Project flow

LaminDB allows tracking data flow on the entire project level.

Here, we walk through exemplified app uploads, pipelines & notebooks following Schmidt et al., 2022.

A CRISPR screen reading out a phenotypic endpoint on T cells is paired with scRNA-seq to generate insights into IFN-γ production.

These insights get linked back to the original data through the steps taken in the project to provide context for interpretation & future decision making.

More specifically: Why should I care about data flow?

Data flow tracks data sources & transformations to trace biological insights, verify experimental outcomes, meet regulatory standards, increase the robustness of research and optimize the feedback loop of team-wide learning iterations.

While tracking data flow is easier when it’s governed by deterministic pipelines, it becomes hard when it’s governed by interactive human-driven analyses.

LaminDB interfaces workflow mangers for the former and embraces the latter.

Setup

Init a test instance:

!lamin init --storage ./mydata

Show code cell output Hide code cell output

✅ saved: User(uid='DzTjkKse', handle='testuser1', name='Test User1', updated_at=2023-12-22 11:25:24 UTC)
✅ saved: Storage(uid='3O6YyjbU', root='/home/runner/work/lamin-usecases/lamin-usecases/docs/mydata', type='local', updated_at=2023-12-22 11:25:24 UTC, created_by_id=1)
💡 loaded instance: testuser1/mydata
💡 did not register local instance on hub

Import lamindb:

import lamindb as ln
from IPython.display import Image, display

💡 lamindb instance: testuser1/mydata

Steps

In the following, we walk through exemplified steps covering different types of transforms (Transform).

Note

The full notebooks are in this repository.

App upload of phenotypic data

Register data through app upload from wetlab by testuser1:

# This function mimics the upload of artifacts via the UI
# In reality, you simply drag and drop files into the UI
def run_upload_crispra_result_app():
    ln.setup.login("testuser1")
    transform = ln.Transform(name="Upload GWS CRISPRa result", type="app")
    ln.track(transform)
    output_path = ln.dev.datasets.schmidt22_crispra_gws_IFNG(ln.settings.storage)
    output_file = ln.Artifact(
        output_path, description="Raw data of schmidt22 crispra GWS"
    )
    output_file.save()


run_upload_crispra_result_app()

Show code cell output Hide code cell output

💡 saved: Transform(uid='6qT7fLNZYwe7sz', name='Upload GWS CRISPRa result', type='app', updated_at=2023-12-22 11:25:26 UTC, created_by_id=1)

💡 saved: Run(uid='wN9HK7PuMGx1ADxQKeO7', run_at=2023-12-22 11:25:26 UTC, transform_id=1, created_by_id=1)

Hit identification in notebook

Access, transform & register data in drylab by testuser2:

def run_hit_identification_notebook():
    # log in as another user
    ln.setup.login("testuser2")

    # create a new transform to mimic a new notebook (in reality you just run ln.track() in a notebook)
    transform = ln.Transform(name="GWS CRIPSRa analysis", type="notebook")
    ln.track(transform)

    # access the upload artifact
    input_file = ln.Artifact.filter(key="schmidt22-crispra-gws-IFNG.csv").one()

    # identify hits
    input_df = input_file.load().set_index("id")
    output_df = input_df[input_df["pos|fdr"] < 0.01].copy()

    # register hits in output artifact
    ln.Artifact(output_df, description="hits from schmidt22 crispra GWS").save()


run_hit_identification_notebook()

Show code cell output Hide code cell output

💡 saved: Transform(uid='8UtqUKuTeTw98Y', name='GWS CRIPSRa analysis', type='notebook', updated_at=2023-12-22 11:25:29 UTC, created_by_id=1)

💡 saved: Run(uid='3MAACmmTQ69HVjripVLE', run_at=2023-12-22 11:25:29 UTC, transform_id=2, created_by_id=1)

Inspect data flow:

artifact = ln.Artifact.filter(description="hits from schmidt22 crispra GWS").one()
artifact.view_flow()

Sequencer upload

Upload files from sequencer:

def run_upload_from_sequencer_pipeline():
    ln.setup.login("testuser1")

    # create a pipeline transform
    ln.track(ln.Transform(name="Chromium 10x upload", type="pipeline"))
    # register output files of the sequencer
    upload_dir = ln.dev.datasets.dir_scrnaseq_cellranger(
        "perturbseq", basedir=ln.settings.storage, output_only=False
    )
    ln.Artifact(upload_dir.parent / "fastq/perturbseq_R1_001.fastq.gz").save()
    ln.Artifact(upload_dir.parent / "fastq/perturbseq_R2_001.fastq.gz").save()


run_upload_from_sequencer_pipeline()

Show code cell output Hide code cell output

💡 saved: Transform(uid='p1xxYqcvrHW1xg', name='Chromium 10x upload', type='pipeline', updated_at=2023-12-22 11:25:30 UTC, created_by_id=1)

💡 saved: Run(uid='lQnw1VQHzVmluFfZmiZT', run_at=2023-12-22 11:25:30 UTC, transform_id=3, created_by_id=1)

scRNA-seq bioinformatics pipeline

Process uploaded files using a script or workflow manager: Pipelines and obtain 3 output files in a directory filtered_feature_bc_matrix/:

def run_scrna_analysis_pipeline():
    ln.setup.login("testuser2")
    transform = ln.Transform(name="Cell Ranger", version="7.2.0", type="pipeline")
    ln.track(transform)
    # access uploaded files as inputs for the pipeline
    input_artifacts = ln.Artifact.filter(key__startswith="fastq/perturbseq").all()
    input_paths = [artifact.stage() for artifact in input_artifacts]
    # register output files
    output_artifacts = ln.Artifact.from_dir(
        "./mydata/perturbseq/filtered_feature_bc_matrix/"
    )
    ln.save(output_artifacts)

    # Post-process these 3 files
    transform = ln.Transform(
        name="Postprocess Cell Ranger", version="2.0", type="pipeline"
    )
    ln.track(transform)
    input_artifacts = [f.stage() for f in output_artifacts]
    output_path = ln.dev.datasets.schmidt22_perturbseq(basedir=ln.settings.storage)
    output_file = ln.Artifact(output_path, description="perturbseq counts")
    output_file.save()


run_scrna_analysis_pipeline()

Show code cell output Hide code cell output

💡 saved: Transform(uid='6y3xwQE3Xuc1Ie', name='Cell Ranger', version='7.2.0', type='pipeline', updated_at=2023-12-22 11:25:32 UTC, created_by_id=1)

💡 saved: Run(uid='YTDcv1NqfdNM8RKHTRFx', run_at=2023-12-22 11:25:32 UTC, transform_id=4, created_by_id=1)

❗ this creates one artifact per file in the directory - you might simply call ln.Artifact(dir) to get one artifact for the entire directory

💡 saved: Transform(uid='SWv8IWrUJz7DRo', name='Postprocess Cell Ranger', version='2.0', type='pipeline', updated_at=2023-12-22 11:25:32 UTC, created_by_id=1)

💡 saved: Run(uid='SporxcN4QFPcjfDV1O1t', run_at=2023-12-22 11:25:32 UTC, transform_id=5, created_by_id=1)

Inspect data flow:

output_file = ln.Artifact.filter(description="perturbseq counts").one()
output_file.view_flow()

Integrate scRNA-seq & phenotypic data

Integrate data in a notebook:

def run_integrated_analysis_notebook():
    import scanpy as sc

    # create a new transform to mimic a new notebook (in reality you just run ln.track() in a notebook)
    transform = ln.Transform(
        name="Perform single cell analysis, integrate with CRISPRa screen",
        type="notebook",
    )
    ln.track(transform)

    # access the output files of bfx pipeline and previous analysis
    file_ps = ln.Artifact.filter(description__icontains="perturbseq").one()
    adata = file_ps.load()
    file_hits = ln.Artifact.filter(description="hits from schmidt22 crispra GWS").one()
    screen_hits = file_hits.load()

    # perform analysis and register output plot files
    sc.tl.score_genes(adata, adata.var_names.intersection(screen_hits.index).tolist())
    filesuffix = "_fig1_score-wgs-hits.png"
    sc.pl.umap(adata, color="score", show=False, save=filesuffix)
    filepath = f"figures/umap{filesuffix}"
    artifact = ln.Artifact(filepath, key=filepath)
    artifact.save()
    filesuffix = "fig2_score-wgs-hits-per-cluster.png"
    sc.pl.matrixplot(
        adata, groupby="cluster_name", var_names=["score"], show=False, save=filesuffix
    )
    filepath = f"figures/matrixplot_{filesuffix}"
    artifact = ln.Artifact(filepath, key=filepath)
    artifact.save()


run_integrated_analysis_notebook()

Show code cell output Hide code cell output

💡 saved: Transform(uid='kXYzVC39aPXspG', name='Perform single cell analysis, integrate with CRISPRa screen', type='notebook', updated_at=2023-12-22 11:25:34 UTC, created_by_id=1)

💡 saved: Run(uid='ZlcMFnt91akUDWjbonhZ', run_at=2023-12-22 11:25:34 UTC, transform_id=6, created_by_id=1)

WARNING: saving figure to file figures/umap_fig1_score-wgs-hits.png

WARNING: saving figure to file figures/matrixplot_fig2_score-wgs-hits-per-cluster.png

Review results

Let’s load one of the plots:

# track the current notebook as transform
ln.track()
artifact = ln.Artifact.filter(key__contains="figures/matrixplot").one()
artifact.stage()

Show code cell output Hide code cell output

💡 notebook imports: ipython==8.19.0 lamindb==0.64.2 scanpy==1.9.6

💡 saved: Transform(uid='1LCd8kco9lZUz8', name='Project flow', short_name='project-flow', version='0', type=notebook, updated_at=2023-12-22 11:25:36 UTC, created_by_id=1)

💡 saved: Run(uid='IFSUBi39tkQADcHFD6n9', run_at=2023-12-22 11:25:36 UTC, transform_id=7, created_by_id=1)

PosixUPath('/home/runner/work/lamin-usecases/lamin-usecases/docs/mydata/.lamindb/VEjlScRfzhrLHFy4ACWV.png')

display(Image(filename=artifact.path))

We see that the image artifact is tracked as an input of the current notebook. The input is highlighted, the notebook follows at the bottom:

artifact.view_flow()

Alternatively, we can also look at the sequence of transforms:

transform = ln.Transform.search("Bird's eye view", return_queryset=True).first()
transform.parents.df()

	uid	name	short_name	version	type	latest_report_id	source_code_id	reference	reference_type	initial_version_id	updated_at	created_by_id
id
4	6y3xwQE3Xuc1Ie	Cell Ranger	None	7.2.0	pipeline	None	None	None	None	None	2023-12-22 11:25:32.078563+00:00	1

transform.view_parents()

Understand runs

We tracked pipeline and notebook runs through run_context, which stores a Transform and a Run record as a global context.

Artifact objects are the inputs and outputs of runs.

What if I don’t want a global context?

Sometimes, we don’t want to create a global run context but manually pass a run when creating an artifact:

run = ln.Run(transform=transform)
ln.Artifact(filepath, run=run)

When does an artifact appear as a run input?

When accessing an artifact via stage(), load() or backed(), two things happen:

1. The current run gets added to artifact.input_of

2. The transform of that artifact gets added as a parent of the current transform

You can then switch off auto-tracking of run inputs if you set ln.settings.track_run_inputs = False: Can I disable tracking run inputs?

You can also track run inputs on a case by case basis via is_run_input=True, e.g., here:

artifact.load(is_run_input=True)

Query by provenance

We can query or search for the notebook that created the artifact:

transform = ln.Transform.search("GWS CRIPSRa analysis", return_queryset=True).first()

And then find all the artifacts created by that notebook:

ln.Artifact.filter(transform=transform).df()

	uid	storage_id	key	suffix	accessor	description	version	size	hash	hash_type	n_objects	n_observations	transform_id	run_id	initial_version_id	visibility	key_is_virtual	updated_at	created_by_id
id
2	UG0nHlbVNd8sHP36QOEV	1	None	.parquet	DataFrame	hits from schmidt22 crispra GWS	None	18368	q54ULUuKxw3LglyzMVYZ8Q	md5	None	None	2	2	None	1	True	2023-12-22 11:25:29.763920+00:00	1

Which transform ingested a given artifact?

artifact = ln.Artifact.filter().first()
artifact.transform

Transform(uid='6qT7fLNZYwe7sz', name='Upload GWS CRISPRa result', type='app', updated_at=2023-12-22 11:25:26 UTC, created_by_id=1)

And which user?

artifact.created_by

User(uid='DzTjkKse', handle='testuser1', name='Test User1', updated_at=2023-12-22 11:25:30 UTC)

Which transforms were created by a given user?

users = ln.User.lookup()

ln.Transform.filter(created_by=users.testuser2).df()

	uid	name	short_name	version	type	reference	reference_type	updated_at	latest_report_id	source_code_id	initial_version_id	created_by_id
id

Which notebooks were created by a given user?

ln.Transform.filter(created_by=users.testuser2, type="notebook").df()

	uid	name	short_name	version	type	reference	reference_type	updated_at	latest_report_id	source_code_id	initial_version_id	created_by_id
id

We can also view all recent additions to the entire database:

ln.view()

Show code cell output Hide code cell output

Artifact

	uid	storage_id	key	suffix	accessor	description	version	size	hash	hash_type	n_objects	n_observations	transform_id	run_id	initial_version_id	visibility	key_is_virtual	updated_at	created_by_id
id
10	VEjlScRfzhrLHFy4ACWV	1	figures/matrixplot_fig2_score-wgs-hits-per-clu...	.png	None	None	None	28814	ijpft7zAYShlKDXYYAD4hw	md5	None	None	6	6	None	1	True	2023-12-22 11:25:35.791756+00:00	1
9	TPV1vpugf6rLaLf9gmgM	1	figures/umap_fig1_score-wgs-hits.png	.png	None	None	None	118999	74WuaFnZeoMTvSpY--lbrA	md5	None	None	6	6	None	1	True	2023-12-22 11:25:35.602804+00:00	1
8	dGFslxyZwkO4Bw4fdjN6	1	schmidt22_perturbseq.h5ad	.h5ad	AnnData	perturbseq counts	None	20659936	la7EvqEUMDlug9-rpw-udA	md5	None	None	5	5	None	1	False	2023-12-22 11:25:34.092514+00:00	1
7	SaoL93JKysWYm535Tl7E	1	perturbseq/filtered_feature_bc_matrix/barcodes...	.tsv.gz	None	None	None	6	UM2cP709Ojx-_U5BgYDhBA	md5	None	None	4	4	None	1	False	2023-12-22 11:25:32.528192+00:00	1
6	LfA1pH0xPK8FIm1WnQFQ	1	perturbseq/filtered_feature_bc_matrix/matrix.m...	.mtx.gz	None	None	None	6	GApoWoSb4r99zQrQwnSy1w	md5	None	None	4	4	None	1	False	2023-12-22 11:25:32.527706+00:00	1
5	lrCAefdEUfloIzpuqMZI	1	perturbseq/filtered_feature_bc_matrix/features...	.tsv.gz	None	None	None	6	X0MOrXGhDmdZ3K_0q7EiQQ	md5	None	None	4	4	None	1	False	2023-12-22 11:25:32.527103+00:00	1
4	LIOgNeuKtXhaERCFpfBA	1	fastq/perturbseq_R2_001.fastq.gz	.fastq.gz	None	None	None	6	Ou1OEPt47npxcqaGh9Jjfg	md5	None	None	3	3	None	1	False	2023-12-22 11:25:31.231303+00:00	1

Run

	uid	transform_id	run_at	created_by_id	report_id	environment_id	is_consecutive	reference	reference_type
id
1	wN9HK7PuMGx1ADxQKeO7	1	2023-12-22 11:25:26.978332+00:00	1	None	None	None	None	None
2	3MAACmmTQ69HVjripVLE	2	2023-12-22 11:25:29.294492+00:00	1	None	None	None	None	None
3	lQnw1VQHzVmluFfZmiZT	3	2023-12-22 11:25:30.798974+00:00	1	None	None	None	None	None
4	YTDcv1NqfdNM8RKHTRFx	4	2023-12-22 11:25:32.081088+00:00	1	None	None	None	None	None
5	SporxcN4QFPcjfDV1O1t	5	2023-12-22 11:25:32.539041+00:00	1	None	None	None	None	None
6	ZlcMFnt91akUDWjbonhZ	6	2023-12-22 11:25:34.987748+00:00	1	None	None	None	None	None
7	IFSUBi39tkQADcHFD6n9	7	2023-12-22 11:25:36.172448+00:00	1	None	None	None	None	None

Storage

	uid	root	type	region	updated_at	created_by_id
id
1	3O6YyjbU	/home/runner/work/lamin-usecases/lamin-usecase...	local	None	2023-12-22 11:25:24.952301+00:00	1

Transform

	uid	name	short_name	version	type	latest_report_id	source_code_id	reference	reference_type	initial_version_id	updated_at	created_by_id
id
7	1LCd8kco9lZUz8	Project flow	project-flow	0	notebook	None	None	None	None	None	2023-12-22 11:25:36.169621+00:00	1
6	kXYzVC39aPXspG	Perform single cell analysis, integrate with C...	None	None	notebook	None	None	None	None	None	2023-12-22 11:25:34.982452+00:00	1
5	SWv8IWrUJz7DRo	Postprocess Cell Ranger	None	2.0	pipeline	None	None	None	None	None	2023-12-22 11:25:32.536609+00:00	1
4	6y3xwQE3Xuc1Ie	Cell Ranger	None	7.2.0	pipeline	None	None	None	None	None	2023-12-22 11:25:32.078563+00:00	1
3	p1xxYqcvrHW1xg	Chromium 10x upload	None	None	pipeline	None	None	None	None	None	2023-12-22 11:25:30.796252+00:00	1
2	8UtqUKuTeTw98Y	GWS CRIPSRa analysis	None	None	notebook	None	None	None	None	None	2023-12-22 11:25:29.290398+00:00	1
1	6qT7fLNZYwe7sz	Upload GWS CRISPRa result	None	None	app	None	None	None	None	None	2023-12-22 11:25:26.975236+00:00	1

User

	uid	handle	name	updated_at
id
2	bKeW4T6E	testuser2	Test User2	2023-12-22 11:25:32.070884+00:00
1	DzTjkKse	testuser1	Test User1	2023-12-22 11:25:30.788478+00:00

Show code cell content Hide code cell content

!lamin login testuser1
!lamin delete --force mydata
!rm -r ./mydata

✅ logged in with email testuser1@lamin.ai (uid: DzTjkKse)

💡 deleting instance testuser1/mydata
✅     deleted instance settings file: /home/runner/.lamin/instance--testuser1--mydata.env

✅     instance cache deleted
✅     deleted '.lndb' sqlite file
❗     consider manually deleting your stored data: /home/runner/work/lamin-usecases/lamin-usecases/docs/mydata