CITE-seq

The core innovation of CITE-seq lies in oligonucleotide-conjugated antibodies. Unlike fluorescent tags, these antibodies carry a unique DNA barcode. After binding to their specific cell-surface protein target, the barcode is read out by the sequencer alongside the cellular mRNA. This converts a protein signal into a quantitative, amplifiable DNA count, enabling highly multiplexed protein detection without spectral overlap issues.

CITE-seq uses standard single-cell encapsulation (e.g., droplet microfluidics) to isolate cells. Crucially, both polyadenylated mRNA and ADT-derived oligos are captured on the same barcoded bead via their shared poly-A tails. This ensures that every transcript and every detected epitope from a single cell share a common cellular barcode, creating a perfectly paired, multimodal dataset without complex registration algorithms.

Surface proteins are canonical markers for immune cell classification (e.g., CD4+ T cells, CD8+ T cells). CITE-seq leverages this by providing robust protein-level validation of cell identities that may be ambiguous from transcriptome data alone. This is critical for resolving closely related populations where marker gene mRNA expression is low or poorly correlated with protein abundance, such as distinguishing naive from memory lymphocyte subsets.

Downstream analysis tools like Seurat v3+ and TotalVI implement weighted nearest neighbor (WNN) algorithms. These methods learn a shared latent space that integrates both RNA and ADT modalities, weighting each modality based on its information content per cell. The result is a unified clustering and visualization (e.g., UMAP) that reflects a more complete biological definition of cell state than either modality alone.

A fundamental challenge in genomics is the imperfect correlation between mRNA and protein levels. CITE-seq directly quantifies this relationship in thousands of single cells. By measuring both analytes simultaneously, researchers can identify genes where post-transcriptional regulation decouples RNA expression from surface protein abundance, providing deep insights into translational control and protein trafficking dynamics.

CITE-seq is often combined with cell hashing, where sample-specific oligo-tagged antibodies against ubiquitous surface markers (like CD45) label cells from different donors. All samples are then pooled into a single run. This eliminates batch effects, reduces reagent costs, and enables computational demultiplexing to assign each cell to its original sample based on its hashtag oligo count, while simultaneously capturing the full CITE-seq protein panel.

The oligonucleotide-conjugated antibodies that make CITE-seq possible. Each ADT carries a unique barcode sequence that is amplified and sequenced alongside the cellular transcriptome. The resulting ADT counts serve as a quantitative proxy for surface protein abundance. Key considerations include:

• Titration: Optimizing antibody concentration to minimize background binding.
• Barcode design: Ensuring orthogonality to prevent cross-hybridization.
• Isotype controls: Essential for distinguishing specific signal from non-specific Fc receptor binding.

The computational process of fusing RNA and ADT data into a unified latent space. Methods like Weighted Nearest Neighbor (WNN) analysis in Seurat v4 learn cell-specific modality weights, giving more influence to the data type with less noise for each cell. This allows for the discovery of cellular states that are not clearly separated by either transcriptome or proteome alone, such as subtle activation states in T-cell populations.

The commercial implementation of CITE-seq reagents by BioLegend. These antibodies are available in A, B, and C oligo formats, each compatible with different single-cell platforms (10x Genomics, etc.). The TotalSeq™-A format is the most common, featuring a poly-A tail that allows capture via the standard 3' single-cell gene expression workflow, making it a direct add-on to existing scRNA-seq protocols.

A computational strategy for mapping CITE-seq data onto scRNA-seq-only datasets. Since most public atlases lack protein measurements, bridge integration uses the shared transcriptome component to transfer protein-level annotations. This effectively imputes surface protein abundance in datasets where ADTs were never measured, enabling the retrospective enrichment of legacy data with proteomic resolution.

While both use oligonucleotide-conjugated antibodies, they serve distinct purposes:

• Cell Hashing: Targets ubiquitously expressed surface markers (e.g., CD45, MHC-I) to encode sample identity for multiplexing and doublet detection.
• CITE-seq: Targets lineage-specific markers (e.g., CD4, CD19) to measure immunophenotype. These techniques are often combined in a single experiment, using hashtag oligos (HTOs) for sample demultiplexing and ADTs for protein quantification.

A critical preprocessing step distinct from RNA normalization. The centered log-ratio (CLR) transformation is the standard method, treating each cell as a composition and dividing each ADT count by the geometric mean of all ADTs in that cell. This accounts for variation in cell size and antibody staining efficiency. Alternative methods like DSB (Denoised and Scaled by Background) explicitly model ambient and non-specific binding using empty droplets and isotype controls.