Managing Datasets in Galaxy
Datasets are the inputs and outputs of each step in an analysis project in Galaxy. Datasets are associated with at least one History, which can be labeled, manipulated, and shared with anyone, whether they have a Galaxy account or not. Watch the "Datasets" video
The tracking information associated with Datasets in a History represent an experimental record of the methods, parameters, and other inputs. These methods are easily extracted into Workflows, making an analysis pathway transparent, reproducible, and reusable.
You have multiple options how to get your files into Galaxy thus making them datasets:
- Upload modal - Interface within Galaxy that suits the best for uploading small files from disk or fetching data from other servers. You can reach it by clicking on its icon (right picture) in the tool panel.
- FTP upload - In case of large files (the upload modal has ~2GB browser limit) or unpredictable connection (support for pausing and resuming) you might want to use FTP. The Galaxy server you want to upload data to has to have an FTP service configured (both Main and Test instances do). See more details at FTPUpload.
- Upper right corner
- Display data in browser "eye icon"
- Edit attributes "pencil icon"
- Delete "'X' icon"
- Lower right corner
- Edit dataset tags
- Edit dataset annotation
- Upper left corner
- Dataset name
- Dataset size/number of lines (actual or estimated)
- format datatypes
- Info (optional)
- Lower left corner
- The size limit for a file loaded using FTP is 50G.
- The size limit for a job's output is (unrelated to quotas):
- The size limit for all data (quotas) on the Galaxy public servers is explained at:
- Administrative instructions for disk quotas
- The format of a dataset is ideally defined by the assigned datatype attribute. Deviations in input dataset format are the first variable to examine when a tool (job) fails. Many of the tools in the "Text Manipulation" tool group can be used to both examine and correct a dataset's format to bring it into alignment with the assigned datatype attribute specification.
- To initially assign a dataset's datatype attribute, the uploaded/imported file can be specified with some import tools or be named with the appropriate file extension. To specify, modify or correct a dataset's datatype attribute after upload, click on the "pencil" icon in the right corner of the dataset's box to reach the "Edit Attributes" form. Use the "Change data type" section of the form to make changes and click on Save. Galaxy will modify the datatype and metadata.
- To transform a dataset format (original → new datatype attribute), use one of the many tools in the Convert Formats group.
- TIP The quickest way to locate tools that manipulate specific formats is to use the Tool Search (top of left Galaxy Tool panel, gear icon menu). For example, type in M-A-F to locate tools in the tool group Convert Formats that transform to/from Multiple Alignment Format.
- For many datatypes, clicking on the eye icon for "Display data in browser" will display the contents or a preview of the contents in as unformatted text in the center pane (exceptions include compressed datatypes such as BAM).
- Direct links to view a dataset within a browser may include:
- To copy the datasets within a history to another history, from the right history pane's top Options menu select Copy Datasets. On the form in the center pane, specify the From and To history/histories.
- From: Select the datasets to be copied in the left column Source History:.
- To: Select the location to copy the datasets in the right column Destination History:.
- Options include a single existing history, multiple existing histories, or a newly created and named history.
- TIP to Copy a Hidden dataset (see below), in the From histories right pane, use gear icon → Unhide Hidden Datasets, then once the datasets refresh, use This dataset has been hidden. Click here to unhide.
- To clone a history is to create an exact copy of the prior history in one step. The new history will be named with the original history's name prefixed by Clone of. Clone is the simplest way to manage datasets when some items in a history need to be retained but the remainder can be deleted (permanently, to reduce disk usage).
- Options are:
- Clone all history items, including deleted items
- Clone only items that are not deleted
- TIP One use of this option is to quickly retain some datasets and permanently delete others (to reduce disk use counted in user quota on Main or Test). First, in the History pane, in the original history, delete individual datasets by clicking on the X delete icon if not to be Cloned, remember to delete Hidden datasets, (see below). Next, Clone the original History. Once complete, the cloned History will contain the datasets to be retained and the original History can be deleted permanently with gear icon → Saved Histories, select original History from the list, and clicking the button Delete Permanently.
- Datasets may be hidden in the default History view as a Workflow option. If you have run a workflow with hidden datasets, choose "gear icon → Include Hidden Datasets or Unhide Hidden Datasets" or use the toogle at the top of the history panel (directly below the history name) to view them.
- When using Clone (see above) to manage datasets to reduce disk usage for quotas, viewing and deleting hidden datasets can be a very important step. Unless deleted, hidden datasets are moved to the new cloned history.
- When using Copy (see above) to manage datasets to reduce disk usage for quotas, hidden datsets will not be in the "From" list of datasets available to transfer unless they are unhidden using gear icon → Unhide Hidden Datasets, then This dataset has been hidden. Click__here_ to unhide.
- Deleting Datasets and Histories
- Watch how it works in the Managing Histories video.
- Deleted datasets and histories can be recovered by users as they are retained in Galaxy for a time period set by the instance administrator. For the Galaxy public instances Main and Test, this is currently several months.
- Permanently deleted datasets and histories cannot be recovered by the user or administrator.
- Deleted datsets can be undeleted or permanently deleted using from the History pane gear icon → Include Deleted Datasets, and then: This dataset has been deleted. Click here to undelete or here to immediately remove it from disk.
- Check for hidden datasets and delete as needed (see section above Hidden for more details)
- Quotas for Datasets and Histories
- Deleted datasets and deleted histories containing datasets are considered when calculating quotas on Main or Test.
- Permanently deleted datasets and permanently deleted histories containing datasets are not considered.
- Imported native Data Library datasets are not considered.
- Datasets can be associated with one or more History, but are only considered once.
- All copies of a dataset must be permanently deleted for it to not be considered.
- Histories/datasets that are shared with you are only partially considered unless you import them.
- Active and Deleted histories can be permanently deleted using from the History pane Options → Saved Histories, then click on Advanced Search, then click on status: all. Check the box for the histories to be discarded and then click on the button Permanently delete.
- WARNING Permanently deleted datasets and histories cannot be recovered by the user or administrator. The best way to avoid losing important data by accident is to clearly name all histories and important datasets.
- Name a dataset:
- Click on the pencil icon in the right History pane) to reach the Edit Attributes form. Here a dataset's primary Name, Info , Annotation, and Notes can be adjusted.
- TIP Copying the Galaxy default Name into the "Info: field, then adding in a custom Name is one way to preserve the tool output original Name while still distinguishing one similarly named dataset from another. This can be useful when reviewing analysis steps and choosing which datasets to retain and which to remove when an analysis is under review or completed.
- Name a history:
- Click near the top of the right history pane where the default text Unnamed history is located. Enter the new name and save.
- From the History pane use Options → Saved Histories, check the histories (one or more) to be renamed, then click on the bottom button Rename. On the Rename form, Current Name is on the left, New Name is on the right. Edit New Name for each history then click on the button Rename Histories.
- Name a dataset:
As of commit 11591, you can search your datasets in a number of ways:
- Open the search by clicking the magnifying glass next to the right of the history name. A search bar will open allowing you to enter your search terms.
- Type any text in the search bar that may help to narrow your search (advanced options are described below) and press enter/return. The list of datasets below the history title will change to include only those that match the search term and exclude those that don't.
- Clear the search by removing the text in the bar and pressing enter, pressing the ESC key while the bar is highlighted or in focus, or by pressing the clear search button on the right of the bar.
- Close the search bar by pressing the magnifying glass button again.
- Searches are case insensitive: 'some' will match both 'some' and 'Some'.
- Searches will persist until cleared. If you switch histories while searching, the list of datasets will still be narrowed to what matches your search terms.
- Terms are space separated. For example, to search for an interval datatype dataset named ~MyDataset when more than
one interval dataset is present and more than one dataset is named ~MyDataset, use
MyDataset interval- order doesn't matter).
- Deleted and hidden datasets are not shown unless you've included them by using the Include deleted datasets and/or Include hidden datasets options in the history options menu (the gear icon at the top left of the history panel).
- To search for terms that include whitespaces (e.g. a dataset named 'My Dataset' which includes a space) you can enclose a search term with double quotes ("My Data") and this will match the dataset in the example but exclude any dataset that may have matched an un-enclosed term (such as '~MyInterval' or 'Some Data').
- Search terms are applied to many of the 'fields' that describe a dataset (i.e. name, format, database, info, 'blurb') and, by default, check all fields against each term. To apply a search term to only one field, use the field name followed by an equals sign and then followed by the term (or double quote enclosed term). E.g.
name="My Dataset" format=interval. Here is a list of field names that can be used:
title: the name or partial name of the dataset(s)
genome_build: the name or partial name of the genome database (e.g. hg19)
file_ext: the file type/datatype of the dataset(s)
description: the text or partial text of description of the dataset(s) - the text just below the dataset title shown when the dataset is expanded.
info: the text or partial text of the info box of the dataset(s) - the text just below the format and database and above the download and info buttons shown when the dataset is expanded.
annotation: the text or partial text of your annotation on the dataset(s)
tag: the text or partial text of any applied you to the dataset(s)
- The field name specifiers can be applied more than once. For example, to find datasets that have both the 'paper1' and '2nd paper' tags (as opposed to those that have one or the other), use
tag=paper1 tag="2nd paper".