skills.rest
epiblastai
epiblastaiOFFICIAL
Explore all skills by this author →
GitHub

multimodal-alignment

Reconcile inconsistent cell barcodes across multimodal LanceDB observation tables.

7|Updated Feb 21, 2026
Search Tags:#single-cell#lancedb#data curation#feature spaces#multimodal alignment#barcode reconciliation#cell matching
One-click install
npx skills add https://github.com/epiblastai/homeobox --skill multimodal-alignment
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: multimodal-alignment
Source: https://github.com/epiblastai/homeobox/tree/main/packages/polycomb/skills/multimodal-alignment
Command: npx skills add https://github.com/epiblastai/homeobox --skill multimodal-alignment

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires lancedb, polycomb, and includes scripts (resource) components.

What problem does it solve?

This skill resolves inconsistent cell barcode formats across multimodal feature spaces so the same physical cells can be joined reliably across gene expression, protein, chromatin, and other assay tables.

Core Features & Use Cases

  • Barcode Reconciliation: Tests exact matching, prefix and suffix stripping, lane cleanup, and reverse-complement normalization.
  • Overlap Optimization: Selects the normalization that maximizes the minimum cross-modality barcode overlap and reports unmatched cells.
  • Audited Curation: Adds and populates a canonical multimodal_barcode column through tracked transactions, with dry-run support and restricted write columns.
  • Use Case: Align CITE-seq gene-expression and antibody-derived-tag tables whose barcodes use different lane prefixes or well suffixes before downstream table joining and multimodal analysis.

Quick Start

Run the barcode reconciliation workflow on a staged dataset LanceDB using the CellIndex observation class, review the overlap statistics in dry-run mode, and apply the audited changes when the alignment is satisfactory.

Frequently Asked Questions about multimodal-alignment

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reconcile inconsistent cell barcodes across multimodal single-cell datasets?▼

Cell barcode reconciliation tests exact matching, prefix and suffix stripping, lane cleanup, and reverse-complement normalization to select the format that maximizes cross-modality barcode overlap.

What is the best way to join CITE-seq gene expression and antibody tables with different barcode formats?▼

Aligning CITE-seq tables uses barcode normalization and audited curation to add a canonical multimodal_barcode column, ensuring reliable joins across gene expression and antibody-derived-tag feature spaces.

Does multimodal barcode alignment support dry-run validation before applying changes?▼

Barcode alignment supports dry-run validation, allowing you to review overlap statistics and unmatched cells before applying tracked transaction updates to the LanceDB observation tables.

Can I use LanceDB and Polycomb APIs for single-cell data curation and barcode matching?▼

LanceDB and Polycomb are required dependencies to evaluate barcode normalizations, apply audited column updates, and manage canonical join keys across staged multimodal observation tables.

When do I need canonical barcode normalization for Multiome or NEAT-seq observation tables?▼

Canonical barcode normalization is needed when physical cells across Multiome, NEAT-seq, or CITE-seq assays use inconsistent formats, preventing reliable joins across multiple feature-space tables.

What are the limitations of matching cell barcodes across different feature spaces?▼

Barcode matching evaluates normalization strategies to maximize minimum cross-modality overlap but will report unmatched cells, indicating physical cells that cannot be reconciled across the feature spaces.

Related Skills

Explore More Data & Analytics Skills→
epiblastaiepiblastai
7

finalize-tables

Finalizes harmonized Lance tables into linked, schema-conformant biomedical collections.

Official
Advanced
AcosmiAcosmi
6

image

Generate, edit, analyze, and convert images via configured vision models.

Community
Advanced
epiblastaiepiblastai
7

prepare-package-for-resolution

Stages raw biomedical tables into Lance for schema resolution.

Official
Advanced
djankiesdjankies

migrating-v3-to-v4

Migrate Zod v3 codebases to v4 with automated scripts and compatibility checks.

Community
Advanced