Ab-Seq/CITE-Seq is a powerful single-cell multi-omics analysis tool that simultaneously reveals in-depth information on both proteins and mRNA.

High-throughput sequencing, capable of detecting a vast number of RNA targets, has greatly enhanced our understanding of complex biological systems. However, to further investigate gene regulation and single-cell heterogeneity, one can choose to simultaneously acquire information on both RNA and protein expression, enabling multi-dimensional in-depth research.
Principle of Cell Surface Protein Detection:

Ab-Seq/CITE-Seq utilizes oligonucleotide-conjugated antibodies (Ab-oligos) to detect protein expression from high-throughput sequencing data. Each antibody clone in Ab-Seq/CITE-Seq is conjugated to a specific oligonucleotide containing an antibody-specific barcode (ABC). Using Ab-Seq/CITE-Seq in combination with a single-cell analysis system allows for the simultaneous detection of mRNA and protein expression in a sample.

Ab-Seq/CITE-Seq antibodies are highly specific. High specificity and low background are critical metrics for evaluating antibodies. Ab-Seq/CITE-Seq products are titrated during development to determine the optimal antibody concentration and usage standards, ensuring superior antibody specificity (recognition of the corresponding antigen) and low background (non-specific binding to non-target cells).

Ab-Seq/CITE-Seq enables the simultaneous detection of protein and RNA expression at the single-cell level, enhancing cellular phenotyping resolution, while also validating and supplementing RNA data, providing a more comprehensive reflection of cell state. Single-cell multi-omics goes beyond traditional bulk analysis by simultaneously resolving multiple omics layers within specific cell populations, significantly improving the efficiency and depth of single-cell research.

Sample Types:
·Tissue, Blood, Cultured cell lines, Prepared single-cell suspensions
·Note: If the client provides tissue samples and lacks the capability to perform tissue dissociation to obtain single-cell suspensions, Novel Bio will endeavor to provide technical and experimental assistance. However, due to the specificity of different sample types, we cannot guarantee the applicability of our methods to all tissue types.

Quality Requirements:
1.Cell viability >70%
2.Concentration: 500-2,000 cells/μl
3.Volume not less than 200μl
4.Cell culture media or buffers must not contain Ca2+ and Mg2+ ions
5.Cell diameter less than 40μm

A strong quantitative correlation exists between AbSeq and flow cytometry, particularly evident for well-resolved markers.

Even low-abundance cell populations, such as CD1c+ conventional dendritic cells (cDCs) and cross-presenting CD141+ cDCs, can be clearly identified by their surface protein phenotypes.

Flow Cytometry Analysis of Cell Surface Markers

Cluster of Differentiation (CD) molecules are cell surface markers used for the identification and characterization of leukocytes. The CD nomenclature was developed and is maintained by the Human Leukocyte Differentiation Antigens (HLDA) workshops, established in 1982. The CD nomenclature is now widely used for humans, mice, and other species. The figure below displays key surface markers for different cell types in humans and mice.

BD AbSeq Oligonucleotide-Conjugated Monoclonal Antibody Library:

Novel Bio, in collaboration with Becton Dickinson (BD), has launched a BD AbSeq oligonucleotide-conjugated monoclonal antibody library containing over 200 antibodies. This library covers multiple fields including tumor immunology, autoimmune diseases, and infectious diseases, providing important research tools for disease diagnosis and treatment. These antibodies are categorized into cell typing, signal transduction, cell proliferation & differentiation, disease diagnosis & monitoring, among others, with each category offering more comprehensive biological data for research in related fields.

1.Experimental Groups:
·Healthy bone marrow samples: 6 cases (3 young + 3 elderly)
·Acute Myeloid Leukemia (AML) patients: 15 cases (6 APL + 9 with NPM1 mutation)

2.Capture Platform:
·BD Rhapsody + BD FACSAria II/Fusion + BD FACSFortessa/LSRII

3.Primary Technical Methods:
AbSeq, Flow Cytometry, qPCR validation, etc.

Single-cell genomics technologies have revolutionized our understanding of complex cellular systems. However, high costs and the lack of purification strategies for newly discovered cell types have hindered their functional characterization and large-scale analysis. This study constructed high-content single-cell proteo-genomic reference maps of human blood and bone marrow. By quantitatively correlating the expression of up to 197 surface markers, it revealed the identity features and biological processes of all major hematopoietic cell types in healthy aging and leukemia patients. These reference maps enable the automated design of cost-effective, high-throughput cytometry panels whose performance surpasses existing top methods, accurately reflecting the complex topology of cellular systems and enabling the purification of precisely defined cell states. Through the systematic integration of cytometry and proteo-genomic data, we were able to measure the functional capacity of precisely localized cell states at the single-cell level. This study not only provides a convenient resource for the cytometry field but also opens a new chapter in the data-driven era.

1.Single-Cell Proteo-Genomic Reference Map of Hematologic Malignancies

Fig. 1

To construct a single-cell transcriptome-surface protein reference map of human bone marrow, the authors performed Abseq experiments: labeling iliac bone marrow mononuclear cells with 97–197 oligo-tagged antibodies, followed by targeted 462-mRNA panel detection or whole transcriptome single-cell RNA sequencing (scRNA-seq) on the BD Rhapsody platform (Fig. 1a). This panel systematically covers various differentiation stages of hematopoietic stem and progenitor cells (HSPCs), cell identity genes, and state markers, and was validated by whole transcriptome data to ensure no cell populations were missed. The authors then applied the 97-surface marker panel to bone marrow samples from 3 young healthy, 3 elderly healthy, and 3 newly diagnosed AML patients (Fig. 1a). Healthy samples were enriched for CD34⁺ cells to detail HSC differentiation, while AML samples were enriched for CD3⁺ cells where necessary to ensure T cell coverage. To validate the reliability of the high-dimensional antibody multiplexing system, the research team designed three control experiments: paired experiments with antibody presence/absence, paired experiments with fresh/cryopreserved-thawed samples, and sequencing depth evaluation experiments, all confirming the robustness of the experimental workflow.

Ultimately, integrating RNA and surface protein information from 70,017 high-quality BM cells identified 45 cell types, covering the entire CD34⁺ HSC differentiation spectrum, CD3⁺/CD56⁺ T-NK populations, CD33⁺ myeloid subsets, and various differentiation states of CD10⁺/CD19⁺/CD38^hi B cells (Fig. 1b, c), and included rare populations such as cytotoxic CD4⁺ T cells and MSCs. Notably, bone marrow cell states were highly consistent between young and elderly donors, while AML patients exhibited heterogeneity (Fig. 1b). Importantly, the combination of RNA and surface protein information provided higher resolution and revealed cell types not easily identifiable from a single data layer.

Fig. 1

In addition to the primary reference dataset, the authors also generated "query" single-cell proteo-genomic datasets, visualized within the context of the primary reference. These datasets specifically included: 1. Analysis of bone marrow (BM) and paired peripheral blood (PB) samples from healthy individuals using a 197-antibody panel to investigate the expression of more surface markers within the constructed reference system; 2. For healthy bone marrow, a strategy combining a 97-antibody panel with whole transcriptome mapping was used to query the expression status of any gene within the range defined by the reference system; 3. Analysis of CD34⁺CD38⁻ bone marrow cells using a 97-antibody panel to enhance the resolution of immature HSPCs; 4. Inclusion of relevant sample data from 12 AML patients.
To make their resource more accessible, the authors developed the Abseq application, a web-based tool that enables visualization of gene and surface marker expression, differential expression testing, and data-driven gating schemes for all datasets in the manuscript. A demo video is available in the supplementary information. The Abseq application is available at: https://abseqapp.shiny.embl.de/

Fig. 2

Although surface markers are widely used to identify cell types, stages, and biological processes, their individual importance remains unclear. To address this, the authors precisely assigned each cell to a cell type and quantified its differentiation stage, stemness score, cytotoxicity score, cell cycle stage, and technical covariates. An unsupervised factor model was further used to introduce covariates for unknown biological processes, with non-technical covariates not influenced by marker expression levels. For each surface marker, the authors calculated the proportion of its expression variance explained by the aforementioned processes (Fig. 2a), thereby identifying markers indicative of key biological properties such as cell type identity, differentiation stage, stemness, cytotoxicity, and cell cycle.

Fig. 3

To validate newly identified markers from the analysis, the authors focused on surface molecules that specifically indicate HSPC differentiation stages. Pseudotime analysis within CD34⁺ HSPCs screened for markers dynamically changing along erythroid, megakaryocyte, monocyte, cDC, or B cell trajectories (the monocyte trajectory also included neutrophil progenitors; mature neutrophils were missing due to density gradient centrifugation; pDC and Eo/Ba trajectories were not analyzed due to low cell numbers) (Fig. 2d). Pseudotime quantified the dynamics of classical marker expression (e.g., pan-differentiation CD38, early B cell CD10, cDC-monocyte CD11c) and revealed new markers like CD326, CD11a, and Tim3 that precisely delineate lineage commitment stages (Fig. 2d; Fig. 3). Validation via FACS-indexed scRNA-seq and single-cell culture confirmed: CD326 specifically predicts erythroid commitment (Fig. 3c–g), Tim3 and CD11a are pan-myeloid commitment markers (Fig. 3c,h-o); CD98 was confirmed as a new pan-differentiation marker for HSPCs (Fig. 2d). Further analysis of the entire B cell differentiation pathway identified specific surface markers for commitment, maturation, isotype switching, and terminal plasma cell generation.

The authors' model provides a comprehensive quantitative understanding of the relationships between cell type identity, differentiation stage, biological processes, and the expression of individual surface markers. A comprehensive overview of surface markers associated with these processes is provided in the supplementary information.

Fig. 4

To investigate the impact of healthy aging on surface protein expression, the authors compared Abseq data from young and elderly healthy individuals. They found that surface molecule expression was highly similar across all bone marrow cell populations in both groups (Fig. 4a,b), suggesting the overall pattern remains stable and tightly regulated during aging. Although cell type frequencies changed only slightly, cytotoxic CD8⁺ T cells accumulated significantly. Furthermore, the surface expression of immunoregulatory molecules like FAS (CD95), CD155, and CD275 changed with age, particularly a decrease in CD27 expression on naive CD8⁺ and CD4⁺ T cell subsets (Fig. 4b,c), indicating selective alterations in the surface presentation of immunoregulatory molecules during healthy aging.

Fig. 4

To decipher the remodeling of surface markers in Acute Myeloid Leukemia (AML), the authors collected single-cell proteo-genomic profiles from 15 patients (6 with t(15;17) acute promyelocytic leukemia, and 9 with NPM1-mutated AML with normal karyotype, 4 of which harbored FLT3-ITD). While the bone marrow (BM) topology remained consistent with healthy references, even three treatment-naïve AML samples already exhibited patient-specific alterations and significant heterogeneity (Fig. 1b). Projecting the leukemic cells onto the healthy reference framework enabled precise localization of their differentiation blockade stages (Fig. 4d). Unsupervised clustering based on differentiation stage abundance identified three AML subtypes: monocytic-type (enriched for blasts with a classical monocytic phenotype), APL (blocked at the promyelocyte stage), and immature-type (enriched for HSC/MPP/ELP-like blasts) (Fig. 4e,f). Differential expression analysis revealed that surface markers distinguishing AML subtypes (e.g., CD133, CD14, CD11b) also marked corresponding healthy stages (Fig. 4g), and uncovered differential expression of potential therapeutic targets PD-L1 (CD274) and CTLA4 (CD152) (Fig. 4h). Comparing AML cells to their healthy counterparts at the same differentiation stage identified leukemia-specific highly expressed markers, including known leukemia stem cell markers such as CD25, Tim3, CD123, and CD45RA (Fig. 4i), which showed significant inter-patient heterogeneity. This "healthy reference projection + cell state differential testing" workflow holds promise as a standard paradigm for scRNA-seq analysis in hematologic malignancies. The computational code is accessible online at https://git.embl.de/triana/nrn

Fig. 5

To overcome limitations of traditional flow cytometry gating, such as trial-and-error experience and insufficient purity of rare subsets, the authors used a machine learning framework for data-driven gating design on the full dataset: machine gating for all cell populations significantly improved purity compared to classical literature-based schemes (Fig. 5a), with the Hypergate algorithm achieving higher recall while maintaining high purity (Fig. 5a). The authors then validated this approach by focusing on two rare and poorly characterized bone marrow populations:

·Cytotoxic CD4⁺ T cells – The Hypergate algorithm suggested a CD4⁺CD28⁻ immunophenotype, with significantly lower expression of CD7, CD25, CD127, and CD197 compared to other CD4⁺ T cell subsets (Fig. 5b-e); this gating was validated by flow cytometry in bone marrow from healthy individuals and various hematologic patients (Fig. 5d), and qPCR after FACS sorting confirmed their cytotoxic gene expression (Fig. 5f).

Fig. 5

·Mesenchymal Stem Cells (MSCs) are a rare, heterogeneous population in BM. While expanded MSCs have been extensively phenotyped, the phenotype of primary human MSCs remains incompletely defined, partly due to their very low frequency. In the authors' dataset, they captured a small number of heterogeneous MSCs, with one subpopulation (MSC-1) showing high expression of the key cytokine CXCL12 (Fig. 5g). The Hypergate algorithm identified CD13 expression combined with CD11a absence as the most effective way to isolate CXCL12-expressing MSCs (Fig. 5h). Flow cytometric analysis of CD13+CD11a- MSCs validated the immunophenotype suggested by the Abseq data and confirmed known and new MSC surface markers discovered by their method (Fig. 5i,j). Furthermore, FACS-based isolation of CD13+CD11a- cells followed by transcriptome analysis revealed strong enrichment for CXCL12 and other key MSC signature genes (Fig. 5k).

These results demonstrate that the authors' method can derive gating strategies directly from data and map surface marker expression for poorly characterized populations. Combined with the single-cell proteo-genomic reference map, the Abseq application allows users to define new data-driven sorting schemes for any population of interest.

Fig. 6

To obtain a flow cytometry strategy most closely reflecting transcriptomic states of HSPCs, the authors trained a decision tree on the Abseq data of CD34⁺ cells from the 'Young1' sample, constructing a novel gating strategy requiring only 12 surface markers to define 14 molecularly precise cell states (Fig. 6a–c). This data-driven strategy significantly outperformed existing expert schemes in identifying lineage-committed progenitors, producing cell populations with higher transcriptional homogeneity (Fig. 6d,e), and yielded functional output comparable to 'consensus gating' but with richer information content. The authors then implemented this 12-marker panel on a classical flow cytometry platform, combined with Smart-seq2 index scRNA-seq for validation: the new gating effectively separated molecularly defined cell states (Fig. 6f,g), and its resolution power in both training and validation data was significantly superior to traditional expert schemes (Fig. 6h). In summary, the results demonstrate that high-content single-cell proteo-genomic maps can be used to derive data-defined cytometry panels that describe the molecular states of complex biological systems with high precision. Furthermore, the authors' gating strategy allows for the faithful identification and prospective isolation of transcriptomically defined progenitor states within the human hematopoietic hierarchy using cost-effective flow cytometry.

Fig. 7

To embed discretely gated FACS data into the continuous differentiation space, the authors developed the open-source algorithm NRN (git.embl.de/triana/nrn). This algorithm unifies flow cytometry and Abseq scales through rank normalization and then projects individual cells onto the proteo-genomic reference UMAP using k-NN. Validation using the 12-marker and an 11-marker panel on FACS-indexed Smart-seq2 data showed that molecularly defined cell types were accurately mapped (Fig. 7b). Projection accuracy was comparable to full transcriptome integration, and the expression dynamics of key lineage genes along pseudotime highly correlated with their location (Fig. 7c), demonstrating that NRN can resolve the continuum of hematopoietic differentiation using flow cytometry data.

Fig. 7

To address the limitation of single-cell genomics in understanding functional differentiation potential, the authors utilized the gating panel defined in Fig. 6, recorded surface markers in single-cell culture assays, and mapped the resulting flow cytometry data back to the Abseq reference map using the NRN algorithm. The results showed: cells with the highest proliferative and multi-lineage potential localized to the phenotypic HSC/MPP compartment; HSPCs distributed along the transcriptomic trajectory sequentially increased the proportion of corresponding progeny (Fig. 7d); functionally unipotent progenitors appeared both on their respective trajectories and within the HSC/MPP compartment (Fig. 7d,g), while oligopotent cells were enriched in the HSC/MPP compartment and showed fate combinations (e.g., erythro-megakaryocytic-eosinophilic/basophilic and lympho-neutro-monocytic-dendritic) significantly higher than random (Fig. 7e,f,g), suggesting stochastic or hierarchical regulation not captured by the transcriptome alone. In summary, hematopoietic lineage commitment primarily proceeds continuously along transcriptomically predicted paths, with early divergence into erythroid and lympho-myeloid branches; the authors' data resource combined with the NRN algorithm allows accurate integration of flow cytometry and single-cell genomic data, enabling the depiction of continuous differentiation using flow cytometry and the mapping of functional data into genomic space.

1.Constructed the first high-resolution single-cell proteo-genomic reference map, integrating 97–197 surface markers with transcriptome data, systematically covering all major hematopoietic cell types and differentiation stages in healthy and leukemic bone marrow and blood, identifying 45 cell types and states.

2.Achieved quantitative association of surface markers with cell identity/biological processes, defining the functional weight of each surface marker in cell type identification, differentiation stage, stemness, cytotoxicity, or cell cycle, and identifying novel stage-specific markers (e.g., CD326, CD11a, Tim3, CD98).

3.Developed data-driven flow cytometry design tools based on machine learning algorithms (e.g., Hypergate algorithm, decision trees) to automatically generate sorting schemes with high purity and recall, significantly outperforming traditional manual gating strategies, especially for rare cell populations (e.g., cytotoxic CD4⁺ T cells, MSCs).

4.Redefined HSPC classification and function, proposing a new HSPC classification scheme that more accurately maps transcriptomically defined differentiation trajectories, validated by functional experiments for its predictive power of differentiation potential, addressing issues of insufficient purity and high functional heterogeneity in classical gating strategies.

5.Established a framework for precise classification and targeted therapy of hematologic malignancies by projecting AML samples onto the healthy reference map, precisely locating the differentiation block stage, and identifying leukemia stem cell-specific markers (e.g., CD25, Tim3, CD123, CD45RA), providing a molecular basis for personalized immunotherapy.

Link to Original Article:https://doi.org/10.1038/s41590-021-01059-0