Tutorial: describing a simple experiment

In this section, we provide a basic example of creating a complete experiment descriptor using the ISA API’s model classes. The descriptor is not complete and realistic, but it demonstrates the range of component classes that you can use to create ISA content, including things like sample characteristics, ontology annotations and units.


As a pre-requisite to using ISA model please make sure you have read and understood the ISA Abstract Model that the ISA formats are based on.

Firstly, we need to import the ISA API’s model classes from the isatools PyPI package.

from isatools.model import *

Next, we build our descriptor encapsulated in a single Python function to simplify the example code. In a real application or script, you might decompose the functionality and hook it up to interactive components to solicit feedback from a user on-the-fly.

def create_descriptor():
    """Returns a simple but complete ISA-Tab 1.0 descriptor for illustration."""

    # Create an empty Investigation object and set some values to the instance variables.

    investigation = Investigation()
    investigation.identifier = "i1"
    investigation.title = "My Simple ISA Investigation"
    investigation.description = "We could alternatively use the class constructor's parameters to set some default " \
                                "values at the time of creation, however we want to demonstrate how to use the " \
                                "object's instance variables to set values."
    investigation.submission_date = "2016-11-03"
    investigation.public_release_date = "2016-11-03"

    # Create an empty Study object and set some values. The Study must have a filename, otherwise when we serialize it
    # to ISA-Tab we would not know where to write it. We must also attach the study to the investigation by adding it
    # to the 'investigation' object's list of studies.

    study = Study(filename="s_study.txt")
    study.identifier = "s1"
    study.title = "My ISA Study"
    study.description = "Like with the Investigation, we could use the class constructor to set some default values, " \
                        "but have chosen to demonstrate in this example the use of instance variables to set initial " \
    study.submission_date = "2016-11-03"
    study.public_release_date = "2016-11-03"

    # Some instance variables are typed with different objects and lists of objects. For example, a Study can have a
    # list of design descriptors. A design descriptor is an Ontology Annotation describing the kind of study at hand.
    # Ontology Annotations should typically reference an Ontology Source. We demonstrate a mix of using the class
    # constructors and setting values with instance variables. Note that the OntologyAnnotation object
    # 'intervention_design' links its 'term_source' directly to the 'obi' object instance. To ensure the OntologySource
    # is encapsulated in the descriptor, it is added to a list of 'ontology_source_references' in the Investigation
    # object. The 'intervention_design' object is then added to the list of 'design_descriptors' held by the Study
    # object.

    obi = OntologySource(name='OBI', description="Ontology for Biomedical Investigations")
    intervention_design = OntologyAnnotation(term_source=obi)
    intervention_design.term = "intervention design"
    intervention_design.term_accession = "http://purl.obolibrary.org/obo/OBI_0000115"

    # Other instance variables common to both Investigation and Study objects include 'contacts' and 'publications',
    # each with lists of corresponding Person and Publication objects.

    contact = Person(first_name="Alice", last_name="Robertson", affiliation="University of Life", roles=[OntologyAnnotation(term='submitter')])
    publication = Publication(title="Experiments with Elephants", author_list="A. Robertson, B. Robertson")
    publication.pubmed_id = "12345678"
    publication.status = OntologyAnnotation(term="published")

    # To create the study graph that corresponds to the contents of the study table file (the s_*.txt file), we need
    # to create a process sequence. To do this we use the Process class and attach it to the Study object's
    # 'process_sequence' list instance variable. Each process must be linked with a Protocol object that is attached to
    # a Study object's 'protocols' list instance variable. The sample collection Process object usually has as input
    # a Source material and as output a Sample material.

    # Here we create one Source material object and attach it to our study.

    source = Source(name='source_material')

    # Then we create three Sample objects, with organism as Homo Sapiens, and attach them to the study. We use the utility function
    # batch_create_material() to clone a prototype material object. The function automatiaclly appends
    # an index to the material name. In this case, three samples will be created, with the names
    # 'sample_material-0', 'sample_material-1' and 'sample_material-2'.

    prototype_sample = Sample(name='sample_material', derives_from=source)
    ncbitaxon = OntologySource(name='NCBITaxon', description="NCBI Taxonomy")
    characteristic_organism = Characteristic(category=OntologyAnnotation(term="Organism"),
                                     value=OntologyAnnotation(term="Homo Sapiens", term_source=ncbitaxon,

    study.samples = batch_create_materials(prototype_sample, n=3)  # creates a batch of 3 samples

    # Now we create a single Protocol object that represents our sample collection protocol, and attach it to the
    # study object. Protocols must be declared before we describe Processes, as a processing event of some sort
    # must execute some defined protocol. In the case of the class model, Protocols should therefore be declared
    # before Processes in order for the Process to be linked to one.

    sample_collection_protocol = Protocol(name="sample collection",
                                          protocol_type=OntologyAnnotation(term="sample collection"))
    sample_collection_process = Process(executes_protocol=sample_collection_protocol)

    # Next, we link our materials to the Process. In this particular case, we are describing a sample collection
    # process that takes one source material, and produces three different samples.
    # (source_material)->(sample collection)->[(sample_material-0), (sample_material-1), (sample_material-2)]

    for src in study.sources:
    for sam in study.samples:

    # Finally, attach the finished Process object to the study process_sequence. This can be done many times to
    # describe multiple sample collection events.


    # Next, we build n Assay object and attach two protocols, extraction and sequencing.

    assay = Assay(filename="a_assay.txt")
    extraction_protocol = Protocol(name='extraction', protocol_type=OntologyAnnotation(term="material extraction"))
    sequencing_protocol = Protocol(name='sequencing', protocol_type=OntologyAnnotation(term="material sequencing"))

    # To build out assay graphs, we enumereate the samples from the study-level, and for each sample we create an
    # extraction process and a sequencing process. The extraction process takes as input a sample material, and produces
    # an extract material. The sequencing process takes the extract material and produces a data file. This will
    # produce three graphs, from sample material through to data, as follows:
    # (sample_material-0)->(extraction)->(extract-0)->(sequencing)->(sequenced-data-0)
    # (sample_material-1)->(extraction)->(extract-1)->(sequencing)->(sequenced-data-1)
    # (sample_material-2)->(extraction)->(extract-2)->(sequencing)->(sequenced-data-2)
    # Note that the extraction processes and sequencing processes are distinctly separate instances, where the three
    # graphs are NOT interconnected.

    for i, sample in enumerate(study.samples):

        # create an extraction process that executes the extraction protocol

        extraction_process = Process(executes_protocol=extraction_protocol)

        # extraction process takes as input a sample, and produces an extract material as output

        material = Material(name="extract-{}".format(i))
        material.type = "Extract Name"

        # create a sequencing process that executes the sequencing protocol

        sequencing_process = Process(executes_protocol=sequencing_protocol)
        sequencing_process.name = "assay-name-{}".format(i)

        # Sequencing process usually has an output data file

        datafile = DataFile(filename="sequenced-data-{}".format(i), label="Raw Data File")

        # Ensure Processes are linked forward and backward. plink(from_process, to_process) is a function to set
        # these links for you. It is found in the isatools.model package

        plink(extraction_process, sequencing_process)

        # make sure the extract, data file, and the processes are attached to the assay

        assay.measurement_type = OntologyAnnotation(term="gene sequencing")
        assay.technology_type = OntologyAnnotation(term="nucleotide sequencing")

    # attach the assay to the study


To write out the ISA-Tab, you can use the isatab.dumps() function:

from isatools import isatab
return isatab.dumps(investigation)  # dumps() writes out the ISA as a string representation of the ISA-Tab

The function listed above is designed to return all three files as a single string output for ease of inspection. Alternatively you could do something like dump(isa_obj=investigation, output_path='./') to write the files to the file system.

Alternatively to write out the ISA JSON, you can use the ISAJSONEncoder class with the Python json package:

import json
from isatools.isajson import ISAJSONEncoder
# Note that the extra parameters sort_keys, indent and separators are to make the output more human-readable.
return json.dumps(investigation, cls=ISAJSONEncoder, sort_keys=True, indent=4, separators=(',', ': '))

The final lines of code is a main routine to invoke the create_descriptor() function.

if __name__ == '__main__':

If you save the file into something like createSimpleISA.py, to execute this script on the command line and view the output, you would run it with:

python createSimpleISA.py

This example can be found in the isatools.examples package in createSimpleISAtab.py and createSimpleISAJSON.py