CITE-seq

The core innovation of CITE-seq is the use of antibodies labeled with DNA barcodes instead of fluorophores. These antibody-derived tags (ADTs) bind to specific cell-surface epitopes. The barcodes are then captured during the standard single-cell RNA-seq library preparation, allowing protein abundance to be read out as countable sequencing reads alongside the transcriptome.

CITE-seq provides true single-cell, single-experiment multi-omics. Key advantages include:

• No signal overlap: Unlike flow cytometry, DNA barcodes avoid spectral overlap issues, enabling highly multiplexed panels (200+ markers).
• Direct correlation: Protein and mRNA levels for the same gene can be directly compared within the same cell, revealing post-transcriptional regulation.
• Unified workflow: Both modalities are captured in a single droplet-based or microwell-based run, minimizing technical batch effects.

Surface proteins are often more stable and lineage-specific than mRNA transcripts. CITE-seq leverages this to resolve cellular identities that are difficult to separate by transcriptomics alone. For example, CD4+ and CD8+ T cells can be definitively partitioned using ADT signals, even when their transcriptomes are highly similar. This immunophenotypic ground truth is critical for annotating novel clusters in complex tissues like tumors.

BioLegend's TotalSeq™ reagents are the commercial implementation of CITE-seq technology. They offer pre-optimized panels for human and mouse immunology, including:

• Universal Panels: Broad panels covering major immune lineages.
• Custom Conjugations: User-defined antibodies can be conjugated to unique barcodes.
• Feature Barcoding: The same chemistry is used for cell hashing and sample multiplexing, reducing costs and doublet rates.

By measuring both mRNA and surface protein, CITE-seq uniquely quantifies post-transcriptional gene regulation at scale. Discordance between transcript and protein abundance—such as high mRNA but low surface protein—can indicate active translational repression or protein degradation. This provides a functional readout of cellular state that is invisible to pure RNA-seq methods.

A supervised machine learning approach that projects cell-type annotations from a well-characterized reference atlas onto a new query dataset. In CITE-seq workflows, protein expression data provides orthogonal validation for label transfer accuracy. Algorithms like Seurat's FindTransferAnchors identify mutual nearest neighbors in a shared correlation space, enabling robust annotation of query cells even when transcriptomic signatures are ambiguous.

Single-cell Variational Inference is a deep generative model based on a variational autoencoder that learns a probabilistic latent representation of gene expression. When extended to multimodal data, models like totalVI jointly model RNA and protein counts from CITE-seq experiments, accounting for batch effects and zero-inflation in both modalities. This provides a principled statistical framework for normalization and imputation.

The computational alignment of multiple single-cell datasets into a shared latent space. For CITE-seq, integration must handle modality imbalance—thousands of RNA features versus dozens of protein features. Methods like Harmony and scANVI correct for technical variation while preserving biological signals. Proper integration enables meta-analyses across studies, technologies, and donors without losing the protein-level resolution that distinguishes cell states.