the main authors by clicking on their names. Add any more relevant papers that you find to the clipboard. Once there are a reasonable number of papers in the clipboard, open the clipboard using the 'clipboard' button in HubMed's menu bar. Select all the articles (using the 'All' checkbox at the top or bottom of the page), then press the 'Rank Relations' button. This will take a little while, as it is fetching the 'related articles' list for each of the selected articles, comparing the lists, and giving scores to articles that appear high up in each of the lists. When finished, it will return a list of the articles that were determined to be most closely related to as many as possible of the original selected articles. From this list, you can then proceed to check additional articles and repeat the 'Rank Relations' process until you have collected all the articles that you consider to be relevant. Pressing the 'Display Checked Abstracts' button will then display all those collected articles on one page, from where they can be saved, printed out or tagged and stored in HubMed's Tag Storage.