Ancillary Tables

These are the underlying raw data tables snapshotted from the sequencing and alignment pipeline. Each table is stored as a directory of Parquet files.

dim_samples

Sample metadata for each physical sample in the experiment. Combines sample run execution data with plan-level identifiers and antigen information. Grain: one row per sample_id.

Column	Type	Description
`sample_id`	`string`	Physical sample identifier.
`sample_plan_id`	`string`	Planned sample part number from the run plan.
`candidate_library_id`	`string`	Identifier of the candidate library consumed by this sample.
`candidate_library_plan_id`	`string`	Part number of the candidate library from the run plan.
`sino_catalog_id`	`string`	Sino catalog identifier for the antigen used in the binding assay.
`antigen_gene_name`	`string`	Gene name of the antigen.
`antigen_sequence`	`string`	Amino acid sequence of the antigen.
`antigen_concentration_nM`	`float`	Antigen concentration in nanomolar.
`translation_condition`	`string`	Translation condition used for the sample (e.g., reducing, disulfide).
`selection`	`boolean`	Whether the sample underwent binding selection. `false` indicates an expression-only sample.

Key relationships:

sample_id joins to aligned_reads.sample_id, labeling_candidates_data.sample_id.
Binding samples (selection = true) correspond to binding_umi_counts.binding_sample_id.
Expression samples (selection = false) correspond to expression_umi_counts.expression_sample_id.

labeling_candidates_data

Candidate designs enriched with sample-run metadata, antigen information, and selection status. This table connects customer-provided candidate names and sequences with the experimental conditions under which they were tested. Grain: one row per (sample_id, candidate_id, sino_catalog_id, antigen_concentration_nM).

Column	Type	Description
`sample_id`	`string`	The sample identifier from the run.
`candidate_name`	`string`	Customer-provided name for the candidate.
`candidate_sequence`	`string`	The full candidate sequence.
`sino_catalog_id`	`string`	Sino catalog identifier for the antigen.
`antigen_gene_name`	`string`	Gene name of the antigen.
`antigen_sequence`	`string`	Amino acid sequence of the antigen.
`antigen_concentration_nM`	`float`	The antigen concentration in nanomolar.
`selection`	`boolean`	Whether this sample underwent binding selection.
`candidate_id`	`string`	Blake-3 content digest of the reference fragment sequence.

Key relationships:

sample_id joins to dim_samples.sample_id.
candidate_id is the shared candidate key across all tables.
candidate_name is the customer-provided name that maps to the internal candidate_id.

aligned_reads

Full-length aligned reads with non-null UMI, filtered to the latest sequencing task attempt per sample and candidate. Joins alignment metrics with read sequences. Grain: one row per (read_id, sample_id).

Column	Type	Description
`read_id`	`string`	UUID of the read coming off the sequencer.
`sample_id`	`string`	The sample identifier.
`candidate_id`	`string`	Blake-3 hash of the reference DNA sequence that this read aligned to.
`variant_id`	`string`	Blake-3 digest of the reconstructed reference sequence after resolving non-UMI ambiguous N bases. Equals `candidate_id` when no ambiguity.
`umi`	`string`	Unique Molecular Identifier extracted from the read.
`candidate_sequence`	`string`	The nucleotide sequence of the read.

Key relationships:

sample_id joins to dim_samples.sample_id.
candidate_id is the shared candidate key across all tables.

binding_umi_counts

Per-UMI read counts for binding samples (samples where selection = true). Each row represents a unique combination of binding sample, candidate, variant, and UMI. Grain: one row per (binding_sample_id, candidate_id, variant_id, umi).

Column	Type	Description
`binding_sample_id`	`string`	The binding sample identifier.
`candidate_id`	`string`	Blake-3 content digest of the candidate reference sequence.
`variant_id`	`string`	Blake-3 digest of the reconstructed reference sequence after resolving non-UMI ambiguous N bases. Equals `candidate_id` when no ambiguity.
`umi`	`string`	Unique Molecular Identifier extracted from the read.
`read_count`	`integer`	Number of reads observed for this sample, candidate, and UMI.

Key relationships:

binding_sample_id corresponds to dim_samples.sample_id where selection = true.
candidate_id is the shared candidate key across all tables.

expression_umi_counts

Per-UMI read counts for expression samples (samples where selection = false). Structure mirrors binding_umi_counts. Grain: one row per (expression_sample_id, candidate_id, variant_id, umi).

Column	Type	Description
`expression_sample_id`	`string`	The expression sample identifier.
`candidate_id`	`string`	Blake-3 content digest of the candidate reference sequence.
`variant_id`	`string`	Blake-3 digest of the reconstructed reference sequence after resolving non-UMI ambiguous N bases. Equals `candidate_id` when no ambiguity.
`umi`	`string`	Unique Molecular Identifier extracted from the read.
`read_count`	`integer`	Number of reads observed for this sample, candidate, and UMI.

Key relationships:

expression_sample_id corresponds to dim_samples.sample_id where selection = false.
candidate_id is the shared candidate key across all tables.

Dataset Guide

Resources

dim_samples

labeling_candidates_data

aligned_reads

binding_umi_counts

expression_umi_counts

​dim_samples

​labeling_candidates_data

​aligned_reads

​binding_umi_counts

​expression_umi_counts

dim_samples

labeling_candidates_data

aligned_reads

binding_umi_counts

expression_umi_counts