AINovice2005/carbon-common-cohort-taxonomy

Carbon Common Cohort Taxonomy

A shared genomic cohort that provides the common reference slice for aligning and comparing enrichment datasets derived from the Carbon pretraining corpus.

Overview

The Carbon enrichment pipeline produces multiple datasets across CPU-based biological enrichment and GPU-based model enrichment. Although each dataset serves a different analytical purpose, meaningful comparisons require a consistent underlying population of genomic intervals.

carbon-common-cohort-taxonomy provides this shared reference layer.

The dataset identifies the genomic intervals included in the common cohort and associates them with their taxonomic context. It enables downstream workflows to align records across CPU-enriched sequences, model-derived likelihood statistics, and embedding datasets without comparing independently sampled populations.

The common cohort therefore acts as an analytical boundary between dataset generation and comparative analysis.

Dataset Purpose

The dataset supports the following workflows:

• Cross-dataset alignment: Connect genomic intervals across enrichment outputs using shared record-level identifiers.
• Taxonomy-aware analysis: Associate genomic intervals with their taxonomic classifications.
• CPU–GPU comparison: Compare biological sequence metrics with model-derived features on a common population.
• Integrated visualization: Support joint analysis of sequence characteristics, likelihood statistics, embeddings, and taxonomy.
• Downstream analytical workflows: Provide a reference table for taxonomic grouping, stratification, and comparative studies.

Schema

Feature	Description
record_id	Identifier linking the genomic interval to its source record.
start	Start coordinate of the genomic interval.
end	End coordinate of the genomic interval.
taxonomy	Taxonomic annotation associated with the genomic record.

Schema note: Confirm the exact data types and taxonomy representation against the published dataset schema before extending this table with additional fields.

Relationship to Other Datasets

The common cohort is intended to serve as a shared reference layer for the enrichment datasets.

Dataset	Role
carbon-cpu-enriched-sequences	Sequence-level biological and taxonomic enrichment.
carbon-likelihood-stats	Model-derived likelihood statistics from GPU enrichment.
carbon-embeddings	Model-derived vector representations for similarity analysis and retrieval.
carbon-common-cohort-taxonomy	Shared genomic cohort and taxonomic reference for cross-dataset alignment.

The common cohort does not replace the enrichment datasets. Instead, it provides the reference population needed to connect and compare their outputs.

Data Lineage

The dataset is part of the Carbon genomic data enrichment pipeline, which augments genomic records with sequence-level biological metrics and model-derived features.

The common cohort provides a shared analytical reference for connecting these enrichment layers and evaluating their behavior on a consistent population.

Related Datasets

• Carbon CPU Enriched Sequences
• Carbon Likelihood Statistics
• Carbon Embeddings