Constraints on regional projections of mean and extreme precipitation under warming.

The projected changes in the hydrological cycle under global warming remain highly uncertain across current climate models. Here, we demonstrate that the observational past warming trend can be utilized to effectively co1nstrain future projections in mean and extreme precipitation on both global and regional scales. The physical basis for such constraints relies on the relatively constant climate sensitivity in individual models and the reasonable consistency of regional hydrological sensitivity among the models, which is dominated and regulated by the increases in atmospheric moisture. For the high-emission scenario, on the global average, the projected changes in mean precipitation are lowered from 6.9 to 5.2% and those in extreme precipitation from 24.5 to 18.1%, with the inter-model variances reduced by 31.0 and 22.7%, respectively. Moreover, the constraint can be applied to regions in middle-to-high latitudes, particularly over land. These constraints result in spatially resolved corrections that deviate substantially and inhomogeneously from the global mean corrections. This study provides regionally constrained hydrological responses over the globe, with direct implications for climate adaptation in specific areas.

A Highly Sensitive and Robust Method for Genome-wide 5hmC Profiling of Rare Cell Populations.

We present a highly sensitive and selective chemical labeling and capture approach for genome-wide profiling of 5-hydroxylmethylcytosine (5hmC) using DNA isolated from ∼1,000 cells (nano-hmC-Seal). Using this technology, we assessed 5hmC occupancy and dynamics across different stages of hematopoietic differentiation. Nano-hmC-Seal profiling of purified Tet2-mutant acute myeloid leukemia (AML) murine stem cells allowed us to identify leukemia-specific, differentially hydroxymethylated regions that harbor known and candidate disease-specific target genes with differential 5hmC peaks compared to normal stem cells. The change of 5hmC patterns in AML strongly correlates with differential gene expression, demonstrating the importance of dynamic alterations of 5hmC in regulating transcription in AML. Together, covalent 5hmC labeling offers an effective approach to study and detect DNA methylation dynamics in in vivo disease models and in limited clinical samples.

Relationship Analysis Between Pericoronary Fat Attenuation Index and Parameters of Single Plaque.

OBJECTIVE: The aim of the study is to investigate the relationship between plaque parameters and pericoronary fat attenuation index (FAI). METHODS: A retrospective collection was performed on 227 patients with coronary heart disease who underwent coronary computed tomography angiography examinations in our hospital from May 2021 to April 2023, with a total of 254 right coronary or left anterior descending coronary arteries exhibiting solitary plaques within the FAI measurement area. Based on whether the proximal coronary FAI value was ≥ -70.0 HU, patients and coronary arteries were divided into FAI-positive group (67 cases, 73 coronary arteries) and FAI-negative group (160 cases, 181 coronary arteries). Quantitative parameters of coronary solitary plaques were collected, including stenosis severity, plaque length, plaque volume, plaque composition ratios, minimal luminal area, and calcification score, as well as qualitative parameters such as plaque types and high-risk plaques. Differences in plaque parameters between the FAI-positive and FAI-negative groups were compared. RESULTS: The proportion of positive remodeling in the FAI-positive group (73 coronary arteries) was higher than that in the FAI-negative group (181 coronary arteries) with statistical significance (89.0% vs 78.5%, P = 0.049). Multivariate analysis revealed that positive remodeling was a risk factor for abnormal FAI values in solitary plaques (odds ratio, 2.271, P = 0.049). CONCLUSIONS: The FAI-positive group had a higher proportion of positive remodeling, and positive remodeling was an independent risk factor for positive FAI values.

Dry and moist dynamics shape regional patterns of extreme precipitation sensitivity.

Responses of extreme precipitation to global warming are of great importance to society and ecosystems. Although observations and climate projections indicate a general intensification of extreme precipitation with warming on global scale, there are significant variations on the regional scale, mainly due to changes in the vertical motion associated with extreme precipitation. Here, we apply quasigeostrophic diagnostics on climate-model simulations to understand the changes in vertical motion, quantifying the roles of dry (large-scale adiabatic flow) and moist (small-scale convection) dynamics in shaping the regional patterns of extreme precipitation sensitivity (EPS). The dry component weakens in the subtropics but strengthens in the middle and high latitudes; the moist component accounts for the positive centers of EPS in the low latitudes and also contributes to the negative centers in the subtropics. A theoretical model depicts a nonlinear relationship between the diabatic heating feedback (α) and precipitable water, indicating high sensitivity of α (thus, EPS) over climatological moist regions. The model also captures the change of α due to competing effects of increases in precipitable water and dry static stability under global warming. Thus, the dry/moist decomposition provides a quantitive and intuitive explanation of the main regional features of EPS.

Dynamic amplification of extreme precipitation sensitivity.

A useful starting hypothesis for predictions of changes in precipitation extremes with climate is that those extremes increase at the same rate as atmospheric moisture does, which is [Formula: see text] following the Clausius-Clapeyron (CC) relation. This hypothesis, however, neglects potential changes in the strengths of atmospheric circulations associated with precipitation extremes. As increased moisture leads to increased precipitation, the increased latent heating may lead to stronger large-scale ascent and thus, additional increase in precipitation, leading to a super-CC scaling. This study investigates this possibility in the context of the 2015 Texas extreme precipitation event using the Column Quasi-Geostrophic (CQG) method. Analogs to this event are simulated in different climatic conditions with varying surface temperature ([Formula: see text]) given the same adiabatic quasigeostrophic forcing. Precipitation in these events exhibits super-CC scaling due to the dynamic contribution associated with increasing ascent due to increased latent heating, an increase with importance that increases with [Formula: see text] The thermodynamic contribution (attributable to increasing water vapor; assuming no change in vertical motion) approximately follows CC as expected, while vertical structure changes of moisture and diabatic heating lead to negative but secondary contributions to the sensitivity, reducing the rate of increase.

Antioxidant activity screening and chemical constituents of the essential oil from rosemary by ultra-fast GC electronic nose coupled with chemical methodology.

BACKGROUND: Traditional chemical methods were mainly used to evaluate the total antioxidant activity of essential oils. How to determine the bioactivity of each compound in mixtures is an interesting research topic. Nowadays, an ultra-fast gas chromatography electronic nose (E-nose) has been gradually used in the detection of volatile compounds, but the screening of the active components of essential oils has not been reported. E-nose coupled with chemical methodology was established using the essential oil from rosemary (EOR) as a specific application example. The proposed method can both identify the chemical constituents of EOR and quickly screen the antioxidant by comparing the change of chromatographic peak area of every component in EOR before and after reaction with free radicals. RESULTS: Among all chemical constituents in EOR, verbenone, eucalyptol and o-cymene showed the strongest scavenging abilities in 1,1'-diphenyl-2-picrylhydrazine (DPPH·), 2,2'-azino-bis(3-ethyl-benzothiazoline-6-sulphonate) (ABTS·+ ) and hydroxyl (·OH) radicals, respectively, with scavenging rates of 67.9%, 39.5%, and 69.9%. The reliability and feasibility of using E-nose to identify chemical constituents of EOR were verified by gas chromatography-tandem mass spectrometry (GC-MS/MS). The GC-MS/MS results showed that the main components of EOR were α-pinene (422.2 µg g-1 ), p-cymene (208.4 µg g-1 ), camphor (203.5 µg g-1 ), verbenone (160.2 µg g-1 ), and eucalyptol (129.1 µg g-1 ). CONCLUSIONS: The E-nose methods can be used as a complementary method to traditional spectrophotometric techniques. Furthermore, this study will be of great significance for the rapid screening of antioxidant active components in essential oils from natural products. © 2020 Society of Chemical Industry.

Genome-wide mapping of 5-hydroxymethylcytosines in circulating cell-free DNA as a non-invasive approach for early detection of hepatocellular carcinoma.

OBJECTIVE: The lack of highly sensitive and specific diagnostic biomarkers is a major contributor to the poor outcomes of patients with hepatocellular carcinoma (HCC). We sought to develop a non-invasive diagnostic approach using circulating cell-free DNA (cfDNA) for the early detection of HCC. DESIGN: Applying the 5hmC-Seal technique, we obtained genome-wide 5-hydroxymethylcytosines (5hmC) in cfDNA samples from 2554 Chinese subjects: 1204 patients with HCC, 392 patients with chronic hepatitis B virus infection (CHB) or liver cirrhosis (LC) and 958 healthy individuals and patients with benign liver lesions. A diagnostic model for early HCC was developed through case-control analyses using the elastic net regularisation for feature selection. RESULTS: The 5hmC-Seal data from patients with HCC showed a genome-wide distribution enriched with liver-derived enhancer marks. We developed a 32-gene diagnostic model that accurately distinguished early HCC (stage 0/A) based on the Barcelona Clinic Liver Cancer staging system from non-HCC (validation set: area under curve (AUC)=88.4%; (95% CI 85.8% to 91.1%)), showing superior performance over α-fetoprotein (AFP). Besides detecting patients with early stage or small tumours (eg, ≤2.0 cm) from non-HCC, the 5hmC model showed high capacity for distinguishing early HCC from high risk subjects with CHB or LC history (validation set: AUC=84.6%; (95% CI 80.6% to 88.7%)), also significantly outperforming AFP. Furthermore, the 5hmC diagnostic model appeared to be independent from potential confounders (eg, smoking/alcohol intake history). CONCLUSION: We have developed and validated a non-invasive approach with clinical application potential for the early detection of HCC that are still surgically resectable in high risk individuals.

Targeting PI3K in cancer: mechanisms and advances in clinical trials.

Phosphatidylinositol-3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) signaling is one of the most important intracellular pathways, which can be considered as a master regulator for cancer. Enormous efforts have been dedicated to the development of drugs targeting PI3K signaling, many of which are currently employed in clinical trials evaluation, and it is becoming increasingly clear that PI3K inhibitors are effective in inhibiting tumor progression. PI3K inhibitors are subdivided into dual PI3K/mTOR inhibitors, pan-PI3K inhibitors and isoform-specific inhibitors. In this review, we performed a critical review to summarize the role of the PI3K pathway in tumor development, recent PI3K inhibitors development based on clinical trials, and the mechanisms of resistance to PI3K inhibition.

5-Hydroxymethylcytosines in Circulating Cell-Free DNA Reveal Vascular Complications of Type 2 Diabetes.

BACKGROUND: Long-term complications of type 2 diabetes (T2D), such as macrovascular and microvascular events, are the major causes for T2D-related disability and mortality. A clinically convenient, noninvasive approach for monitoring the development of these complications would improve the overall life quality of patients with T2D and help reduce healthcare burden through preventive interventions. METHODS: A selective chemical labeling strategy for 5-hydroxymethylcytosines (5hmC-Seal) was used to profile genome-wide 5hmCs, an emerging class of epigenetic markers implicated in complex diseases including diabetes, in circulating cell-free DNA (cfDNA) from a collection of Chinese patients (n = 62). Differentially modified 5hmC markers between patients with T2D with and without macrovascular/microvascular complications were analyzed under a case-control design. RESULTS: Statistically significant changes in 5hmC markers were associated with T2D-related macrovascular/microvascular complications, involving genes and pathways relevant to vascular biology and diabetes, including insulin resistance and inflammation. A 16-gene 5hmC marker panel accurately distinguished patients with vascular complications from those without [testing set: area under the curve (AUC) = 0.85; 95% CI, 0.73-0.96], outperforming conventional clinical variables such as urinary albumin. In addition, a separate 13-gene 5hmC marker panel could distinguish patients with single complications from those with multiple complications (testing set: AUC = 0.84; 95% CI, 0.68-0.99), showing superiority over conventional clinical variables. CONCLUSIONS: The 5hmC markers in cfDNA reflected the epigenetic changes in patients with T2D who developed macrovascular/microvascular complications. The 5hmC-Seal assay has the potential to be a clinically convenient, noninvasive approach that can be applied in the clinic to monitor the presence and severity of diabetic vascular complications.

Diagnostic and grading accuracy of 18F-FDOPA PET and PET/CT in patients with gliomas: a systematic review and meta-analysis.

BACKGROUND: Positron emission tomography (PET) and PET/computed tomography (PET/CT) imaging with 3,4-dihydroxy-6-[18F] fluoro-L-phenylalanine (18F-FDOPA) has been used in the evaluation of gliomas. We performed a meta-analysis to obtain the diagnostic and grading accuracy of 18F-FDOPA PET and PET/CT in patients with gliomas. METHODS: PubMed, Embase, Cochrane Library and Web of Science were searched through 13 May 2019. We included studies reporting the diagnostic performance of 18F-FDOPA PET or PET/CT in glioma patients. Pooled sensitivity, specificity, and area under the summary receiver operating characteristic (SROC) curve were calculated from eligible studies on a per-lesion basis. RESULTS: Eventually, 19 studies were included. Across 13 studies (370 patients) for glioma diagnosis, the pooled sensitivity and specificity of 18F-FDOPA PET and PET/CT were 0.90 (95%CI: 0.86-0.93) and 0.75 (95%CI: 0.65-0.83). Across 7 studies (219 patients) for glioma grading, 18F-FDOPA PET and PET/CT showed a pooled sensitivity of 0.88 (95%CI: 0.81-0.93) and a pooled specificity of 0.73 (95%CI: 0.64-0.81). CONCLUSIONS: 18F-FDOPA PET and PET/CT demonstrated good performance for diagnosing gliomas and differentiating high-grade gliomas (HGGs) from low-grade gliomas (LGGs). Further studies implementing standardized PET protocols and investigating the grading parameters are needed.

Robotic-Assisted Radical Hysterectomy Results in Better Surgical Outcomes Compared With the Traditional Laparoscopic Radical Hysterectomy for the Treatment of Cervical Cancer.

OBJECTIVE: The aim of this study was to compare the surgical outcomes of robotic-assisted radical hysterectomy (RRH) with traditional laparoscopic radical hysterectomy (TLRH) for the treatment of early-stage cervical cancer in a large retrospective cohort of a total of 933 patients. METHODS: We have enrolled 100 patients into the RRH and 833 patients into the TLRH group. The surgical outcomes include operating time, blood loss, transfusion rate, pelvic lymph node yield, hospitalization days, duration of bowel function recovery, catheter removal before and after 3 weeks, conversion to laparotomy, and intraoperative and postoperative complications. Follow-up results were also analyzed for all patients. RESULTS: Both groups have similar patient and tumor characteristics but patients with a larger lesion size were preferably enrolled in the TLRH treatment group. The treatment with RRH was generally superior to TLRH with respect to operating time, blood loss, length of hospitalization, duration of bowel function recovery, and postoperative complications. On follow-up of patients, there were no relapses reported in the RRH group compared with 4% of relapse cases and 2.9% of deaths because of metastasis in the TLRH group. No conversion of laparotomy occurred in the RRH group. No significant difference was found with respect to intraoperative complications and blood transfusion between both groups. CONCLUSIONS: The results from this study suggest that RRH is superior to TLRH with regard to surgical outcome and may pose a safe and feasible alternative to TLRH. The operating time and lymph node yield is acceptable. Our study is one of the largest single-center studies of surgical outcomes comparing RRH with TLRH during cervical cancer treatment and will significantly contribute to the safety of alternative treatment options for patients. Furthermore, the difference detected between TLRH and RRH group is further strengthened by the great expertise of the surgeon performing laparoscopic surgeries.

Role of surface heat fluxes underneath cold pools.

The role of surface heat fluxes underneath cold pools is investigated using cloud-resolving simulations with either interactive or horizontally homogenous surface heat fluxes over an ocean and a simplified land surface. Over the ocean, there are limited changes in the distribution of the cold pool temperature, humidity, and gust front velocity, yet interactive heat fluxes induce more cold pools, which are smaller, and convection is then less organized. Correspondingly, the updraft mass flux and lateral entrainment are modified. Over the land surface, the heat fluxes underneath cold pools drastically impact the cold pool characteristics with more numerous and smaller pools, which are warmer and more humid and accompanied by smaller gust front velocities. The interactive fluxes also modify the updraft mass flux and reduce convective organization. These results emphasize the importance of interactive surface fluxes instead of prescribed flux boundary conditions, as well as the formulation of surface heat fluxes, when studying convection.

Analysis of the miRNA-mRNA networks in malignant transformation BEAS-2B cells induced by alpha-particles.

The aim of this study was to determine the toxicity induced by irradiation with alpha-particles on malignant transformation of immortalized human bronchial epithelial cells (BEAS-2B) using miRNA-mRNA networks. The expression of BEAS-2B cells was determined by measuring colony formation, mtDNA, mitochondrial membrane potential (MMP), and ROS levels. Changes in BEAS-2B cell gene expression were observed and quantified using microarrays that included an increase in 157 mRNA and 20 miRNA expression and a decrease in 77 mRNA and 48 miRNA. Bioinformatic software was used to analyze these different mRNA and miRNA, which indicated that miR-107 and miR-494 play an important role in alpha-particles-mediated cellular malignant transformation processes. The pathways related to systemic lupus erythematosus, cytokine-cytokine receptor interaction, MAPK signaling pathway, regulation of actin cytoskeleton, and cell adhesion molecules (CAMs) were stimulated, while those of ribosome, transforming growth factor (TGF)-beta signaling pathway, and metabolic pathways were inhibited. Data suggest that miRNA and mRNA play a crucial role in alpha-particles-mediated malignant transformation processes. It is worth noting that three target genes associated with lung cancer were identified and upregulated PEG 10 (paternally expressed gene 10), ARHGAP26, and IRS1.

A smart tailor-made G-clip reporter for sensitive detection of G-triplet-containing sequences.

Taking advantage of the intrinsic characteristics of G-triplet-containing sequences, a pioneering tailor-made clip-like reporter containing three-fourths of a G-quadruplex is established. The reporter can clip the G triplet in the target sequence through a recognition process to form a complete G-quadruplex structure.

Methylene blue as a G-quadruplex binding probe for label-free homogeneous electrochemical biosensing.

Herein, G-quadruplex sequence was found to significantly decrease the diffusion current of methylene blue (MB) in homogeneous solution for the first time. Electrochemical methods combined with circular dichroism spectroscopy and UV-vis spectroscopy were utilized to systematically explore the interaction between MB and an artificial G-quadruplex sequence, EAD2. The interaction of MB and EAD2 (the binding constant, K ≈ 1.3 × 10(6) M(-1)) was stronger than that of MB and double-stranded DNA (dsDNA) (K ≈ 2.2 × 10(5) M(-1)), and the binding stoichiometry (n) of EAD2/MB complex was calculated to be 1.0 according to the electrochemical titration curve combined with Scatchard analysis. MB was proved to stabilize the G-quadruplex structure of EAD2 and showed a competitive binding to G-quadruplex in the presence of hemin. EAD2 might mainly interact with MB, a positive ligand of G-quadruplex, through the end-stacking with π-system of the guanine quartet, which was quite different from the binding mechanism of dsDNA with MB by intercalation. A novel signal read-out mode based on the strong affinity between G-quadruplex and MB coupling with aptamer/G-quadruplex hairpin structure was successfully implemented in cocaine detection with high specificity. G-quadruplex/MB complex will function as a promising electrochemical indicator for constructing homogeneous label-free electrochemical biosensors, especially in the field of simple, rapid, and noninvasive biochemical assays.

A G-quadruplex based platform for label-free monitoring of DNA reaction kinetics.

Research on the kinetic characteristics and mechanisms of DNA reactions is crucial for bioengineering and biosensing. A G-quadruplex, which can form a peroxidase-mimicking DNAzyme with hemin, was for the first time used to establish a versatile platform for kinetic investigations on DNA reactions. G-quadruplex sequence EAD2 was incorporated into the corresponding nucleic acid reaction as product. The kinetic curves can be obtained rapidly and simply via the quantification of created DNAzyme. In this paper, the kinetics of isothermal linear strand displacement amplification reactions with different DNA lengths and isothermal exponential amplification reactions were successfully elucidated via the G-quadruplex based monitoring platform. As a safe and accessible alternative to the traditional methods, this robust, label-free, time-saving and high-throughput platform shows great potential for the exploration of more novel DNA reactions or circuits in an ingenious manner.

Hybridization chain reaction-based fluorescence immunoassay using DNA intercalating dye for signal readout.

A novel format of fluorescence immunosorbent assay based on the hybridization chain reaction (HCR) using a DNA intercalating dye for signal readout was constructed for the sensitive detection of targets, both in competitive and sandwich modes. In this platform, the capture and recognition processes are based on immunoreactions and the signal amplification depends on the enzyme-free, isothermal HCR-induced labelling event. After a competitive or a sandwich immunoreaction, a biotinylated capture DNA was bound to a biotinylated signal antibody through avidin, and triggered the HCR by two specific hairpins into a nicked double helix. Gene Finder (GF), a fluorescent probe for double-strand DNA, was intercalated in situ into the amplified chain to produce the fluorescence signal. The limit of detection (LOD) for rabbit IgG in competitive mode by HCR/GF immunoassay was improved at least 100-fold compared with the traditional fluorescence immunoassay using the fluorescein isothiocyanate-labelled-streptavidin or fluorescein isothiocyanate-labelled second antibody as the signal readout. The proposed fluorescence immunoassay was also demonstrated by using α-fetoprotein as the model target in sandwich mode, and showed a wide linear range from 28 ng mL(-1) to 20 µg mL(-1) with a LOD of 6.0 ng mL(-1). This method also showed satisfactory analysis in spiked human serum, which suggested that it might have great potential for versatile applications in life science and point-of-care diagnostics.

Initial experience with radiomics of carotid perivascular adipose tissue in identifying symptomatic plaque.

Background: Carotid atherosclerotic ischemic stroke threatens human health and life. The aim of this study is to establish a radiomics model of perivascular adipose tissue (PVAT) around carotid plaque for evaluation of the association between Peri-carotid Adipose Tissue structural changes with stroke and transient ischemic attack. Methods: A total of 203 patients underwent head and neck computed tomography angiography examination in our hospital. All patients were divided into a symptomatic group (71 cases) and an asymptomatic group (132 cases) according to whether they had acute/subacute stroke or transient ischemic attack. The radiomic signature (RS) of carotid plaque PVAT was extracted, and the minimum redundancy maximum correlation, recursive feature elimination, and linear discriminant analysis algorithms were used for feature screening and dimensionality reduction. Results: It was found that the RS model achieved the best diagnostic performance in the Bagging Decision Tree algorithm, and the training set (AUC, 0.837; 95%CI: 0.775, 0.899), testing set (AUC, 0.834; 95%CI: 0.685, 0.982). Compared with the traditional feature model, the RS model significantly improved the diagnostic efficacy for identifying symptomatic plaques in the testing set (AUC: 0.834 vs. 0.593; Z = 2.114, p = 0.0345). Conclusion: The RS model of PVAT of carotid plaque can be used as an objective indicator to evaluate the risk of plaque and provide a basis for risk stratification of carotid atherosclerotic disease.

Cisplatin-induced oxPAPC release enhances MDSCs infiltration into LL2 tumour tissues through MCP-1/CCL2 and LTB4/LTB4R pathways.

Lung cancer is the leading global cause of cancer-related death, however, resistance to chemotherapy drugs remains a huge barrier to effective treatment. The elevated recruitment of myeloid derived suppressor cells (MDSCs) to tumour after chemotherapy has been linked to resistance of chemotherapy drugs. Nevertheless, the specific mechanism remains unclear. oxPAPC is a bioactive principal component of minimally modified low-density lipoproteins and regulates inflammatory response. In this work, we found that cisplatin, oxaliplatin and ADM all increased oxPAPC release in tumour. Treating macrophages with oxPAPC in vitro stimulated the secretion of MCP-1 and LTB4, which strongly induced monocytes and neutrophils chemotaxis, respectively. Injection of oxPAPC in vivo significantly upregulated the percentage of MDSCs in tumour microenvironment (TME) of wild-type LL2 tumour-bearing mice, but not CCL2-/- mice and LTB4R-/- mice. Critically, oxPAPC acted as a pro-tumor factor in LL2 tumour model. Indeed, cisplatin increased oxPAPC level in tumour tissues of WT mice, CCL2-/- and LTB4R-/- mice, but caused increased infiltration of Ly6Chigh monocytes and neutrophils only in WT LL2-bearing mice. Collectively, our work demonstrates cisplatin treatment induces an overproduction of oxPAPC and thus recruits MDSCs infiltration to promote the tumour growth through the MCP-1/CCL2 and LTB4/LTB4R pathways, which may restrict the effect of multiple chemotherapy. This provides evidence for a potential strategy to enhance the efficacy of multiple chemotherapeutic drugs in the treatment of lung cancer by targeting oxPAPC.

A Deep Learning Model for Predicting Molecular Subtype of Breast Cancer by Fusing Multiple Sequences of DCE-MRI From Two Institutes.

RATIONALE AND OBJECTIVES: To evaluate the performance of deep learning (DL) in predicting different breast cancer molecular subtypes using DCE-MRI from two institutes. MATERIALS AND METHODS: This retrospective study included 366 breast cancer patients from two institutes, divided into training (n = 292), validation (n = 49) and testing (n = 25) sets. We first transformed the public DCE-MRI appearance to ours to alleviate small-data-size and class-imbalance issues. Second, we developed a multi-branch convolutional-neural-network (MBCNN) to perform molecular subtype prediction. Third, we assessed the MBCNN with different regions of interest (ROIs) and fusion strategies, and compared it to previous DL models. Area under the curve (AUC) and accuracy (ACC) were used to assess different models. Delong-test was used for the comparison of different groups. RESULTS: MBCNN achieved the optimal performance under intermediate fusion and ROI size of 80 pixels with appearance transformation. It outperformed CNN and convolutional long-short-term-memory (CLSTM) in predicting luminal B, HER2-enriched and TN subtypes, but without demonstrating statistical significance except against CNN in TN subtypes, with testing AUCs of 0.8182 vs. [0.7208, 0.7922] (p=0.44, 0.80), 0.8500 vs. [0.7300, 0.8200] (p=0.36, 0.70) and 0.8900 vs. [0.7600, 0.8300] (p=0.03, 0.63), respectively. When predicting luminal A, MBCNN outperformed CNN with AUCs of 0.8571 vs. 0.7619 (p=0.08) without achieving statistical significance, and is comparable to CLSTM. For four-subtype prediction, MBCNN achieved an ACC of 0.64, better than CNN and CLSTM models with ACCs of 0.48 and 0.52, respectively. CONCLUSION: Developed DL model with the feature extraction and fusion of DCE-MRI from two institutes enabled preoperative prediction of breast cancer molecular subtypes with high diagnostic performance.