Table 1 Selected published articles of CAR T-cell therapy with single-cell sequencing technologies
From: Deciphering and advancing CAR T-cell therapy with single-cell sequencing technologies
Technology | Tumor type | Target antigen | Sample | Sample source | Research level | Cell number | Experimental design | Conclusion | Citation |
|---|---|---|---|---|---|---|---|---|---|
CAR T-cell product heterogeneity | |||||||||
scRNA-seq, bulk RNA-seq | B-ALL | CD19 | Activated and inactivated CAR T-cells | 5 healthy donors | Cell | 83,123 | Bulk RNA-seq and scRNA-seq were performed on antigen-specific stimulated and unstimulated CAR T-cells with non-costimulatory domains and different costimulatory domains (CD28,4-1BB). | Compared to CD28 CAR T-cells, 4-1BB CAR T-cells enriched in a central memory cell phenotype and fatty acid metabolism genes. And 4-1BB CAR T-cells also had increased expression of MHC II genes, ENPP2, and IL-21 axis genes, and decreased PD1. | [59] |
Bulk RNA-seq, CITE-seq, scATAC-Seq | B-ALL | CD19 | Post-infusion CAR T-cells | 71 patients | Human | 37,423 | RNA-seq was performed on a classified CAR T-cell subsets from 71 patients, followed by matching CITE-seq and scATAC-seq on T cells from 6 of these patients. | The TCF7 regulon was not only associated with the favorable naïve T-cell state, but maintained in effector T cells among patients with long-term CAR T-cell persistence; chronic IFN signaling regulated by IRF7 was associated with poor CAR T-cell persistence across T cell subsets. | [34] |
scRNA-seq, CITE-seq | ALL | CD19 | Activated and inactivated CAR T-cells | 1 healthy donor and 2 patients | Cell | 23,349 | scRNA-seq and CITE-seq were performed on CAR T-cells that were stimulated specifically or non-specifically with cells expressing CD19 or MSLN respectively. | Delineate the global cellular and molecular CAR T-cell landscape at baseline or in the activated state. Besides, healthy donor-derived CAR T-cells manifestated stronger functional activities correlated with the upregulation of MHC II genes than patient-derived CAR T-cells. | [61] |
scRNA-seq, ATAC-seq, flow cytometry | MM | BCMA | IPs | 13 healthy donors, public bulk RNA-seq and scRNA-seq datasets | Human | 43,981 | Detect the antitumor response of CARHigh and CARLow T-cells in vitro and in vivo. RNA-seq and ATAC-seq were performed on CARHigh and CARLow T-cells and scRNA-seq was performed on IPs from 3 independent donors. Analyze the correlation between CARHigh T-cell gene signature and clinical response through bulk RNA-seq and scRNA-seq datasets. | Reveal the different profiles between CARHigh and CARLow T-cells in phenotypic, functional, transcriptomic and epigenomic levels and provide mechanistic insights behind differential functionality of these cells. Particularly, CARHigh T-cells were related to tonic signaling and a cell exhausted phenotype as well as an increase in tumor cytotoxicity in vitro. Patients treated with CAR T-cell products enriched in CARHigh T cells showed a significantly worse clinical response. | [71] |
scRNA-seq, flow cytometry | ALL | CD19 | Activated CAR T-cells | 3 donors | Cell | 31,000 | Flow cytometry and scRNA-seq were performed on antigen-stimulated CAR T-cells that were not transduced and that were transduced by different lentiviral vectors (CD8-LV, VSV-LV). | VSV-LV CAR T-cells produced a more significant central memory phenotype, while CD8-LV CAR T-cells showed stronger cytotoxic activity. | [72] |
Antigen-specific stimulation of CAR T-cells | |||||||||
scRNA-seq, single-cell cytokine assay, single-cell cytotoxicity assay | BCL | CD19 | Activated and inactivated CAR T-cells | 3 healthy donors | Cell | 3,817 | scRNA-seq, single-cell cytokine assay, single-cell cytotoxicity assay were performed on antigen-specific stimulated or unstimulated CAR T-cells. | The activation states of CAR T-cells were highly mixed with TH1,TH2, Treg, and GM-CSF-expressing T cell responses in the same single cells and largely independent of differentiation status. | [75] |
Flow cytometry, scRNA-seq | MM | BCMA, TACI | Activated and inactivated CAR T-cells | 3 healthy donors | Cell | 53,191 | Combine flow cytometry and scRNA-seq to characterize three stages of the CAR T-cell production process, namely the starting leukapheresis sample, CAR-T-cell product, and the cellular product upon specific antigen stimulation. | CAR T-cell products from different donors showed a similar cellular composition, and only half of CAR-expressing cells displayed transcriptional changes upon CAR-specific antigen exposure. Particularly, a small proportion of antigen-responding CAR-expressing cells showed no transcriptional response to specific antigen stimulation and some stimulated CAR-expressing cells exhibited exhaustion features. | [74] |
Dynamic performance of CAR T-cells | |||||||||
scRNA-seq, scTCR-seq | B-ALL | CD19 | IPs, post-infusion CAR T-cells | 15 patients | Human | 184,791 | scRNA-seq and scTCR-seq were performed on IPs and PBMCs and BM-derived CAR T-cells at multiple time points after infusion (week 1–4/8, month 3/6). | TIGIT+CD27−CD62LLow was identified and validated in the pre-infusion product cell subsets, resulting in a highly efficient post-infusion CAR T-cell phenotype. | [76] |
TCR-seq, scRNA-seq | CLL, NHL | CD19 | IPs, post-infusion CAR T-cells | 10 patients | Human | 62,167 | TCR-seq was performed on CD8+ CAR T-cells before infusion and on day 7–14/26–30 after infusion, and scRNA-seq was performed on CD8+ CAR T- cells on day 7–14/26–30/83–112 after infusion . | Clonal diversity of CAR T-cells was highest in the IPs and declined following infusion. Clones expanding after infusion mainly originated from infused clusters with higher expression of cytotoxicity and proliferation genes. | [36] |
CyTOF, scRNA-seq, CITE-seq, scTCR-seq | CLL | CD19 | Post-infusion CAR T-cells | 2 patients | Human | 1,437 | CyTOF was performed on CAR T-cells from 2 patients followed for 9.3 and 7.2 years, followed by scTCR-seq, CITE- seq, and scRNA-seq for late-stage CD4+ CAR T-cells. | There were two different phases of anti-leukemia response: an initial stage dominated by CD8+ and γδ CAR T-cells, followed by a long-term remission stage characterized by Ki67hiCD4+ CAR T-cells that exhibited proliferative and cytolytic phenotypes. | [81] |
Cellular interactions with CAR T-cells | |||||||||
scRNA-seq, flow cytometry | PCL | BCMA | IPs, post-infusion CAR T-cells | 1 patient | Human | 55,488 | scRNA-seq was performed on CAR T -cells and endogenous T cells isolated from PBMC at three phases (day 0/8/15). | The mixed cell subsets mainly characterized by high metabolism of CD4+ T-cells in the initial stage would transition from highly amplified to cytotoxic CD8+ T-cells in the amplification stage. CD8+ memory-like cell states with high expression of RP genes were found in the cells on the final remission stage. | [80] |
Image analysis, flow cytometry, scRNA-seq | BCL | CD19 | BM cells | Mice | Mouse | NA | Flow cytometry and scRNA-seq were performed on BM cells from CAR T- cell-treated and untreated mice after 3 days of treatment. | CAR T-cells relied on cytokine-mediated crosstalk with the TME for optimal activity. IFN-γ produced by CAR T-cells enhanced endogenous T cells and sustained CAR T-cell cytotoxicity. | [85] |
scRNA-seq, flow cytometry | GBM | IL13Rα2 | CD45+ cells from the brains of untreated or CAR T-cell treated mice | Patients and mice | Human and Mouse | NA | Flow cytometry was performed on CAR T-cells after cell stimulation to measure IFN-γ production. scRNA-seq was performed on CD45+ cells from the brains of CAR T-cell treated or untreated mice. | IFN-γ production by CAR T-cells and IFN-γ responsiveness of host immune cells were critical for tumor immune landscape remodeling to promote a more activated and less suppressive tumor microenvironment. | [86] |
scRNA-seq, bulk RNA-seq, ATAC-seq | PAAD | MSLN | Activated CAR T-cells, post-infusion CAR T-cells | Healthy donors and patients | Cell | 16,000 | Bulk RNA-seq and ATAC-seq were performed on M5CAR T-cells after CAE and T cell exhaustion related gene enrichment and scRNA-seq was performed on day 0/20/28 with CAE driver dysfunction. | CAE drove CAR T-cell exhaustion and promoted CD8+ CAR T-cell to NK-like CAR T-cell transition. ID3 and SOX4 were upregulated during the process of exhaustion and knocking out ID3 and SOX4 in CAR T-cells slowed dysfunction and improved antitumor immunity. | [87] |
scATAC-seq, ChIP-seq, flow cytometry | B-ALL, Mm | CD19, BCMA | Activated CAR T-cells, post-infusion CAR T-cells | 2 patients | Human | 10,929 | scATAC-seq was performed on CAR T-cells cultured in vitro and CAR T-cells of 2 Mm patients who received BCMA CAR T-cell therapy at the peak and decline of amplification. | Panoramic chromatin accessibility after CAR T- cell differentiation and exhaustion was depicted; BATF and IRF4 were key regulators of CAR T- cell exhaustion and downregulation of BATF and IRF4 contributed to reducing CAR T-cell exhaustion and enhancing CAR T-cell therapeutic efficacy. | [47] |
Primary resistance | |||||||||
scRNA-seq | LBCL | CD19 | IPs | 24 patients | Human | 137,326 | scRNA-seq was performed on IPs from 24 LBCL patients treated with CAR T- cells. | Within the IPs of patients with PR/PD, exhausted CD8+ and CD4+ T cells were significantly enriched, while memory type CD8 T cells were significantly enriched within the IPs of patients who achieved CR. | [88] |
scRNA-seq, flow cytometry | NHL | CD19 | IPs, post-infusion CAR T-cells | 17 patients | Human | 94,000 | scRNA-seq and flow cytometry were performed on CAR T-cells from 17 NHL patients at different time points (IPs, day 14/30). | CD8+ CAR T-cells expressing exhausted marker TIGIT were associated with poor clinical response in NHL patients and targeted inhibition of TIGIT could improve the antitumor function of CAR T-cells. | [89] |
scRNA-seq, TCR-seq, flow cytometry | BCL | CD19 | Pre-infusion and post-infusion PBMCs, IPs | 32 patients | Human | 602,577 | scRNA-seq and scTCR-seq performed on 105 pre-treatment and post-treatment PBMC samples at different time points (day 30 before treatment, day 7 after treatment ) and IPs collected from 32 individuals with BCL treated with axi-cel or tisa-cel. | Expansion of proliferative memory-like CD8 clones was a hallmark of tisa-cel response, whereas axi-cel responders displayed more heterogeneous populations. Besides, the higher number of CAR Treg cells was associated with disease progression. | [70] |
scRNA-seq, genome-wide CRISPR/Cas9 knockout screening | B-ALL | CD19 | IP, post-infusion CAR T-cells | 2 patients | Human | NA | Genome-wide CRISPR/Cas9-based knockout screening was performed on the CD19+ human ALL cell line. scRNA-seq was performed on IPs from both responsive and unresponsive patients and on T cells at peak CAR T-cell amplification. | Death receptor signaling was identified as a key regulator of primary resistance to CAR T-cells in ALL. scRNA-seq confirmed that the CAR T-cells from patients with primary resistance expressed much higher levels of exhaustion markers. | [91] |
Relapse | |||||||||
CITE-seq, scRNA-seq, TCR-seq, flow cytometry | MM | BCMA | Pre-infusion and post-infusion BM cells | 23 patients | Human | 151,054 | CITE-seq, scRNA-seq, TCR-seq, and flow cytometry were performed on BM samples from patients with long and short PFS before infusion and at day 28 and month 3 after infusion. | Short PFS was associated with the lower diversity of pretherapy TCR repertoire, presence of hyperexpanded clones with exhaustion phenotype, and BAFF+PD-L1+ myeloid cells in the marrow. And long PFS was associated with an increased proportion of CLEC9A+ DCs, CD27+TCF1+ T cells with diverse T-cell receptors, and emergence of T cells expressing marrow-residence genes. | [96] |
scRNA-seq, scTCR-seq, cytokine multiplex profiling | MCL | CD19 | Pre-infusion and post-infusion MCL cells and non-tumor cells of TME | 20 patients | Human | 40,091 | scRNA-seq and scTCR-seq were performed on 39 samples collected over a long period of time from 15 CAR T-cell treated patients, and cytokine multiplex profiling was performed on 80 consecutive samples from 20 patients. | After relapse, CD4+ and CD8+ CTLs acquired expression of TIGIT and exhibited less cytotoxic. Besides, MCL tumor cells also increased TIGIT expression and then led to weaker antitumor immune surveillance. And elevated sIL2R in relapsed patients may contribute to therapeutic resistance by inhibiting T-cell expansion. | [98] |
Positive relapse | |||||||||
scRNA-seq, CITE-seq, flow cytometry, multiplexed secretomic assay | ALL | CD19 | Activated and inactivated CAR T-cells, pre-infusion CAR T-cells | 61 patients | Cell | 101,326 | scRNA-seq and CITE-seq were performed on (CAR-specific stimulation or TCR-mediated activation, not activated) IPs of 12 pediatric r/r ALL patients. Flow cytometry and multiplexed secretomic assay were performed on the pre-infusion CAR T-cells from other 49 patients. | The lack of TH2 function in CAR T-cell products was a novel mechanism for CD19-positive relapse and early memory-like T-cell subsets TSCM and TCM were significantly reduced in positive relapse patients. | [73] |
Negative relapse | |||||||||
scRNA-seq | B-ALL | CD19 | Leukemia cells | 1patient | Human | NA | scRNA-seq was performed on leukemic cells form BM before and after the patient received CD19 CAR-T cell therapy. | CD19-negative leukemic cells were present before CAR T-cell therapy and the relapse resulted from the selection of these rare pre-existing CD19-negative subclones. | [101] |
scRNA-seq | B-ALL | CD19 | CAR T-cells, leukemic cells | * | Cell | 1,039 | scRNA-seq was performed on the surviving leukemic cells cocultured with different methods (CAR T-cells and T cells) for 24 h. | Existing CD19low leukemic cells sustained decreased CD19 expression through transcriptional programs of physiologic B-cell activation and germinal center reaction, which helped achieve immune escape. | [102] |
Bulk RNA-seq, scRNA-seq, flow cytometry | B-ALL | CD19 | HSPCs | 2 second-trimester fetus | Human | 30,000 | Bulk RNA-seq, scRNA-seq and flow cytometry were performed on specific HSPC populations from FBM. CD34+CD19−CD22+ cells were detected in BM form B-ALL patients who relapsed or achieved CR. And use FISH and xenograft modeling to assess whether CD34+CD19−CD22+ cells initiate leukemogenesis. | CD22 preceded CD19 in normal B-cell development and CD34+CD19−CD22+progenitors underlie phenotypic escape after CD19-directed immunotherapies. | [100] |
CRS | |||||||||
scRNA-seq | B-ALL | CD44v6, CD19 | Post-infusion CD45+immune cells | 8 mice | Cell | 6,511 | scRNA-seq was performed on CD45+ cells at two periods (day 2/7 after fever) after CAR T-cell infusion. | Human circulating monocytes, rather than CAR T-cells, were primarily responsible for the systemic release of IL-6, which ultimately caused CRS. | [108] |
Single-cell cytokine profiling, flow cytometry | NHL | CD19 | IPs | 20 patients | Human | 500,000 | Single-cell multiplex cytokine profiling was performed on the IPs of 20 NHL patients. | Higher product PSI was associated with clinical response and severe CRS while higher numbers of IL-17A-producing polyfunctional CAR T (Th17)-cells was associated with severe ICANS. | [58] |
ICANS | |||||||||
scRNA-seq | NA | CD19 | Human brain, lung pericytes, PBMCs, mice brain cells | Mice and the BRAIN Initiative Cell Census Network | Cell | # | scRNA-seq was performed on brain cells of 4 healthy mice and verified the analysis results of human brain, peripheral blood and pulmonary parietal cells from a scRNA-seq database. | CD19 was expressed in human brain mural cells that are critical for BBB integrity and this cell population might contribute to the neurotoxicity of CD19-directed immunotherapy including CAR T-cells therapy. | [113] |
scRNA-seq | BCL | CD19 | IPs, post-infusion CAR T-cells | 72 patients | Human | 956,647 | After analysis of the HHV-6 public data sets of the two studies, scRNA-seq was performed on the IPs and CAR T-cells of 72 patients at different time periods after infusion as well as on PBMCs of 1 patient with HHV-6B overexpression on day 7/14/21 after treatment. | Reactivated HHV-6 carried by CAR T- cells may enter the CNS through OX40 receptors on BBB endothelial cells, resulting in the development of HHV-6 encephalitis, which has similar symptoms and requires differential diagnosis from ICANS. | [114] |
CyTOF, scRNA-seq, scTCR-seq, CITE-seq | LBCL | CD19 | Post-infusion CAR T-cells | 32 patients | Human | 6,316 | CyTOF was performed on CAR T-cells from 32 patients on day 7/21 after infusion, and CAR T-cells from 9 of them were analyzed by scRNA-seq, scTCR-seq, and CITE-seq. | CD4+Helios+ CAR T-cells on day 7 after infusion manifested hallmark features of Treg cells and were associated with progressive disease and less severe neurotoxicity. | [90] |
On-target, off-tumor effects | |||||||||
scRNA-seq, flow cytometry | NA | B-lineage- derived malignant cells, AML, and solid tumors related target antigens | Cells in normal tissues/organs | Healthy donors, public scRNA-seq datasets | Cell | # | Analyze 121 target antigen expression patterns of CAR T-cells in 18 tissues and organs derived from normal human samples, and then compare the expression levels of antigens in malignant cells and nonmalignant cells. | The expression patterns of 121 target antigens in normal tissues or organs were obtained at the single-cell level, which facilitated revealing the reason for on-target, off-tumor toxicity in special tissues/organs. | [116] |
scRNA-seq | # | * | Cells from the human cell landscape and the adult human cell atlas | 40 donors | Cell | 427,118 | Analyze the expression of 591 CAR targets in various cell types across different normal tissues from the public scRNA-seq databases. | A more stringent cutoff by defining a CAR target as a potentially risky gene had identified targets in the public databases that caused potential on-target, off-tumor toxicity. | [117] |
Engineered CAR T-cells | |||||||||
scRNA-seq, genome-wide CRISPR/Cas9 knockout screening | GBM | IL13Rα2 | Activated and inactivated engineered and control CAR T-cells | Healthy donors | Cell | 37,898 | Genome-wide CRISPR/Cas9-based knockout screening was performed on CAR T-cells to identify essential regulators of effector activity. scRNA-seq was performed on engineered (knockout of TLE4 or IKZF2) and control CAR T-cells with or without stimulation by tumor cells. | CRISPR screening identified targets including TLE4 and IKZF2, knockout of which resulted in the preservation or expansion of certain CAR T-cell subsets displaying transcriptional signatures of superior effector function and inhibited exhaustion responses. | [130] |
scRNA-seq, CRISPR/Cas9 genome editing system | NB | GD2 | Activated and inactivated engineered CAR T-cells | 2 healthy donors | Cell | 79,317 | scRNA-seq was performed on antigen-specific stimulated or unstimulated CAR T-cells. | The feasibility of preparing TRAC-targeted CAR T-cells by CRISPR/Cas9 technology and virus-free method was proved and TRAC-integrated CAR T-cells showed higher proportion of memory phenotype, less depletion phenotype and lower degree of differentiation. | [132] |
scRNA-seq, CRISPR/Cas9 genome editing system | B-NHL | CD19 | IPs, engineered CAR T-cells, PBMCs, | 3 patients | Cell | 63,789 | scRNA-seq was performed on CAR T-cells from 3 NHL patients before CAR T-cell infusion and on day 7/12/28/29 after infusion. | The feasibility of preparing PD1-targeted CAR T-cells by CRISPR/Cas9 technology and virus-free method was proved and non-viral, PD1-integrated CAR-T cells exhibited enhanced antitumor ability. | [131] |
scRNA-seq | MM, PAAD | MSLN, CD19 | Post-infusion tumors, engineered CAR T-cells, non-naive CD8+ T cells | Mice | Mouse | NA | scRNA-seq was performed on tumors on day 7 after treatment with CAR T-cells. | CAR T-cells delivered RN7SL1 via extracellular vesicles, which was selective to immune cells in TME and could directly elicit favorable changes in myeloid/DC subsets that helped activate endogenous CD8+ T cells. | [133] |
scRNA-seq | MM | TRP1 | TILs untreated or treated with CAR T-cells | Mice | Mouse | 9,767 | scRNA-seq was performed on TILs of MM mice untreated or treated with CAR T-cells. | Superkine IL-2 and IL-33 expressing CAR T-cells exhibited a potent, universal antitumor response and shifted the TME from immune suppressive to immune stimulatory in the absence of preconditioning. | [134] |
Combination therapy | |||||||||
scRNA-seq | LCA | ROR1 | IPs, pre-infusion and post-infusion lung tumors | Mice | Mouse | 16,672 | scRNA-seq was performed on IPs and lung tumors at three time points (untreated, 6 h after injection of Ox/Cy before CAR T-cell infusion and day 10 after CAR T-cell infusion). | The lymphodepletion regimen Ox/Cy activated lung tumor macrophages to produce multiple T-cell-recruiting chemokines that facilitated infiltration of CAR T-cells into tumors and remodel the immunosuppressive TME. | [120] |
scRNA-seq, flow cytometry | BRCA | Her2 | Post-infusion CD45+immune cells, CD4+CAR T-cells | Mice | Mouse | 128,000 | scRNA-seq was performed on CD45+ immune cells and CD4+ CAR T-cells isolated from TME on day 7/10 after Th/Tc17 CAR T -cell infusion with or without DMXAA. | DMXAA promoted CAR T-cell migration and persistence by generating a chemokine milieu that promoted CAR T-cell recruitment and modulating the immunosuppressive TME through alterations in the balance of immune-stimulatory and suppressive myeloid cells. | [123] |
CITE-seq, RNA-seq, flow cytometry, ATAC-seq, scRNA-seq | CC, OV | Lewis Y | Untreated and pretreated T cells | Mice | Mouse | NA | CITE-seq, RNA-seq and ATAC-seq were performed on T cells from vehicle- or CDK4/6i-treated CC tumors. scRNA-seq was performed on in vitro–activated T cells and in vivo-derived T cells respectively. Animal models were used to explore the efficacy of CDK4/6i combined with CAR T-cells. | CDK4/6i-pretreated T cells exhibited increased memory phenotype and immune persistence. And combination of CDK4/6i and CAR T-cells in ovarian cancer mice significantly improved the effectiveness and persistence of tumor control. | [126] |
Locoregional delivery of CAR T-cells | |||||||||
CyTOF | B-ALL, NHL, DLBCL | CD19 | IPs, PMBCs, BM, post-infusion CAR T-cells | 3 patients | Human | NA | CyTOF was performed to analyze the trafficking and functional proteins expression in CAR T-cells across patients’ tissues, including leukapheresis T cells, IPs, CAR T-cells in peripheral blood, BM, and CSF post infusion. | CAR T-cell product showed increased expression of trafficking and activation molecules, and patients’ CAR T cells from peripheral blood, BM and CSF showed spatiotemporal alteration in trafficking, activation, maturation, and exhaustion expression, with distinct signature in the CSF niche. | [140] |
scRNA-seq | CNSL | CD19 | Post-infusion CAR T-cells | Mice | Mouse | NA | scRNA-seq was performed on CAR T -cells isolated from mice BM with both CNS and systemic lymphoma 68 days after ICV or IV CAR T-cell infusion. | Compared with IV, exposure of CAR T-cells to CSF after ICV infusion led to a metabolic reprogramming that favored the formation of memory and exhibited enhanced antilymphoma activity. | [138] |
scRNA-seq | DIPG, DMG | GD2 | IPs, post-infusion CSF cells | 4 patients | Human | 65,598 | scRNA-seq was performed on IPs and CSF cells from patients after IV and ICV administration. | Transcriptomic analyses of IPs and CSF showed heterogeneity in response between participants and administration routes. Particularly, ICV administrations were associated with less immunosuppressive cell populations in CSF compared with IV infusions. | [141] |
CAR target selection | |||||||||
Machine learning, scRNA-seq, CITE-seq | # | # | Tumor, tumor-infiltrating normal and reference normal cells | 9 patients, public scRNA-seq datasets | Cell | # | scRNA-seq, random forest, convolutional neural networks were performed on identification of targets for logical switch-based CAR therapy and CITE-seq was performed on transcriptome-coupled epitope mapping. | The large-scale tumor-normal single-cell meta-atlas were leveraged to select gene pairs ( AND, OR and NOT switch targets) that contributed most to discrimination between individual malignant and normal cells. And the results were validated in ovarian cancer and colorectal cancer. | [154] |