Recent advancements in molecular pathology research have identified recurrent genetic alterations associated with lymphoma. Frequent mutations such as MYC gene rearrangements, BCL2 translocations, inactivating mutations in TP53, and truncating mutations in the PEST domain of NOTCH1 have been increasingly recognized. These mutations play a key role in the occurrence and development of lymphoma. MYC gene rearrangement disrupts the metabolic homeostasis of oxidative phosphorylation, leading to uncontrolled tumor proliferation. Such cases are often accompanied by genomic instability and reduced chemosensitivity (6). FL is characterized by an inherited alteration of IGH-BCL2 translocation, which puts the BCL2 gene under the control of immunoglobulin enhancers, resulting in overexpression. BCL2-positive patients typically respond poorly to chemotherapy and have shorter survival (7). Mutations or inactivation of the TP53 gene are relatively common in relapsed and refractory lymphoma. Such gene abnormalities are closely associated with increased tumor aggressiveness, treatment tolerance, and a high risk of recurrence, which seriously affect the prognosis of patients (8). NOTCH1 is a driver of MCL (9). Target gene networks such as HES1/HEY1 regulate cell proliferation and differentiation, angiogenesis, and drug resistance development, becoming key drivers of aggressive disease progression.

Genomic structural abnormalities can contribute to the malignant transformation of lymphoma. A prominent example is chromosomal translocation, with the IGH-BCL2 translocation being the most frequently observed in follicular lymphoma. This translocation leads to BCL2 gene overexpression under the control of the IGH enhancer, therefore inhibiting apoptosis and facilitating lymphoma initiation and progression. Copy number variations (CNVs) are also quite common. For instance, the deletion of the cyclin-dependent kinase inhibitor 2A (CDKN2A) gene is an independent factor associated with poor progression-free survival (PFS) and overall survival (OS) (10). Such patients often have a poor prognosis and reduced responsiveness to chemotherapy (11). Deletion of the p16INK4a protein encoded by CDKN2A leads to an uncontrolled cell cycle, which promotes tumor cell proliferation (12). These structural variants not only reveal the core mechanism of lymphoma progression but also provide molecular loopholes for targeted therapy.

In cancer treatment, targeted molecular therapies are used not only as first-line treatments but also often as second-line regimens. Their mechanism of action involves the selective targeting of intrinsic molecular susceptibilities within malignant cells. These therapies include small-molecule inhibitors and protein degradation technologies. Common small-molecule inhibitors include BTK inhibitors, BCL2 inhibitors, and EZH2 inhibitors. The main action of BTK inhibitors has been systematically described by the teams of Broccoli A and Tian G (40, 41). Ibrutinib is an oral Bruton’s tyrosine kinase (BTK) inhibitor. It irreversibly inhibits the B-cell receptor (BCR) signaling pathway by covalently binding to the cysteine (Cys) 481 site, marking a milestone in treating B-cell tumors. First-generation drugs, such as Ibrutinib, have been approved by the European Medicines Agency (EMA) to treat relapsed or refractory mantle cell lymphoma (R/R MCL). Clinical evidence indicates that the timing of therapeutic intervention exerts a substantial influence on patient prognosis (42, 43). The SYMPATICO study showed that the combination of Venetoclax can extend the median PFS of relapsed MCL patients to 31.9 months (44). Patients treated with ibrutinib at the first recurrence had a 2.5-fold improvement in median PFS compared with late-line use (25.4 vs 10.3 months) (45). To address the therapeutic limitations of drug resistance, Chen Q et al. conducted a study demonstrating that the CDK1 inhibitor RO-3306 increases sensitivity to BTKi in DLBCL. These findings propose a novel therapeutic strategy for overcoming resistance mediated by the C481S mutation (46). For first-generation drugs like Ibrutinib, clinical response declines due to resistance mutations such as C481S. Second-generation inhibitors Acalabrutinib and Zanubrutinib address this by optimizing kinase selectivity to improve specificity and reduce off-target effects. Zanubrutinib improves oral bioavailability by 3.9% compared to Ibrutinib, showing higher ORR and longer PFS in relapsed MCL (47, 48). Chihara et al. reported that the combination of Ibrutinib with PD-1 inhibitors yielded an ORR of 77.8% in patients with primary or secondary central nervous system (CNS) lymphoma. However, treatment was associated with grade 3 to 4 hematologic toxicities (49). In the future, comprehensive molecular profiling of CNS lymphoma is needed to refine diagnostic criteria and optimize therapeutic strategies. Furthermore, dynamic monitoring of ctDNA holds the potential for guiding timely interventions against resistant subclones, thus enabling more precise therapeutic targeting. Venetoclax is a highly selective BCL-2 protein antagonist. It works by competitively binding to the BH3 domain of BCL-2. This relieves BCL-2’s inhibition of pro-apoptotic proteins like BAX/BAK. It also restarts the tumor cell apoptosis program (50). BCL-2 overexpression often coexists with EZH2 gain-of-function mutations. Targeted combination with Tazemetostat (an EZH2 inhibitor) can synergistically induce dual regulation of the epigenetic-apoptotic axis (51, 52).

Coughlin CA et al. revealed that BCL10 mutations in DLBCL contribute to treatment resistance via two distinct mechanisms. First, they activate the NF-κB signaling pathway and establish a positive feedback loop involving IL-6 and STAT1/2, which increases the expression of anti-apoptotic BCL-2 family proteins such as BCL-xL and BFL1, thus conferring resistance to Venetoclax monotherapy. Simultaneously, these mutations alter the BCR signaling network, suppressing the inhibitory effects of BTK inhibitors on downstream kinases (53). To address this resistance mechanism, the combination of non-covalent BTK inhibitors Pirtobrutinib and Venetoclax reveals significant synergistic effects. It blocks NF-κB/AP-1 transcriptional activation mediated by BCL10 mutations while suppressing the expression of anti-apoptotic proteins BCL-2 and MCL-1. This significantly inhibits the growth of BCL10-mutated lymphoma cells, showing potent antitumor activity and providing a new strategy to overcome BCL10 mutation-related drug resistance (53).

The EZH2 inhibitor Tazemetostat targets the catalytic subunit of the histone methyltransferase PRC2 complex. It competitively inhibits its binding to the cofactor S-adenosylmethionine (SAM), thus blocking H3K27me3-mediated epigenetic silencing (54). In FL and DLBCL, EZH2 gain-of-function mutations (i.e., Y641F) drive silencing of tumor suppressor genes like CDKN2A by improving H3K27me3 catalytic efficiency. These mutations are significantly associated with tumor progression and poor prognosis (55). Izutsu K et al. used Tazemetostat as monotherapy for EZH2-mutated FL, yielding 70.6% ORR (56). This demonstrated good long-term efficacy and safety, supporting its use as a third-line treatment option. The ongoing SYMPHONY-1 Phase III trial (NCT04224493) for relapsed/refractory FL evaluates its synergistic effects when combined with lenalidomide-rituximab. Importantly, the EZH2 resistance mutation profile reveals dynamic evolutionary characteristics. A recent study identified that the W113C mutation within the SET domain increases the binding affinity of SAM, thus attenuating the inhibitory potency of Tazemetostat by altering the positioning and structural stability of residue R685 (55). This resistance mechanism shows molecular heterogeneity compared to the classic Y641F mutation. This finding highlights the need for systematic screening of non-hotspot EZH2 mutations in clinical practice and drives the development of new-generation PRC2 inhibitors. Currently, MAK683, CPI-1205, SHR2554, PF-06821497, and other agents are in Phase 2 or 3 antitumor trials. Efficacy validation of Tazemetostat combination regimens in DLBCL (NCT05618366 and NCT05604417) and analysis of resistance mechanisms will provide critical evidence-based support for targeted therapy.

Proteolysis-Targeting Chimera (PROTAC) technology constructs bifunctional molecules composed of a target protein-ligand and an E3 ligase ligand. These molecules form a ternary complex of target protein-PROTAC-E3 ligase, triggering ubiquitination of the target protein and proteasome-mediated specific degradation (57). This technology has demonstrated significant advantages in targeting oncoproteins (BCL6 and MYC) refractory to traditional therapies (58). Malarvannan M et al. comprehensively evaluated the design protocols and optimization strategies employed in developing PROTACs (59). As a representative example, the BCL6-specific degrader DZ-837 was rationally designed based on the N-phenyl-4-pyrimidineamine scaffold. It can effectively reduce BCL6 protein levels in DLBCL cells. It can also be used in combination with Ibrutinib to overcome resistance mutations. PROTAC molecules directed against MYC have shown dual therapeutic efficacy in mouse models: prolonging survival and inhibiting tumor growth (60). Compared with gene editing technologies like CRISPR-Cas9, the PROTAC system has unique, reversible regulatory properties. Its mechanism of action does not rely on permanent genomic modifications. It provides more clinical application potential in terms of operational cycle, cost control, and off-target risk (61). Fu et al. further expanded this platform and successfully constructed an EZH2-targeted molecular library. Among these agents, the bifunctional degrader ZJ-20 induces the degradation of the EZH2 enhancer, thus facilitating the disassembly of the entire PRC2. It shows excellent antiproliferative activity and favorable pharmacokinetic properties (62). Studies have developed new PROTAC molecules based on the reversible non-covalent BTK inhibitor ARQ531. These molecules can effectively circumvent resistance mechanisms mediated by C481S mutations. Their efficacy is significantly higher than traditional covalent BTK-PROTACs (63). Particularly, the innovative STAT3D PROTAC series molecules conjugate a STAT3-specific decoy with thalidomide. This not only efficiently reduces STAT3 protein levels but also precisely inhibits the expression of key oncogenes regulated by STAT3, such as BCL2L1, CCND2, and MYC. As a result, it suppresses lymphoma cell proliferation while inducing cell apoptosis (64). This targeted degradation strategy showed minimal off-target activity against other members of the STAT family, such as STAT1 and STAT5, thus contributing to improved therapeutic specificity and reinforcing the direct anti-tumor efficacy in B-cell lymphoma. Together, these developments have validated the important application value of PROTAC technology in cancer therapy (64).

Recently, CD20-directed therapy has been the core strategy in managing B-cell lymphoma. The following section provides a systematic overview of conventional therapeutic approaches targeting CD20, CD30, and CD52 antigens in treating lymphoma ( Table 1 ).

Advancements in multi-targeted therapeutic strategies are driving significant innovation in the clinical management of lymphoma. In this context, Esquinas et al. engineered novel chimeric CD79b-targeted CAR-T cells. They found that these CAR-T cells can induce CD79b and CD19 co-downregulation in NHL cells. Among them, CARLY3 has become an advantageous candidate for NHL treatment. It has excellent targeting specificity, durable tumor clearance ability, and minimal antigen loss rate (71).

Traditional single-target CAR-T has an antigen escape defect. Dual-target CAR-T (CD19/CD22) addresses this by significantly reducing recurrence risk through a synergistic targeting mechanism. It shows improved efficacy in relapsed/refractory B-cell lymphoma (72). CD20×CD3 bispecific antibodies are novel T cell-engaging antibodies (such as glofitamab and epcoritamab). They are approved by the FDA/EMA for treating DLBCL and FL. However, approximately half of the patients exhibit primary resistance (73, 74). A retrospective analysis by Kyvsgaard ER et al. identified NOTCH1 mutation-induced loss of CD20 antigen expression as a critical mechanism contributing to therapeutic resistance, highlighting the role of genetic alterations in modulating treatment efficacy (74). The novel bispecific antibody TG-1801 targets both the CD47-SIRPα axis and CD19. This improves the ADCC/ADCP effects of anti-CD20 antibodies. It also achieves multi-dimensional antitumor activity by regulating the GPR183-PI3Kδ pathway (75). Kolbe C et al. proposed a combinatorial therapeutic strategy involving anti-CD20 bispecific antibodies with CD39/CD73 inhibition. This regimen improves therapeutic efficacy by counteracting adenosine-mediated immunosuppression, restoring immune effector function within the TME. It increases tumor killing rates by 2.1-3.5 fold, CD8⁺ T cell expansion by 3.8 fold, and IFN-γ secretion by 4.2 fold. Patients with high CD39-expressing effector memory T cells (TEM) show a 67% higher response rate. This suggests that CD39 may serve as a predictive biomarker for efficacy (76). Resistance to BTK inhibitors (BTKi) and CAR-T is closely related to Myc pathway activation, with HSP90AB1 and CDK9 as key regulatory nodes (77). Jiang V et al. reported that the CDK9 inhibitor enitociclib effectively inhibits protein synthesis and induces apoptosis, successfully overcoming resistance to sequential therapies in a MCL model (78). A combined inhibition approach targeting both HSP90 and CDK9 offers a promising strategy for addressing therapeutic resistance. Moreover, CD19/CD70 dual-target CAR-T therapy achieved complete remission in patients with relapsed/refractory primary central nervous system DLBCL, sustaining 17 months of disease-free survival without neurotoxicity, highlighting the potential of multi-targeted immunotherapies (79). These advances, comprising targeted therapies, immune modulation, and resistance reversal, have significantly enhanced the precision treatment landscape for lymphoma and laid a solid foundation for future clinical translation.

Clinical trials for novel monoclonal antibodies, bispecific antibodies, and CAR-T therapies are currently being conducted globally. The following table ( Table 2 ) summarizes some representative ongoing trials (March 2025).

Recent breakthroughs in universal CAR-T (UCAR-T) technology drive cell therapy toward an “off-the-shelf” model. This therapeutic strategy utilizes gene editing to ablate endogenous TCR and MHC molecules in T cells, generating universal cellular products compatible with multiple patients. Preliminary clinical trials have demonstrated both the efficacy and safety of this approach in treating lymphoma. Based on this foundation, research groups have engineered a modular universal chimeric antigen receptor T-cell (MU-CAR-T) platform incorporating an Sd/Gv covalent linkage system. This system enables the modular assembly of diverse single-chain variable fragments, facilitating flexible and precise multi-antigen targeting. MU-CAR-T cells have shown potent cytotoxic activity against HIV-infected and T-cell lymphoma cells, while also suppressing tumor progression in vivo through enhanced immune infiltration and cytokine release (80). This modular framework streamlines manufacturing processes and quality control protocols and establishes a comprehensive therapeutic paradigm with broad applicability to oncological and viral pathologies.

To optimize immunocompatibility, Zhu S and his team used lentiviral transduction technology. They developed CD38-targeted UCAR-T cells overexpressing LLT1. The introduction of LLT1 promotes CAR-T cell proliferation and antitumor activity. It also effectively defends against rejection by allogeneic NK and T cells. This improved UCAR-T demonstrates higher survival rates and tumor clearance capacity. It reduces inflammatory responses, providing a key solution to the immune rejection problem in universal therapies (81). Deep mechanistic studies on host immune rejection have led to Dasatinib-resistant UCAR-T (KM UCAR-T) development. This involves introducing the T316I mutation in the LCK gene and knocking out the TRAC and B2M genes. When combined with a Dasatinib administration strategy, this approach has led to the development of a novel cell therapy with both anti-rejection and anti-tumor properties (82). In vitro studies have demonstrated that KM UCAR-T cells retain their activation and functional capacity in the presence of Dasatinib. Furthermore, in xenograft mouse models, these cells show effective tumor clearance and resistance to host immune responses (82). The integration of multiple strategies (multiplex antigen recognition, immune modulation, and optimization of manufacturing processes) has significantly expanded the clinical potential of UCAR-T therapies.

Immune checkpoint therapy has introduced a transformative approach to cancer treatment by modulating T cell activation pathways and reshaping the anti-tumor immune response. Beyond the established targets of CTLA-4 and PD-1/PD-L1, identifying emerging immune checkpoints such as LAG-3, TIM-3, and TIGIT offers promising avenues to overcome current therapeutic limitations and improve clinical efficacy (83). LAG-3 acts as a key co-inhibitory receptor that inhibits T-cell function by binding to MHC-II molecules (84, 85). LAG-3 and PD-1 work synergistically to drive the exhaustion of CD8⁺ T cells, suppressing the IFN-γ-dependent anti-tumor immunity. Blocking both pathways simultaneously can promote IFN-γ release. It reduces T-cell exhaustion and enhances tumor clearance (86). Relatlimab is the first FDA-approved LAG-3 inhibitor. It is combined with a PD-1 antibody to improve efficacy (87). In R/R HL patients who failed PD-1 treatment, the combination of Relatlimab and Pembrolizumab achieved an ORR of 31%, and the PFS at 12 months reached 39%. Dual blockade of the LAG-3/PD-1 pathway can reverse the exhaustion of CD8⁺ T cells, promote the release of IFN-γ, and enhance the tumor clearance ability (88).

TIM-3 acts as a transmembrane immunomodulatory molecule. It drives T cell function exhaustion by binding ligands such as Galectin-9. Its inhibitor, sabatolimab, has shown significant potential in the treatment of myeloid tumors (89–91). The recently developed small-molecule compound ML-T7 effectively disrupts PtdSer/CEACAM1 signaling, thus enhancing the functional activity of CAR-T cells. Moreover, by targeting the FG-CC0 cleft of TIM-3, ML-T7 contributes to remodeling the immune microenvironment. Its therapeutic efficacy as a monotherapy is comparable to that of TIM-3-targeting antibodies, highlighting its potential as a promising immunomodulatory agent. The combination of PD-1 inhibitors can significantly improve the tumor suppression rate. It has the advantage of oral administration (92). It has been further found that the TIM-3/Galectin-9 signaling axis forms a negative feedback loop. This occurs through IFN-γ-mediated upregulation of Galectin-9. This finding provides theoretical support for developing combination immunotherapy (93).

TIGIT, a co-inhibitory receptor that interacts with CD155 and CD112, plays a critical role in immune evasion by suppressing the activity of NK cells and T lymphocytes. Its underlying mechanisms and therapeutic potential in lymphoma have increasingly attracted research interest, positioning it as an emerging focus in tumor immunology (94, 95).In FL, TIGIT is significantly overexpressed in tumor-infiltrating T cell subsets. These include Treg, follicular helper T cells, and exhausted T cells. It is closely related to a poor survival prognosis. Importantly, anti-PD-1 therapy can specifically regulate TIGIT⁺ depleted T cell subsets. This suggests the potential value of a dual-target blockade strategy (96). The mechanism of recurrence after CAR-T therapy in mantle cell lymphoma showed that TIGIT expression was significantly upregulated on depleted T cells and cytotoxic T cells (CTLs). Tumor cells use monocyte-expressed CD155/PVR to promote immunosuppressive signaling by upregulating TIGIT expression. This interaction is associated with the accumulation of MDSCs, collectively contributing to an immunosuppressive TME. These findings underscore the pivotal role of TIGIT in mediating resistance to CAR-T therapy (97). In DLBCL, the co-expression of TIGIT and PD-1 in tumor-infiltrating T cells (TILs) is particularly prominent. These TILs not only have close contact with malignant B cells but also reveal functional defects in cytokine secretion. Combined blockade of TIGIT and PD-1 can completely clear A20 lymphoma and significantly prolong survival in most mice. This highlights the clinical translation prospect of synergistic immune checkpoint inhibition (98). Further analysis of TME revealed that TIGIT was highly frequently expressed in TME cells from small B-cell lymphoma and ALCL, and its expression levels were significantly correlated with OS and PFS (99). In chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), negative expression of TME cells is associated with shorter OS. Subtype-specific differences exist in the co-expression patterns of TIGIT and PD-L1. These differences provide a molecular basis for accurately screening the beneficiary population for TIGIT inhibitor combined with PD-1/PD-L1 blockade (99).

These novel immune checkpoints and the classic PD - 1/CTLA - 4 form a multi-dimensional regulatory network. Their combined blockade strategies have demonstrated the potential to enhance the anti-tumor immune response in pre-clinical models. Current research centers on precisely identifying predictive biomarkers, developing novel combinatorial strategies, such as co-administration of TIM-3 inhibitors with PD-1 blockade, and exploring new drug modalities, including the small-molecule agent ML-T7. These efforts aim to address the clinical limitation of suboptimal response rates associated with existing immune checkpoint inhibitors and to establish a foundation for next-generation tumor immunotherapy paradigms.

Growing insights into the TME have increased focus on the mechanistic interaction between microbial populations and tumor dynamics. The intestinal microbiota has emerged as a key modulator in the onset, progression, and therapeutic responsiveness of lymphoma. Its influence on tumor immune microenvironment homeostasis is exerted through several regulatory pathways, including modulation of host immune responses, metabolic output, and pro-inflammatory signaling networks (111).

Routy B et al. found that fecal microbial transplantation (FMT) could enhance immunotherapy sensitivity. It does so by remodeling the microbiota structure. FMT intervention increased the response rate of PD-1 inhibitors in tumor-bearing mice by 2.3 times. The mechanism involved an increase in intestinal microbiota-mediated infiltration of CD8⁺ T cells. It also involved downregulating immunosuppressive factors such as IL-10 (112). The abundance of specific probiotics, such as Bifidobacteria and Akkermansia muciniphila, is positively correlated with the sensitivity of immunotherapy. Microbial metabolites such as short-chain fatty acids affect the progression of lymphoma by regulating the inflammatory pathway (113–115). In a clinical study (NCT04567446) involving 33 patients with B-cell lymphoma, the objective response rate of the Akkermansia muciniphila (Akk bacteria) high-abundance group (;1%) reached 78% after 6 months of CAR-T therapy. This rate was significantly higher than that of the low-abundance group. The abundance of Akk bacteria was positively correlated with CD8⁺ T cell infiltration and IFN-γ release in tumors/bone marrow. In the mouse model, the tumor volume of the Akk bacteria-supplemented group was reduced by 58%, 72%, and 65% on days 10, 34, and 46, respectively. Survival was also significantly prolonged in this group. Clinical translational studies have further verified the clinical value of microbiota regulation. In a multicenter cohort analysis, the diversity of the gut microbiota in patients with allogeneic hematopoietic cell transplantation (allo-HCT) was significantly correlated with survival. The high microbial diversity group showed a 29% to 51% reduction in mortality risk. Microbiota disruption was characterized by a significant loss of diversity accompanied by single-genus dominance. This pattern was a predictive indicator for developing graft-versus-host disease (116). Mendelian randomization studies have revealed species-specific effects from the perspective of causal associations. Faecalibacterium prausnitzii reduces the risk of Hodgkin lymphoma. Coprococcus antagonizes follicular lymphoma. Ruminococcaceae UCG-002 increases the risk of DLBCL by upregulating the inflammatory factor MIG (117, 118). Moreover, prospective studies of patients with DLBCL have shown the following. Dynamic changes in the abundance of Enterobacteriaceae during R-CHOP treatment are significantly associated with the risk of bacterial infection. An initial Enterobacteriaceae abundance greater than 4.5% is an independent prognostic indicator of infection risk. This observation supports the clinical adoption of early intervention strategies to preempt infection onset (119). These findings reveal the molecular mechanism by which the gut microbiota regulates the immune system against malignant tumors, which provides an important scientific basis for formulating personalized lymphoma treatment plans by adjusting the microbial community.

Recent advancements in nanotechnology are paving the way toward personalized therapeutic strategies for malignant tumors. Among these innovations, nucleic acid-based drug delivery systems, particularly mRNA vaccines, have emerged as a prominent area of investigation. A main challenge in this domain is overcoming intracellular delivery barriers associated with nucleic acid molecules while achieving targeted accumulation within the tumor immune microenvironment.

Kranz LM et al. developed charge-optimized RNA lipid complexes (RNA-LPX) for delivery system optimization. These complexes trigger an immune response to virus-like infections through ligand-independent DC-targeting mechanisms (120). Sasaki K’s team manipulated the size of mRNA lipid nanoparticles (LNPs) to 200–500 nm by microfluidic technology, and the optimal formulation A-11-LNP was screened to significantly enhance the RNA uptake and antigen presentation efficiency of spleen DCs (121). Mannose-modified lipid nanoparticles (STLNPs-Man) can increase mRNA delivery efficiency by a factor of 4. This is achieved through the DC mannose receptor-mediated endocytosis pathway. They also enhance the synergistic effect with immune checkpoint inhibitors by downregulating T cell CTLA-4 expression. This downregulation of CTLA-4 expression improves the combined efficacy (122). Moreover, a research group has engineered biomimetic nanoparticles cloaked with dendritic cell membranes (DPNs). They can enhance lysosomal escape efficiency through membrane fusion mechanisms (123). A team innovatively integrated a nanoplatform of CRISPR/Cas9 gene editing and photothermal therapy. It induces immunogenic cell death through PD-L1 knockout synergistically with sub-high temperature. This significantly promotes DC maturation and cytotoxic lymphocyte infiltration (124).

For the mechanism of drug resistance, a research team has developed an RGD-targeted peptide nanoplatform. This nanoplatform can specifically deliver Bcl-2 antisense oligonucleotides to αvβ3 integrin-high expression tumor cells. It achieves efficient gene silencing and adapts to various nucleic acid drugs (125). The nucleolin-targeted nanodrug PA-HM@DOX/ICG was functionalized with the AS1411 aptamer. This aptamer-based modification enables precise targeting of DLBCL cells by exploiting the overexpression of nucleolin on their surface. The combination therapy using this nanodrug resulted in a tumor inhibition rate of 91.5%. It also caused a significant reduction in toxicity (126). Carvalho S.M. and others integrated ZnS quantum dot imaging and CD20/CD19 double antibody targeting. This integration was based on a chitosan-functionalized nanoplatform. It made the apoptosis rate of NHL cells reach 82.4%. Meanwhile, the damage to normal B cells was less than 15% (127). Technological advancements, including improved drug delivery efficiency, integration of combinatorial therapeutic modalities, and the realization of synchronous diagnostic and therapeutic functions, have significantly expanded the application potential of nanomedicine in treating malignant tumors of the lymphatic system.

The rapid development of AI-assisted diagnosis and treatment platforms has pushed lymphoma to a new stage. AI technology integrates multi-omics data, such as genomics and transcriptomics, with deep learning algorithms. AI technology has achieved systematic innovation from molecular typing to treatment decision-making. Considering the EcoTyper framework as an example, a high-resolution TME map was constructed for the first time. This map contains 13 cell types and 44 cell states. It was created by collaboratively analyzing the bulk and single-cell RNA-seq data of hundreds of DLBCL cases. This breakthrough overcomes the limitations of traditional classification. It also reveals the dynamic interaction network between malignant B cells and TME (128). In therapeutic development, the Auto-RapTAC platform has significantly reduced the screening cycle for PROTAC molecules to just eight days by employing a modular, automated design. Using this high-throughput system, six potent degraders targeting CDK2, CDK12, and BCL6 have been successfully developed, thus substantially expediting the discovery and optimization of heterobifunctional drugs (129). AI can improve the efficiency of PROTAC development in the following ways. It uses machine learning to analyze structure-effect data and build predictive models. It performs virtual screening of potentially efficient candidates. It generates and evaluates the optimal design of new structures. It mines and analyzes experimental data to reveal key influencing factors. It assists in optimizing the synthesis route and improving preparation efficiency. For example, AlphaFold uses deep residual convolutional neural networks to efficiently capture complex patterns in protein data for accurate structure prediction. This provides key information for PROTAC design (130).

A deep learning-based whole-slide analysis tool for H&E staining demonstrated an overall accuracy of 0.932 in differentiating between FL, DLBCL, and CHL. Its multi-class diagnostic performance was comparable to that of expert pathologists and offered enhanced interpretability through heat map visualization. Virtual H&E staining technology achieved a staining quality pass rate of 92%, with diagnostic concordance closely matching that of conventional chemical staining (90% vs. 92%). These advancements collectively lay a robust foundation for the transition toward digital pathology (131, 132). (https://xulymphoma.shinyapps.io/PCDI_pred/) (127).

In terms of prognosis prediction, researchers integrated multi-omics data from 339 DLBCL patients. They aimed to construct a PCD index model. This model dynamically correlates ctDNA burden, minimal residual disease (MRD) status, and immune microenvironment characteristics. It does so through an online tool. The tool is designed to accurately identify patients with chemotherapy resistance to R-CHOP. It also guides targeted therapy selection (https://xulymphoma.shinyapps.io/PCDI_pred/) (133). Ferrández M.C. et al. developed a deep learning PET/CT prognostic model. This model demonstrated an area under the characteristic curve (AUC) of 0.71 in 1132 patients. The performance was significantly better than the International Prognostic Index (IPI). The model can predict treatment response without tumor segmentation (134). Song C’s team built a multimodal model based on 2.5D transfer learning. The integration of clinical features with radiomics data has enabled the development of a predictive model capable of accurately differentiating lymphoma from tuberculous lymphadenitis in patients with HIV/AIDS, achieving an AUC of 0.920 (135). AI has driven transformative progress in multiple pivotal areas, including elucidating molecular pathogenesis, accelerating drug discovery pipelines, and optimizing clinical decision-making processes. These advancements are progressively shaping a comprehensive precision medicine paradigm for lymphoma, thus promoting individualized diagnostic and therapeutic interventions from early detection through post-treatment monitoring.

ctDNA can be an important indicator for liquid biopsy. Analysis of the ctDNA mutational spectrum enables clinical target identification, dynamic monitoring of therapeutic efficacy, and early detection of disease recurrence. The ctDNA-based assessment of MRD has shown substantial advantages in individualized lymphoma treatment. Due to its high-sensitivity detection capabilities, this approach allows real-time tracking of tumor dynamics, thus providing timely and reliable data to inform and refine clinical decision-making.

In FL, high-sensitivity droplet digital PCR (ddPCR) revealed an EZH2 mutation frequency of 41.5%, significantly higher than the 27% detected through conventional single-point tissue biopsy. Liquid biopsy further identified undetected mutations in an additional 14 tissue samples. Dynamic monitoring showed that ctDNA levels in treatment-responsive patients declined by over 100-fold. Mutations could be detected and reproduced 6 months before recurrence, and EZH2 wild-type clones were more likely to infiltrate the bone marrow. This has expanded the scope of benefit for the precise use of inhibitors such as tazemetostat (136). In different lymphoma subtypes, ctDNA has distinct application values in disease monitoring and efficacy evaluation. For extranodal NK/T-cell lymphoma (ENKTL), ctDNA sequencing has identified BCOR, TP53, and DDX3X mutations associated with prognosis. The dynamics of these mutations can distinguish between remission and relapse (137). Intravascular large B-cell lymphoma (IVLBCL) is characterized by significantly elevated ctDNA concentrations compared to DLBCL. High-frequency mutations are frequently observed in MYD88 (56%) and CD79B (44%), with BCL6 mutations significantly enriched in cases with CNS involvement (138). In PTCL, ctDNA was detected in 95.7%, one-year PFS and OS were significantly reduced in the high-concentration group, and dynamic monitoring was consistent with imaging evaluation in 81.9% (139). ctDNA also plays a key role in predicting treatment response. Chen et al. developed a detection system for 29 TP53 mutation probes in patients with CAR-T-treated NHL. They found that the median PFS in the high ctDNA group was only 1.4 months. This was significantly shorter than that in the low ctDNA group (140). In cHL and DLBCL, ctDNA plasma concentrations were significantly negatively correlated with tumor mutational burden. Continuous monitoring of ctDNA could track clonal evolutionary trajectories in these malignancies. However, the threshold for ctDNA detection in FL needed to be optimized to improve clinical applicability (141, 142). Despite the limitations of some studies, such as small sample sizes and a constrained scope of detectable genetic alterations, blood-based ctDNA analysis has successfully addressed several inherent drawbacks of conventional tissue biopsies. In DLBCL, ctDNA technology enables simultaneous guidance of therapeutic regimen selection, dynamic monitoring of treatment responses, and early detection of disease recurrence. These capabilities substantially enhance the precision and continuity of clinical management, spanning the entire course from initial diagnosis to post-therapeutic surveillance.

The ctDNA-based assessment of MRD has shown substantial advantages in personalizing lymphoma treatment. Due to its high sensitivity and high-specificity detection capabilities, this modality enables real-time surveillance of tumor dynamics, thus providing timely and reliable molecular data to inform precise clinical adjustments to therapeutic strategies. The ctDNA-MRD real-time monitoring system can evaluate the treatment effect and predict recurrence by detecting changes in ctDNA levels, and guide the adjustment of treatment strategies. For patients with CAR-T-treated R/R LBCL, the median OS of ctDNA-positive patients at day 28 was only 6.7 months. None of the ctDNA-negative patients experienced disease progression. The sensitivity and specificity of MRD detection reached 83%-100%. This systematically verified the early warning value of ctDNA in evaluating CAR-T efficacy. Compared with traditional PET/CT, ctDNA can identify individuals at high risk of recurrence in advance and enable timely adjustment of treatment strategies (143). Soscia R et al. analyzed 73 patients with DLBCL. The analysis was based on MRD monitoring using immunoglobulin (IG) gene rearrangements in ctDNA. It further demonstrated the prognostic stratification efficacy of this monitoring method. MRD-negative patients had significantly better PFS than positive patients at mid-treatment and treatment end (144). This non-invasive detection technology not only enables effective stratification of patient groups based on prognostic risk but also facilitates clinical trial design incorporating dynamic risk adjustment. It holds significant promise as a foundational tool for developing personalized diagnostic and therapeutic strategies in clinical practice.

ctDNA mutation profiling and ctDNA-MRD real-time monitoring systems are essential technologies within liquid biopsy. Both methods are non-invasive and enable dynamic disease assessment. ctDNA mutation profiling is primarily used for early detection and prediction of treatment response, while ctDNA-MRD monitoring is applied during therapy to track disease progression and provide early relapse warnings. Despite their application at distinct clinical stages, these methodologies jointly contribute to advancing personalized medicine.

Positron emission tomography/computed tomography (PET/CT) is an image-based biomarker development method that enables accurate assessment based on the Lugano criteria through fusion imaging of metabolic activity (PET) and anatomical structure (CT) (145). However, traditional PET/CT detects lesions based on macroscopic tumor burden. This approach may miss microscopic residual lesions and lead to a false negative risk. It also has limitations such as radiation exposure and high economic costs. With the development of radiomics technology, in-depth analysis based on texture features is breaking through the traditional evaluation framework. The radiomics model established by Kim JJ et al. showed improved predictive accuracy relative to traditional parameters, such as metabolic tumor volume (MTV), in evaluating the therapeutic response to axi-cel CAR-T therapy in patients with relapsed or refractory DLBCL. The high shape complexity (PC value) of non-round/irregular nodules is strongly associated with poor prognosis. This provides a new dimension for non-invasive evaluation of immunotherapy response (137). In another study of mediastinal lymphoma subtypes (GZL/PMBCL/cHL), FDG-PET texture analysis was performed. It revealed that the metabolic activity of primary mediastinal large B-cell lymphoma (PMBCL) was significantly higher than that of other subtypes. The random forest model showed a dichotomous AUC of 0.87 for this analysis. The heterogeneous characteristics of PMBCL identified through this analysis could assist in pathological typing and puncture localization (146). These innovative methods transform medical imaging information into quantifiable diagnostic indicators by analyzing the differential characteristics of tumor cell metabolism. Although these technologies require further large-scale clinical validation and fundamental research, their application in disease stratification, treatment response assessment, and the development of personalized therapeutic strategies is driving innovation in clinical management models for lymphoma.

The National Cancer Institute’s (NCI) Molecular Analysis Therapy Selection (MATCH) trial uses an adaptive design model. It detects gene mutations, copy number variations, and fusion events in tumor tissues based on next-generation sequencing (NGS) technology. The trial combines gain-of-function/deletion mutations to accurately match lymphoma patients with specific molecular markers to the corresponding targeted therapy group (147). The establishment of a nationwide network of clinical laboratories has enabled the standardization of testing protocols across institutional samples, thus validating the feasibility of large-scale molecular subtyping in clinical translational settings (148).

The NCI-MATCH trial has demonstrated substantial clinical value in managing various solid tumors, including breast and colorectal cancers. Based on the molecular classification expertise developed through this platform, research efforts are now focused on establishing a gene-guided therapeutic framework for lymphoma. This approach aims to identify more individualized treatment strategies for patients with relapsed or refractory disease by analyzing tumor-specific molecular alterations. In the NCI-MATCH trial, Mansfield AS et al. reported cross-tumor efficacy in rare malignancies harboring ALK or ROS1 rearrangements. Among patients with ALK rearrangements, an objective response rate of 50% and a median PFS of 3.8 months were observed, outcomes that significantly surpassed those achieved with conventional chemotherapy regimens (149). Treatment with tazemetostat targeting EZH2/SMARCB1 mutant subsets in pediatric lymphoma resulted in disease control for six months or more. One patient experienced a sustained objective response. This demonstrates the therapeutic potential of epigenetic regulators in specific molecular subtypes (150). The trial uses a multi-arm design and integrates experts from different fields to simultaneously verify the effectiveness of other therapies. It accelerates the scientific research process and enhances the adaptability of treatment options, helping lymphoma patients obtain more personalized disease management strategies.

The incorporation of real-world evidence (RWE) into clinical research on lymphoma has emerged as a critical paradigm for validating therapeutic efficacy. The Flatiron Health database, an advanced platform integrating electronic health records (EHRs) with structured diagnostic and treatment data, offers distinct advantages. Its robust data mining capabilities enable comprehensive, multi-dimensional analyses, thus enhancing the evaluation of the clinical value and effectiveness of lymphoma treatment strategies. In a real-world cohort study involving 4336 patients with MCL, the median real-world overall survival (rwOS) was 35 months. This rwOS was significantly better than that observed with conventional chemotherapy. A high-risk subgroup (5% of the cohort) harboring a 17p deletion/TP53 mutation had a poor prognosis. BTK inhibitors have shown the capacity to partially reverse adverse outcomes in this subgroup, providing an evidence-based treatment option for older patients or those unable to tolerate intensive chemotherapy (151). In primary DLBCL, Breinholt et al. demonstrated that the prognostic relevance of TP53 mutations is consistent across cohorts. However, the clinical significance of PRDM1 and NOTCH2 mutations remains to be validated in larger patient populations (152). These findings highlight the need for dynamic optimization of molecular classification systems based on real-world data and evolving biological insights. Similarly, studies of FL and histological transformation (HT) have revealed a significant increase in the heterogeneity of tumor mutational burden and driver genes (such as EZH2, and CREBBP) as the disease progresses. This increased heterogeneity may affect the use of targeted drugs (such as EZH2 inhibitors). It highlights the need for dynamic monitoring of molecular profiles before treatment (153). The research method based on actual diagnosis and treatment records can use massive case information to evaluate the efficacy and safety of drugs and combine genetic testing results with patient recovery data to promote the discovery of new disease surveillance indicators.