Free Text Search

The listing operations are useful for listing the IDs and names, but not for searching through data content. For example, if we wish to search for an Experiment with associated metadata, we must list all the experiments, execute info and analyze the result of each experiment. Also, the list_samples command needs the correct metadata content to find the expected sample. In short, it is not an optimal way to search data. The search command is the optimal method for data searching in DeepBluebecause it simplifies the task of finding data.

The search command has three parameters: the free text for which to do search, the data type (optional), and the user_key. The data type defines the collection on which the search will be made. The available collections are: annotations, biosources, epigenetic_marks, experiments, genomes, projects, samples, sample fields, techniques, tilings, and column_types. Multiple data types can be combined, i.e., it is possible to search for the term "cancer" in all experiments and samples. The search command will search for the free text input in all data content. This means that the command will look at the data name, description, other data type attributes, and also at the extra metadata.

The search command result is a set of tuples, ordered by relevance, containing the data id, data name, and data type. The command returns a maximum of 100 values. The command returns a set containing the IDs and names; the info command offers more information about the retrieved items.

In the following example, we search for all samples containing either the word "cancer" or "blood" in their description:

server.search("cancer blood", "samples", user_key)

The following example searches for cancer or blood or methylation. This means that an experiment with only the word "methylation" in its metadata will also be returned. Experiments containing all three given words will have higher scores and will appear before those experiments with only one or two of the given words:

server.search("cancer blood methylation", "experiments", user_key)

More complex searches can also be performed:

• Words that must be in the metadata should be placed in double quotes
• Text can be exactly matched by enclosing in escaped double quotes, e.g. "white cells"
• The hyphen symbol (-) can be used to exclude a word

To search all experiments including "cancer" and "blood" and "methylation":

server.search("\"cancer\" \"blood\" \"methylation\"", "samples", user_key)

And, to search for "methylation" and not "histone" and not "modification" in the epigenetic marks:

server.search("methylation -histone -modification", "epigenetic_marks", user_key)

To search for "methylation" and "cancer" and not "rrbs" and not "histone" and not "modification" in experiments:

server.search("\"methylation\" \"cancer\" -rrbs -histone -modification ",
              "experiments", user_key)

Please be aware that the terms inside \" \" are searched for exactly as they are entered, turning off the default functionality of searching for similar terms.

More Examples

Search can be used together with the info command:

(s, result) = server.search("methylation  -histone -modification cancer ",
                            "experiments", user_key)
for r in result[:3]:   # Select first three elements
    print server.info(r[0], user_key)

In this example, we search for experiments whose metadata does not contain the term "rrbs":

(s, result) = server.search("methylation -rrbs  -histone -modification cancer ",
                             "experiments", user_key)
for r in result[:3]:   # Select first three elements
    print server.info(r[0], user_key)