Catalog columns#

Every column of the catalog schema. "Facet" columns are categorical filters (Catalog.query(<column>=...), values from sddb-cli facets <column>); "range" columns are numeric and filter with __gte / __lte. source is not stored in the catalog file: the client derives it on load.

Facet values#

A string facet value is first looked up as given in that column of the whole catalog. If it is not there, case, spaces, hyphens and underscores are ignored. A technology value may also be an assay label, so technology="10x Xenium" selects Xenium, and technology="PhenoCycler-Fusion" selects both PhenoCyclerFusion1 and PhenoCyclerFusion2. A value that still does not resolve raises ValueError with the closest known value ("did you mean 'Xenium'?"). A known value that matches nothing after the other filters gives an empty cohort.
hne_image and if_image take True or False. A null value means unknown and matches neither.
source is the lowercased study_id prefix before ":". In the 2026-10-28 catalog (6087 datasets) the values are geo (4113), hest (1070), 10x (175), abc (115), bia (4), metaspace (3), bruker (3), doi (2) and figshare (2). 600 datasets have no study_id, so their source is null.
query(redistributable=True) keeps datasets whose license is known and allows redistribution (license_redistributable is true and license_unknown is false). redistributable=False keeps the rest, including every dataset with an unknown license, where false means the terms are not known. Every dataset in the catalog is served, whatever its license.

Column	Dtype	Facet	Range
`uid`	`string`
`canonical_key`	`string`
`study_id`	`string`	yes
`collection_name`	`string`
`issue_url`	`string`
`organism`	`string`	yes
`organism_id`	`string`
`tissue`	`string`	yes
`tissue_id`	`string`
`disease`	`string`	yes
`disease_id`	`string`
`development_stage`	`string`	yes
`technology`	`string`	yes
`assay`	`string`	yes
`technology_version`	`string`
`panel_size`	`Int64`		yes
`validation_status`	`string`	yes
`tier`	`string`	yes
`license`	`string`	yes
`license_url`	`string`
`license_noncommercial`	`boolean`	yes
`license_redistributable`	`boolean`
`license_unknown`	`boolean`
`non_integer_counts`	`boolean`
`n_obs`	`Int64`		yes
`n_features`	`Int64`		yes
`total_counts`	`Float64`		yes
`zarr_url`	`string`
`card_url`	`string`
`vitessce_url`	`string`
`size_bytes`	`Int64`		yes
`container_tag`	`string`
`sddb_version`	`string`
`published_at`	`string`
`slug`	`string`
`title`	`string`
`pathological`	`boolean`
`product`	`string`
`modality`	`string`
`modalities`	`list[string]`	yes
`biomaterial_type`	`string`
`chemistry_version`	`string`
`staining_method`	`list[string]`
`preservation_method`	`string`	yes
`hne_image`	`boolean`	yes
`if_image`	`boolean`	yes
`ftu_annotation`	`boolean`
`mpp`	`Float64`		yes
`image_width`	`Int64`		yes
`image_height`	`Int64`		yes
`magnification`	`string`	yes
`sdata_repr`	`string`
`n_transcripts`	`Int64`		yes
`capture_area`	`string`
`publication_date`	`string`
`doi`	`string`	yes
`pmid`	`string`
`publication_year`	`string`
`publication_verify_status`	`string`
`sample_id`	`string`
`donor_id`	`string`
`database_version`	`string`
`sddb_id`	`string`
`disease_details`	`string`
`default_table`	`string`
`dataset_url`	`string`
`thumbnail_url`	`string`
`collections`	`list[string]`	yes
`sample_title`	`string`
`timepoint`	`string`
`timepoint_kind`	`string`
`timepoint_days`	`Float64`		yes
`treatment`	`string`
`genotype`	`string`
`condition`	`string`
`obs_unit`	`string`
`segmentation_source`	`string`
`segmentation_method`	`string`
`same_section_group`	`string`
`same_section_relation`	`string`
`same_section_release`	`string`
`same_section_panel_of`	`list[string]`
`source`	`string`	yes

Quality columns#

Column	Description
`validation_status`	Result of schema validation, `pass` or `fail`. In the current catalog every dataset passes validation, so this filter changes nothing yet. Filter on `tier` (`silver`, `bronze`) for quality.
`tier`	Curation tier: `gold`, `silver` or `bronze`, from most to least curated. Some datasets have no tier yet.
`non_integer_counts`	True when a count table is stored as deposited with fractional values (mostly x.5 or x.25). The values aren't rounded, and the depositor doesn't state the cause.

Study and condition columns#

All optional: older catalogs without them still load, and they are null when the source has no such metadata.

Dataset columns#

Column	Description
`study_id`	Identifier of the study the dataset belongs to.
`sample_title`	Free-text title of the individual sample.
`timepoint`	Sampling timepoint as written in the source (e.g. `day 7`).
`timepoint_kind`	Kind of timepoint (e.g. days post injury, age, treatment duration).
`timepoint_days`	Timepoint normalised to days (numeric; supports range filters).
`treatment`	Treatment or perturbation applied to the sample.
`genotype`	Genotype of the sample.
`condition`	Condition label for the sample (e.g. control vs. disease).

Collection columns (`collections.parquet`)#

Column	Description
`overall_design`	Study-level design summary text.
`pubmed_id`	PubMed ID of the associated paper.
`source_url`	URL of the study's source record.
`parent`	Slug of the parent collection, if any.
`source`	Origin of the collection (e.g. GEO, HEST).
`year`	Publication year (integer).
`contrast_axes`	JSON object `{axis: {value: count}}` of the condition axes that vary across members.
`assays`	Distinct assays across member datasets.
`modalities`	Distinct modalities across member datasets.
`organisms`	Distinct organisms across member datasets.
`tissues`	Distinct tissues across member datasets.
`diseases`	Distinct diseases across member datasets.
`timepoint_kinds`	Distinct timepoint kinds across member datasets.
`condition_values`	Distinct condition values across member datasets.