Background: Pain empathy engages multiple neural networks. Neuroimaging research has identified anterior insula (aINS) and the sensorimotor network in processing pain empathy among women with primary dysmenorrhea (PDM). However, it remains unclear how the interaction between the two networks modulates pain empathy during menstrual pain. Methods: This case-control study included 56 women with PDM and 55 healthy controls (HCs). Participants were instructed to take functional MRI (fMRI) scans during the menstrual phase and view images depicting painful and non-painful situations to elicit empathy, as well as the short form of McGill Pain Questionnaire to elicit pain intensity. Functional connectivity (FC) analyses of the aINS-related and sensorimotor networks, and their interactions, were conducted to investigate their relationship with pain and pain empathy. Mediation analysis was performed to investigate the role of FC mediating menstrual pain and empathy. Results: PDM patients exhibited higher scores of pain empathy, which was significantly correlated with pain intensity. With comparable bilateral aINS-precentral/postcentral FC, left aINS-postcentral connectivity showed positive correlations with both pain intensity and empathy in PDM patients, whereas. the positive interaction between the left aINS- and precentral-centered networks was inversely associated with pain empathy in HCs. Furthermore, correlation and mediation analyses revealed that left aINS-right postcentral FC mediated the relationship between menstrual pain and pain empathy in PDM women. Conclusion: aINS-sensorimotor network interactions play a critical role in pain empathy, which may nonetheless engage in divergent regulatory processes under menstrual pain, shaping psychological adaptation and empathic responses.

Chronic non-bacterial prostatitis (CNP), a prevalent and debilitating urological disorder affecting 8.4% of men aged 15-60 years, presents significant clinical challenges due to the paucity of targeted therapies and poor patient adherence. To address this unmet medical need, we developed an innovative multifunctional nanoplatform (QM (Zn) NPs) by integrating Ti3C2 MXene with a quercetin-zinc coordination complex (Que-Zn) for precision CNP therapy. This system leverages chondroitin sulfate (Chs)-mediated CD44 targeting to achieve selective accumulation in inflamed prostate tissue, thereby enhancing Zn2+ bioavailability while enabling co-delivery of MXene and Que-Zn therapeutic payloads. Upon localization, QM (Zn) NPs orchestrate a coordinated therapeutic cascade: MXene scavenges reactive oxygen species (ROS) via electron-deficient sites, while Que-Zn drives M1-to-M2 macrophage repolarization and facilitates Zn2+ cellular uptake. The accumulated intracellular Zn2+ critically upregulates metallothionein 1 (Mt1), activating the IKK/NF-κB/IκB axis to resolve inflammation and oxidative damage. Transcriptomic analysis unequivocally identified Mt1 as the pivotal mediator of Zn2+-driven microenvironment reprogramming. Notably, QM (Zn) NPs not only significantly alleviated pelvic pain by mitigating neuronal oxidative stress but also exhibited excellent biocompatibility. This work pioneers a targeted nano-theranostic strategy that synergistically restores zinc homeostasis, quenches ROS, and reprograms immune responses, thereby establishing a transformative paradigm for CNP management.

BackgroundGastroenteropancreatic neuroendocrine tumors (GEP-NETs) display marked clinical heterogeneity, and conventional prognostic indicators such as TNM stage and Ki-67 index provide limited individualized risk discrimination. We aimed to develop and externally validate interpretable machine learning models for survival prediction.MethodsThis retrospective study enrolled 337 patients with histologically confirmed GEP-NETs, randomly partitioned into a training set (n = 236) and an independent test set (n = 101) with stratification by mortality. Six survival models, including unpenalized Cox regression as a baseline, LASSO Cox, random survival forest (RSF), gradient-boosted survival analysis (GBSA), Extra Trees, and DeepSurv, were trained with 5×5 nested cross-validation. Discrimination, probabilistic accuracy, calibration, and clinical utility were comprehensively evaluated. SHAP analysis identified key predictors and informed nomogram construction. External validation was performed in 51,225 SEER patients.ResultsOn the test set, Extra Trees achieved the highest discrimination (C-index 0.843, 95% CI 0.762-0.914) and balanced calibration (1-year ICI 0.031). Time-dependent AUC values were 0.871, 0.847 and 0.935 at 1, 3 and 5 years. The Extra Trees-based stratification identified high-risk patients with substantially worse survival (HR 8.93, 95% CI 3.35-23.84, P<0.001). The model outperformed TNM stage (C-index 0.772), Ki-67 (0.705) and tumor grade (0.699). Decision curve analysis demonstrated greater net benefit across threshold probabilities of 5%-50%. SHAP analysis identified T stage, tumor grade, M1 status, TNM stage and Ki-67 as the top predictors. External validation in SEER preserved discriminative ability (C-index 0.719) and excellent calibration (slopes 0.901-1.058).ConclusionInterpretable machine learning models outperform conventional approaches in predicting GEP-NET survival. Extra Trees showed the best internal discrimination and calibration, whereas DeepSurv achieved the highest external C-index. The simplified SHAP-derived nomogram provides a practical and well-calibrated but exploratory tool for individualized prognosis that requires prospective validation before clinical use.

Pulmonary fibrosis (PF) is a progressive and ultimately fatal lung disorder characterized by irreversible parenchymal scarring. Current therapeutic interventions are limited by suboptimal efficacy and significant adverse effects, highlighting a critical unmet need for the development of effective antioxidant therapeutic strategies. In this study, we engineered a novel nanotherapeutic platform composed of chondroitin sulfate (Chs)‐functionalized molybdenum disulfide nanosheets (CLM NSs) for targeted PF therapy. The system was synthesized through lipoic acid (LA)‐mediated conjunction of Chs to MoS2 nanosheets, significantly enhancing their physiological stability. CLM NSs utilize a CD44 receptor‐mediated targeting mechanism: initial pulmonary accumulation is achieved by attaching to circulating macrophages during the early inflammatory phase, followed by direct recognition of CD44 overexpressed on diseased alveolar epithelial cells in the late phase, enabling precise lesion‐specific accumulation. The nanosheets demonstrated robust broad‐spectrum antioxidant capacity, efficiently scavenging reactive oxygen/nitrogen species (ROS/RNS), restoring pulmonary redox homeostasis, and significantly inhibiting fibroblast activation and epithelial‐mesenchymal transition (EMT). In a bleomycin‐induced murine PF model, CLM NSs exerted significant therapeutic effects, markedly attenuating fibrotic progression, preserving alveolar architecture, and reducing pathological collagen deposition. Mechanistically, their therapeutic action involves inhibition of the critical ROS‐STAT1‐CXCL10 signaling axis, consequently suppressing macrophage recruitment and the inflammatory cascade. This nanoplatform, integrating efficient targeting, potent antioxidant activity, and excellent biocompatibility, represents a highly promising strategy for PF treatment.

BACKGROUND:Colorectal cancer (CRC) is associated with a high mortality rate. Previous studies have shown that FOXQ1, MMP11, and CST1 play significant roles in various cancers, influencing the invasion and metastasis of tumors. However, their effects on colorectal cancer have not been fully investigated. The purpose of this research was to examine the expression of FOXQ1, MMP11, and CST1 in colorectal cancer (CRC) and to systematically as-sess how these factors relate to clinicopathological characteristics and patient survival outcomes. METHODS:This study retrospectively gathered paraffin-embedded samples from 110 CRC patients who underwent surgery between 2017 and 2018. Meanwhile, relevant data were obtained from public databases to analyze expression differences of FOXQ1, MMP11, and CST1 between tumor tissues and normal lung tissues. We examined the expression of FOXQ1, MMP11, and CST1 using immunohistochemistry. Furthermore, the associations among FOXQ1, MMP11, CST1, clinical-pathological parameters, and prognosis were systematically analyzed. Further verification of the in vitro results was conducted through qRT-PCR. RESULTS:Expression of FOXQ1, MMP11, and CST1 in patients was high, with 83.6%, 67%, and 74.5%, respectively. Through rigorous quantitative analysis of clinical-pathological parameters, the study confirmed that these biomarkers have a close and clinically significant correlation with the progression of TNM staging and the occurrence of lymph node metastasis (p < 0.05). Bioinformatics analysis and qRT-PCR verification both indicated that the expression levels of FOXQ1, MMP11, and CST1 in colorectal cancer (CRC) tissues were significantly higher than those in adjacent non-cancerous tissues. CONCLUSIONS:The research data indicate that the abnormal overexpression of FOXQ1, MMP11, and CST1 in CRC tissues is significantly correlated with poor clinical prognosis in patients. There may be a synergistic effect influencing the invasion and metastasis of tumor cells, positioning them as potential novel therapeutic targets for patients with CRC.

Background Colonic mucinous adenocarcinoma is a highly aggressive subtype of colon cancer. Epidemiological evidence has linked estrogen exposure to certain gastrointestinal malignancies, yet its pathophysiological role in the pathogenesis of colonic mucinous adenocarcinoma remains unclear. This study aimed to investigate the effect of estrogen on tumor growth in a rat model of colonic mucinous adenocarcinoma and its association with intestinal mucosal epithelial proliferation-related factors and apoptosis. Results Compared with the normal control group, the model control group showed significantly increased tumor proliferative activity and elevated intestinal epithelial apoptosis. After estrogen intervention, tumor volume and proliferative activity were inhibited, the expressions of Ki-67 and proliferating cell nuclear antigen were downregulated, and the apoptosis rate of intestinal epithelial cells decreased; an estrogen antagonist could enhance some of the above effects. Conclusions Estrogen can slow the growth of colonic mucinous adenocarcinoma by inhibiting intestinal mucosal epithelial proliferation, exhibiting potential anti-tumor effects. Its mechanism may be related to regulating the proliferation-apoptosis balance of intestinal mucosal epithelium, providing experimental evidence for colon cancer therapy research.How to cite: Wang K, Yue Y, Liu X, et al. Estrogen attenuates colonic mucinous adenocarcinoma growth in a rat model by inhibiting proliferation and promoting apoptosis. Electron J Biotechnol 2026;83. https://doi.org/10.1016/j.ejbt.2026.100725.

PURPOSE: Adrenal myelolipoma (AML) is a benign tumor composed of intermixed adipose and hematopoietic tissues, but its detailed cellular composition remains unclear. CXCL12-abundant reticular (CAR) cells have been found to promote hematopoiesis during AML development; however, the mechanism of adipogenesis in AML remains unclear. This study aimed to characterize the cellular composition of AML and elucidate the potential mechanisms underlying its development, with a particular focus on adipocyte origin. METHODS: Four AML specimens and matched adjacent adrenal tissues were subjected to single-nucleus RNA sequencing, and an additional ten paired samples were analyzed using immunostaining. Complementary in vitro experiments were performed to validate the proposed mechanisms of AML development. RESULTS: Unsupervised clustering revealed that AML is composed predominantly of T cells, B cells, neural-like cells, CAR cells, adipocytes, and nearly all types of myeloid cells. RNA velocity analysis suggested that CAR cells might be the lineage source of adipocytes. Consistently, CAR cells isolated from AML tissues demonstrated adipogenic capacity in vitro. Expression analysis showed high levels of adrenocorticotropic hormone receptor and androgen receptor in adrenal cortical cells and endothelial cells respectively. CellChat analysis further revealed extensive paracrine signals from adrenal cortical cells and endothelial cells to CAR cells. Overexpression of androgen receptor in endothelial cells transcriptionally upregulated key regulators of adipogenesis including COL4A1 and PDGFD. CONCLUSIONS: AML comprises a heterogeneous population of immune and stromal cells, with CAR cells likely serving as the primary source of adipocytes. Moreover, androgen may regulate CAR cells to induce adipocytes via endothelial cells.

Cytoreductive surgery is the treatment protocol for colorectal cancer. Nonetheless, a major medical challenge remains to fully eliminate malignant tumor cells, along with a number of complications such as peritoneal adhesion and tumor peritoneal metastasis. The occurrence of peritoneal adhesions compromises not only the ability to do subsequent surgery, but also the efficacy of adjunct chemotherapy. More and more evidence suggest that the process of mesothelial-mesenchymal transition (MMT) influenced by transforming growth factor-β1 (TGF-β1) has a role to play in these disturbances, therefore making TGF-β1 a viable target for therapy. This study has designed a hydrogel-based physical barrier drug delivery system loaded with RNA interference technology, designated as FC@MT. The 5-fluorouracil (5-FU), which is known for its antitumor effects, was firmly linked to the FCGCM hydrogel matrix through the formation of hydrogen bonds. Meanwhile, APTES-modified mesoporous silica nanoparticle (AMSN)/TGF-β1 siRNA complexes were incorporated to facilitate the cellular uptake of siRNA and enable their escape from lysosomes. The localized co-delivery of 5-FU and TGF-β1 siRNA induces residual tumor cells killing by silencing TGF-β1 expression and reverses MMT. The combination of FC@MTs was shown to have a synergistic anti-peritoneal metastasis and anti-adhesion effects, which could be an effective strategy to enhance the clinical therapeutics of CRC.

IntroductionAcute stress can disrupt gastrointestinal function. Interstitial cells of Cajal (ICCs) regulate gastric motility; however, the effects underlying their alterations during stress remain unclear. We aimed to explore the potential impact of acute stress on alterations in gastric ICCs.MethodsAnimals were assigned to one control and four study groups (0, 24, and 72 h). Acute stress was induced via right chest puncture using a Hopkinson bar. Lung injury was assessed using hematoxylin-eosin staining, gastric ICCs were examined using immunohistochemistry, immunohistofluorescence, and TdT-mediated dUTP nick-end labeling assays. Transmission electron microscopy was used to evaluate ultrastructural changes. Western blotting and real-time polymerase chain reaction were used to analyze SCF/c-kit, protein 53 (p53), and the Wnt/β-catenin signaling pathways. Enzyme-linked immunosorbent assay was used to quantify nuclear factor kappa B (NF-κB) p65, interleukin (IL)-1β, IL-6, IL-9, IL-10, tumor necrosis factor-alpha (TNF-α), and cortisol. Spectrophotometry was used to determine the levels of reactive oxygen species (ROS) and malondialdehyde (MDA).ResultsThe lung tissue exhibited congestion, edema, and the destruction of the alveolar structure following puncture. The gastric ICCs tested positive for the surface markers CD117/c-kit and CD34. The population and apoptosis of the gastric ICCs were altered in the acute stress condition. Moreover, acute stress caused dysregulated SCF/c-kit, Wnt/β-catenin, p53 and NF-κB pathways. IL-1β, IL-6, IL-9, IL-10, TNF-α, and cortisol demonstrated altered expression levels, whereas ROS and MDA were upregulated in the acute stress condition.ConclusionAcute stress damages gastric ICCs through the Wnt/β-catenin, JAK/STAT, and NF-κB signaling pathways, as well as p53-and SCF/c-kit pathway-mediated abnormalities and oxidative stress, contributing to functional alterations in gastric ICCs.

Suppurative otitis media (SOM) is a condition characterized by the invasion of pathogenic bacteria into the middle ear via the eustachian tube (ET) or tympanic membrane (TM), significantly impacting the health of patients. Ear drops are frequently employed in clinical practice. Still, their efficacy is limited by their inability to penetrate the TM and their susceptibility to clearance through the ET, which does not address the comprehensive treatment needs. In this study, a novel injectable hydrogel (HGCCB hydrogel) is fabricated by the radical copolymerization and dynamic covalent bonding using chlorogenic acid-grafted methacrylate hyaluronic acid (HA-GCGA) and phenylboronic acid-modified ciprofloxacin (Cip-PBA) without any chemical crosslinkers. HGCCB hydrogel demonstrates favorable injectability, significant antibacterial activity, and anti-inflammatory properties. Importantly, dynamic boronate ester bonds within the HGCCB hydrogel facilitate the rapid release of Cip-PBA upon injection into the tympanic cavity, effectively targeting bacterial pathogens. Meanwhile, HAGCGA acts as a prodrug, enabling the sustained release of chlorogenic acid through ester bond hydrolysis, thereby achieving a sequential drug release profile. Furthermore, the intratympanic injection of HGCCB hydrogel successfully treated SOM in a rat model, mitigating bacterial infection and inflammatory responses in the middle ear. This functional hydrogel presents a straightforward and effective strategy for treating SOM.

Sequential drug delivery hydrogels (SDDHs), which mimic the natural spatiotemporal regulation of physiological processes, have emerged as promising platforms for enhancing therapeutic efficacy and minimizing drug interactions. However, existing strategies relying on spatial structuring or affinity-based mechanisms face challenges in complexity and applicability. Inspired by the metastable nature of noncovalent and dynamic covalent interactions, this study proposes a novel sequential release system leveraging the energy disparity between lowbinding-energy hydrogen bonds (-0.5 to -50 kcal/mol) and high-binding-energy disulfide bonds (-70 kcal/ mol). This design enables spontaneous, bioenvironment-suitable drug release without requiring multilayered architecture or specific triggers. Experimental validation demonstrates significantly distinct release profiles for dual drugs and improved healing outcomes in biofilm infected wounds, highlighting the potential of integratingbonds driven strategies for dynamic therapeutic delivery.

Postoperative adhesions are a common complication following abdominal surgery, for which current preventive strategies demonstrate limited efficacy. To address this, a novel Bletilla striata polysaccharide (BSP)-based interpenetrating network (IPN) hydrogel as an active barrier was developed. This design integrates the pharmacological activity and biocompatibility of BSP with the robust mechanical and antibiofouling properties of poly(N-hydroxyethyl acrylamide) (PHEAA), based on the pathological characteristics of postoperative adhesions. The physical network of BSP and the chemical network of PHEAA were interpenetrated and entangled through hydrogen bonding, forming a stable BSP/PHEAA IPN structure. A specific BSP/PHEAA hydrogel formulation was selected for its optimal mechanical strength, fouling resistance, degradability, and excellent biocompatibility, stability, and hemostatic properties. The efficacy of the hydrogel in preventing both primary and recurrent postoperative adhesions was assessed using rat cecal-abdominal wall abrasion and postadhesiolysis recurrence models, an effect presumably attributable to its combined barrier function and inherent bioactivity. Molecular biology studies confirmed that the BSP/PHEAA hydrogel modulates the balance between M1 and M2 macrophages and alleviates inflammation. Furthermore, it regulated the MMP-9/TIMP-1 balance by suppressing NF-κB p65 expression, thereby inhibiting fibrosis. Thus, this hydrogel demonstrates significant potential as an active barrier for preventing both primary and recurrent adhesions.

Triple-negative breast cancer (TNBC) is a highly aggressive malignancy lacking targeted therapeutic modalities. Aptamers are promising targeting agents with high specificity and affinity. In this study, we identified a chemically modified nucleic acid aptamer (AptT1) that specifically recognized and internalized into TNBC cells. Mechanistic studies revealed that ANXA2 was an essential protein in facilitating AptT1 internalization via clathrin-mediated endocytosis. Furthermore, we constructed an aptamer-drug conjugate (ApDC) to selectively deliver a cytotoxic agent (SN38) to TNBC cells and demonstrated significant efficacy and safety in a TNBC xenograft model. These findings suggest that AptT1-based ApDC holds great potential as an innovative targeted strategy for TNBC.

Calcium-regulated heat-stable protein 1 (CARHSP1) has been identified as a cold shock domain (CSD) protein family member, participating in the regulation of ribosomal translation, mRNA degradation, and the rate of transcription termination. However, there is an extremely limited understanding of the function of CARHSP1 as an RNA binding protein (RBP) in prostate cancer (PCa). The expression pattern of CARHSP1 and the correlation between the CARHSP1 expression and clinical prognosis in PCa patients were analyzed by using multiple public databases. In vitro and in vivo functional assays were conducted to assess the role of CARHSP1. The mechanisms of CARHSP1 function on IL-17RA were identified by RNA pull-down and RNA stability assays. A co-culture model of Jurkat cells and PCa cells was established to investigate the potential role of CARHSP1 in tumor immunity of PCa. CARHSP1 was highly expressed in PCa, and correlated with advanced characteristics of PCa and unfavorable prognosis in PCa patients. Moreover, knockdown of CARHSP1 significantly dampened the capacity of proliferation, migration, invasion, and immune evasion of PCa cells in vitro and in vivo. Mechanistically, the RNA-binding protein CARHSP1 selectively bound to the mRNA of IL-17RA, resulting in the increased expression of both IL-17RA mRNA and protein. Downregulating expression of CARHSP1 shortened the half-life of IL-17RA mRNA and reduced its expression. Subsequently, the downstream pathways of IL-17RA, JAK-STAT3 signaling pathway and NF-κB signaling pathway, were activated by CARHSP1 and contributed to the malignant phenotype of PCa cells. In conclusion, our results demonstrated that the increased expression of CARHSP1 in PCa is correlated with advanced clinical characteristics and unfavorable prognosis, and CARHSP1 may promote the progression of PCa through enhancing the mRNA stability of IL-17RA and activating its downstream pathways. These results suggest that CARHSP1 is an important regulator of tumor microenvironment in PCa, and CARHSP1-IL-17RA axis could be potential novel therapeutic targets for PCa.

Extracellular vesicles (EVs) play a crucial role as important mediators of intercellular communication in the progression of tumors. The capture and analysis of tumor-derived EVs offer new possibilities for the application of cancer liquid biopsies. This study aims to construct a DNA tetrahedral nanostructure that specifically recognizes HER2 and CD63, enabling the effective enrichment and detection of HER2-expressing EVs (HEVs). We enriched HEVs from cell lines and 13 random clinical samples and validated their characteristics by dynamic light scattering, transmission electron microscopy, and Western blotting. Further, we detected HEVs levels in clinical samples. The HEVs levels in HER2-positive breast cancer patients were significantly higher than those in healthy/benign controls (mean, 4.737 vs 4.160 vs 4.144 U/μL, P < 0.0001), displaying a concentration gradient across different HER2 expression levels. This study establishes an approach for HEV detection, thus providing a new tool for the diagnosis of HER2-positive breast cancer.

Triple-negative breast cancer is therapeutically challenging due to the low expression of tumor markers and 'cold' tumor immunosuppressive microenvironment. Here, we present a dual-targeting peptide-drug conjugate (PDC) for tumor inhibition. Our PDC efficiently and selectively delivers cytotoxic Monomethyl Auristatin E (MMAE) into tumor cells via C-X-C chemokine receptor type 4 (CXCR4) and folate receptor 1 (FOLR1) for synergistic inhibition of growth and metastasis. Our results show that the dual-targeting PDC has potent antitumor activity in cultured human cells and several murine transplanted tumor models without apparent toxicity. The combination of dual-targeting PDC and radiotherapy modulates the tumor immunosuppressive microenvironment by increasing CD8+ T cell infiltration and attenuating the proportion of myeloid-derived suppressor and regulatory T cells. Therefore, our dual-targeting PDC represents a promising new strategy for cancer therapy that rebalances the immune system and promotes tumor regression.

Macrophage extracellular traps (METs) are protein network structures released by macrophages during immune defense initiation and inflammation regulation. Evidence shows that these protein structures can help kill pathogens. However, overproduction of METs can harm organisms. Therefore, novel strategies to inhibit the excessive production of METs during immune reactions need to be developed. This study introduces a novel type of anti-inflammatory and antibacterial nanocomposites (i.e., ZIF-8/Ce@rutin NPs) developed by doping the nanozyme Ce into the antibacterial zeolitic imidazole framework-8 (ZIF-8) and then loading the reactive oxygen species (ROS) scavenger rutin via electrostatic adsorption. Results showed that ZIF-8/Ce@rutin exerted bactericidal effects on gram-positive and -negative bacteria, thereby inhibiting MET production from the source. Furthermore, ZIF-8/Ce@rutin scavenged ROS and induced macrophage polarization into the anti-inflammatory M2 type, thereby inhibiting MET progression. This research shows that ZIF-8/Ce@rutin can restrain MET overproduction, providing a theoretical basis for using ZIF-8/Ce@rutin to treat inflammatory and infectious diseases.

Background: The pathological response to neoadjuvant chemotherapy (NAC) is an established predictor of long-term outcomes in breast cancer. However, conventional binary assessment based solely on pathological complete response (pCR) fails to capture prognostic heterogeneity across molecular subtypes. This study aimed to develop an interpretable deep learning model that integrates multiple clinical and pathological variables to predict both recurrence and metastasis development following NAC treatment. Methods: We conducted a retrospective analysis of 832 breast cancer patients who received NAC between 2013 and 2022. The analysis incorporated five key variables: tumor size changes, nodal status, Ki-67 index, Miller–Payne grade, and molecular subtype. A Multi-Layer Perceptron (MLP) model was implemented on the PyTorch platform and systematically benchmarked against SVM, Random Forest, and XGBoost models using five-fold cross-validation. Model performance was assessed by calculating the area under the curve (AUC), accuracy, precision, recall, and F1-score, and by analyzing confusion matrices. Results: The MLP model achieved AUC values of 0.86 (95% CI: 0.82–0.93) for HER2-positive cases, 0.82 (95% CI: 0.70–0.92) for triple-negative cases, and 0.76 (95% CI: 0.66–0.82) for HR+/HER2-negative cases. SHAP analysis identified post-NAC tumor size, Ki-67 index, and Miller–Payne grade as the most influential predictors. Notably, patients who achieved pCR still had a 12% risk of developing recurrence, highlighting the necessity for ongoing risk assessment beyond binary response evaluation. Conclusions: The proposed deep learning system provides precise and interpretable risk assessment for NAC patients, facilitating individualized treatment approaches and post-treatment monitoring plans.