Retrieving Data

🤓 The ultimate guide for data analysts using Isabl!

Introduction to Filters

Filters enable you to subset the data of your interest. For example you can use filters to retrieve all the BAM files of a given project, or get all VCFs from a given variant calling application. Filters are field-value pairs and can be used both in the Command Line and within Python. Check out this examples:

# A request to the API

curl http://{my-isabl-instance}/api/v1/experiments?sample__identifier={the-sample-id}

# Using Isabl CLI

isabl get-outdirs -fi application.name BWA_MEM -fi status SUCCEEDED

# Using Isabl SDK

import isabl_cli as ii
samples = ii.get_instances('samples', individual__species="HUMAN")

Note that fields can traverse the relational model. To do so concatenate the fields with __ (e.g. samples__disease__acronym=AML, or a dot in the Command Line application.name=PINDEL)

Filters Modifiers

As indicated in the previous hint, filter fields can traverse database relationships. However, all filters can be augmented using lookups:

Related Filters

{related_field}__{related_filter}="query"

Here is a quick representation of Isabl's relational model, hence related filters:

Lookup Types

Furthermore, all query parameters in this API support advanced lookup types:

Lookup Type	Description	Example
!	Negate any query	name!=isabel
[i]exact	Exact match	name__exact=isabl name__iexact=IsAbL
[i]contains	Value contains query	name__contains=isa name__icontains=iSa
[i]startswith	Value starts with query	name__startswith=isab name__istartswith=iSab
[i]endswith	Value starts with query	name__endswith=bl name__iendswith=bL
in	Comma separated query	name__in=isabl,besuhof
isnull	Value is null	name__isnull=true
regex	Use regex pattern	name__regex=isabl
gt	Greater than	total__gt=1
gte	Greater or equal	total__gte=1
lt	Less than	total__lt=1
lte	Less than or equal	total__lte=1

Datetime Lookups

Moreover, Datetime query parameters support extra lookups:

Lookup Type	Description	Example
``	No lookup, ISO format required	created=
date	Filter by date YYYY-MM-DD.	created__date=2016-06-04 created__date__gt=2016-06-04
day	Filter by day DD	created__day=04
month	Filter by month MM	created__month=06
year	Filter by year YYYY	created__year=2016
time	Filter by time HH-MM-SS	created__time=21:00:51

Full Relational Model

UML Diagram of the the db schema

To get a full description of all available filters please visit Isabl's Redoc API documentation at https://isabl.github.io/redoc/ or https://isabl.mskcc.org/api/v1 (replacing isabl.mskcc.org with your own host). Another useful way to explore the relational model is by using isabl get-metadata.

Common Filters

Here are some common and useful filters for Isabl.

Limit vs Count Limit

The filter count_limit enables you to limit the total number of instances that will be retrieved. For example to get the output directory for the first 10 successful analyses you could do:

isabl get-outdirs -fi status SUCCEEDED -fi count_limit 10

On the other side, limit will determine how many instances should be retrieved at the same time. For example, the following command would retrieve paths to all successful analyses in batches of 10000:

isabl get-outdirs -fi status SUCCEEDED -fi limit 100000

Has BAM File

To get for example all experiments that have a BAM file for GRCh37 you could do:

experiments = ii.experiments(has_bam_for="GRCh37")

Performance Filters

The following filters can be used to (quite dramatically) improve the performance for some queries:

Filter

Description

Usage

distinct

If you set distinct to false, the each result within the query won't be guaranteed to be unique, yet the response will be faster.

distinct=false

paginator

By activating the cursor pagination, you would be able to traverse queries results, but you won't know the total number of results.

paginator=cursor

paginator=cursor is still experimental, please report an issue if you have trouble.

Isabl Command Line Client

Filters in the command line are usually provided using the -fi or --filters flags. Relations or lookups can be provided using double underscores or dots (e.g. application.name or application__name). Here is a list of Isabl commands available to retrieve information:

Command	Description	Example
get-count	Get count of database instances given a particular query. For example, how many failed analyses are in the system?	isabl get-count analyses -fi status FAILED
get-metadata	Retrieve instances metadata in multiple formats. To limit the number of fields you are interested in use -f (i.e. --fields).	isabl get-metadata samples -fi category TUMOR -f disease
get-data	This command will retrieved the raw data linked to experiments as imported in Isabl (e.g. BAM, FASTQ, CRAM). Use --verbose to see what experiments have missing data.	isabl get-data -fi projects.pk.in 102,103
get-bams	Get the official bam registered for a given list of experiments. Use --assemblyif there are BAMs available for different versions of the genome. Use has_bam_for to filter those experiments with a registered BAM.	isabl get-bams -fi has_bam_for GRCh37
get-reference	Isabl supports the linkage of auxiliary resources to the assembly instances. By defaultget-reference gives you the path to the reference FASTA, however you can retrieve other linked resources.	isabl get-reference GRCh37
get-bed	Retrieve a BED file linked to a particular sequencing technique. By default, the targets BED file is returned, to get the baits BED use --bed-type.	isabl get-bed HEMEPACT --assembly GRCh37
get-paths	Retrieve the storage directory for any instance in the database. Use --pattern to retrieve files within those directories.	isabl get-paths projects 102
get-outdirs	This command is a short cut of isabl get-paths analyses. Learn more about retrieving results here.	isabl get-outdirs -fi name PINDEL -fi status SUCCEEDED
get-results	Retrieve analyses results produced by applications. Use --app-resultsto list all available choices for a given application.	isabl get-results -fi application.pk 1 -r command_script

Dynamically Explore Metadata

Another useful way to explore the relational model is by using isabl get-metadata:

isabl get-metadata experiments --fx

Expand and navigate with arrow keys, press e to expand all and E to minimize. Learn more at fx documentation. Use --help to learn about other ways to visualize metadata (e.g. tsv).

Furthermore, you can limit the amount of information you are retrieving by passing the list of fields you are interested in:

isabl get-metadata analyses -f application.name -f status

Assembly Resources

By default, the command get-reference helps you retrieve the assembly reference genome.

isabl get-reference GRCh37    # retrieve reference genome

However, by means of the --data-id flag, the command get-reference also allows you to retrieve the indexes generated during import. To get a list of available files per assembly use --resources:

$ isabl get-reference GRCh37 --resources

    genome_fasta         Reference Genome Fasta File.
    genome_fasta_fai     Index generated by: samtools faidx
    ...

Then get the one you are interested in with:

isabl get-reference GRCh37 --data-id genome_fasta_fai

Retrieving Application Results

You can use get-outdirs within the command line to systematically explore output directories. For example:

isabl get-outdirs -fi status FAILED | xargs tree -L 2

Further more you can retrieve files within those directories by using --pattern:

isabl get-outdirs -fi status FAILED --pattern 'head_job.*'

Additionally, you can retrieve results directly registered by the application:

for i in `isabl get-results -fi status FAILED -r command_err`; do
   echo exploring $i;
   cat $i;
done

To visualize what results are available for a given application run:

isabl get-results --app-results <application primary key>

You can retrieve the application primary key from the front end.

Isabl Software Development Kit

Importantly, isabl-cli can also be used as a Software Development Kit within python:

Try from an ipython session

import isabl_cli as ii  # ii stands for `interactive isabl` 😎

If you are using ipython, use ? to get help on a method (e.g. ii.get_instances?)

Getting Instances

To get started, we can retrieve specific instances from the database:

# retrieve an experiment with a system id (primary keys also work)
experiment = ii.Experiment("DEM_10000_T01_01_TD1")

# we can also get an analysis using it's primary key (we'll limit retrieved fields)
analysis = ii.Analysis(10235, fields="status,application")

# same thing for assemblies
assembly = ii.Assembly("GRCh37")

These instances are Munch, in other words they are dot-dicts (like javascript). So you can do both analysis["status"] or analysis.status.

A more general way to retrieve any object in the database is using get_instance:

project = ii.get_instance("projects", 100)  # the signature is (endpoint, identifier)

Some examples of things you can do with these instances:

# get the target experiment or tumor
target = analysis.targets[0]

# print the experiment sample class
print(experiment.sample.category)

# see available analysis fields
print(analysis.keys())

# see all available data
print(vars(assembly))

To get multiple instances you can do:

# get all TUMOR experiments in project 102
experiments = ii.get_experiments(projects=102, sample__category="TUMOR")

# get the first 10 SUCCEEDED analyses in the same project
analyses = ii.get_experiments(projects=102, status="SUCCEEDED", count_limit=10)

# get all the projects where I'm the owner
projects = ii.get_projects(owner__startswith="besuhof")

Similarly to isabl get-count , you can determine the number of available results for a given query:

ii.get_instances_count("analyses", status="FAILED")

Getting all Samples from an Individual

To retrieve all samples and experiments for a given individual:

# you can also use the individual system_id
individual = ii.get_tree(10000)

# them all samples are available at
samples = individual.sample_set

# and all experiments for a given sample
experiments = samples[0].experiment_set

You can also retrieve multiple trees:

individuals = ii.get_trees(projects=267)

Create, Delete, and Modify Instances

If you have permissions, you will be able to systematically alter instances in the database:

# create a disease
ii.create_instance("diseases", name="Osteosarcoma", acrynom="OS")

# update an individual's gender
ii.patch_instance(individual.pk, gender="UNKNOWN")

# delete an analysis
ii.delete_instance("analyses", analysis.pk)

With great power, comes... yeah you know how it goes. Just be careful.

Isabl SDK Utils

Here are other useful utilities available in isabl-cli:

Method	Description
ii.api.chunks	Given a list of elements, return a list of chunks of n elements from the original list.
ii.api.api_request	Perform an authenticated request to Isabl API.
ii.api.retry_request	Retry an HTTP request multiple times with a delay between each failure.

Isabl Web

Isabl Web is a great tool to retrieve information and understand the state of affairs within the system. Simply type something in the search bar to retrieve instances across multiple schemas:

Multiple panels will be stacked horizontally as you request more information:

image

Projects Detail Panel

The projects detail panel conveys all assets and stakeholders linked to a particular project:

Live Tables are directly wired to the API and will enable you to search and filter on specific columns. For the later, simply click in the column name:

Directly searching on the sample Identifier column.

The Samples View

The samples tree panel provides access to all assets generated on a given individual.

A data generation process tree that resulted in 4 experiments (or ultimately bams), produced from two samples of the same individual.

By clicking on a given node in the tree, you can retrieve more metadata, filter out available analyses on that instance, and even get access to BAM files:

Although the BAM file is an output of the bwa-mem analysis, Isabl enables registering default bams to an experiment. Thus a link is available in the sample panel.

Analyses Results

We can retrieve different types of results for all analyses generated by Isabl. For example accessing a project level quality control report:

A project level Quality Control report. Can you find the Experiment and Individual-level reports?

Similarly we can retrieve other types of results such as a VCF: