Automated detection of MRI-negative temporal lobe epilepsy with ROI-based morphometric features and machine learning.

Objective: Temporal lobe epilepsy (TLE) predominantly originates from the anteromedial basal region of the temporal lobe, and its prognosis is generally favorable following surgical intervention. However, TLE often appears negative in magnetic resonance imaging (MRI), making it difficult to quantitatively diagnose the condition solely based on clinical symptoms. There is a pressing need for a quantitative, automated method for detecting TLE. Methods: This study employed MRI scans and clinical data from 51 retrospective epilepsy cases, dividing them into two groups: 34 patients in TLE group and 17 patients in non-TLE group. The criteria for defining the TLE group were successful surgical removal of the epileptogenic zone in the temporal lobe and a favorable postoperative prognosis. A standard procedure was used for normalization, brain extraction, tissue segmentation, regional brain partitioning, and cortical reconstruction of T1 structural MRI images. Morphometric features such as gray matter volume, cortical thickness, and surface area were extracted from a total of 20 temporal lobe regions in both hemispheres. Support vector machine (SVM), extreme learning machine (ELM), and cmcRVFL+ classifiers were employed for model training and validated using 10-fold cross-validation. Results: The results demonstrated that employing ELM classifiers in conjunction with specific temporal lobe gray matter volume features led to a better identification of TLE. The classification accuracy was 92.79%, with an area under the curve (AUC) value of 0.8019. Conclusion: The method proposed in this study can significantly assist in the preoperative identification of TLE patients. By employing this method, TLE can be included in surgical criteria, which could alleviate patient symptoms and improve prognosis, thereby bearing substantial clinical significance.

Systematic Review of Machine Learning Applied to the Secondary Prevention of Ischemic Stroke.

Ischemic stroke is a serious disease posing significant threats to human health and life, with the highest absolute and relative risks of a poor prognosis following the first occurrence, and more than 90% of strokes are attributable to modifiable risk factors. Currently, machine learning (ML) is widely used for the prediction of ischemic stroke outcomes. By identifying risk factors, predicting the risk of poor prognosis and thus developing personalized treatment plans, it effectively reduces the probability of poor prognosis, leading to more effective secondary prevention. This review includes 41 studies since 2018 that used ML algorithms to build prognostic prediction models for ischemic stroke, transient ischemic attack (TIA), and acute ischemic stroke (AIS). We analyzed in detail the risk factors used in these studies, the sources and processing methods of the required data, the model building and validation, and their application in different prediction time windows. The results indicate that among the included studies, the top five risk factors in terms of frequency were cardiovascular diseases, age, sex, national institutes of health stroke scale (NIHSS) score, and diabetes. Furthermore, 64% of the studies used single-center data, 65% of studies using imbalanced data did not perform data balancing, 88% of the studies did not utilize external validation datasets for model validation, and 72% of the studies did not provide explanations for their models. Addressing these issues is crucial for enhancing the credibility and effectiveness of the research, consequently improving the development and implementation of secondary prevention measures.

Construction of a scalable DNA computing nano-system for large-scale and complex logical operations.

The predictability of Watson-Crick base pairing provides unique structural programmability to DNA, facilitating the development and application of biomolecules in biocomputing. However, in DNA-based biocomputing, the scale of operation that can be achieved by an existing reaction system is very limited. How to expand the operation range of a logic circuit and realize the integration and extensibility of circuits is always the key problem to be solved in this field. In this work, by designing a multifunctional DNA-nanostructure-based reaction platform, which can realize an output of up to 2n scalable fluorescence signals, combined with the construction of an input "library" and a modular distribution strategy of output signals, for the first time, we successfully performed the calculation of both square roots and cube roots of consecutive integers within a decimal number of "10" and in each result of the operation, two digits after the decimal point are preserved (). We believe that the design concept presented in this work can help effectively solve the urgent problems of biological computing in terms of computational scaling, integration and scalability, and can open up new horizons for the design of new functional devices and complex computing circuits.

Detection of mild cognitive impairment in type 2 diabetes mellitus based on machine learning using privileged information.

Type 2 diabetes mellitus (T2DM) patients may develop into mild cognitive impairment (MCI) or even dementia. However, there is lack of reliable machine learning model for detection MCI in T2DM patients based on machine learning method. In addition, the brain network changes associated with MCI have not been studied. The aim of this study is to develop a machine learning based algorithm to help detect MCI in T2DM. There are 164 participants were included in this study. They were divided into T2DM-MCI (n = 56), T2DM-nonMCI (n = 49), and normal controls (n = 59) according to the neuropsychological evaluation. Functional connectivity of each participant was constructed based on resting-state magnetic resonance imaging (rs-fMRI). Feature selection was used to reduce the feature dimension. Then the selected features were set into the cascaded multi-column random vector functional link network (RVFL) classifier model using privileged information. Finally, the optimal model was trained and the classification performance was obtained using the testing data. The results show that the proposed algorithm has outstanding performance compared with classic methods. The classification accuracy of 73.18 % (T2DM-MCI vs NC) and 79.42 % (T2DM-MCI vs T2DM-nonMCI) were achieved. The functional connectivity related to T2DM-MCI mainly distribute in the frontal lobe, temporal lobe, and central region (motor cortex), which could be used as neuroimaging biomarkers to recognize MCI in T2DM patients. This study provides a machine learning model for diagnosis of MCI in T2DM patients and has potential clinical significance for timely intervention and treatment to delay the development of MCI.

Correlation between parameters related to sarcopenia and gray matter volume in patients with mild to moderate Alzheimer's disease.

BACKGROUND: Alzheimer's disease (AD) is a neurodegenerative disease characterized by brain atrophy and closely correlated with sarcopenia. Mounting studies indicate that parameters related to sarcopenia are associated with AD, but some results show inconsistent. Furthermore, the association between the parameters related to sarcopenia and gray matter volume (GMV) has rarely been explored. AIM: To investigate the correlation between parameters related to sarcopenia and cerebral GMV in AD. METHODS: Demographics, neuropsychological tests, parameters related to sarcopenia, and magnetic resonance imaging (MRI) scans were collected from 42 patients with AD and 40 normal controls (NC). Parameters related to sarcopenia include appendicular skeletal muscle mass index (ASMI), grip strength, 5-times sit-to-stand (5-STS) time and 6-m gait speed. The GMV of each cerebral region of interest (ROI) and the intracranial volume were calculated by computing the numbers of the voxels in the specific region based on MRI data. Partial correlation and multivariate stepwise linear regression analysis explored the correlation between different inter-group GMV ratios in ROIs and parameters related to sarcopenia, adjusting for covariates. RESULTS: The 82 participants included 40 NC aged 70.13 ± 5.94 years, 24 mild AD patients aged 73.54 ± 8.27 years and 18 moderate AD patients aged 71.67 ± 9.39 years. Multivariate stepwise linear regression showed that 5-STS time and gait speed were correlated with bilateral hippocampus volume ratios in total AD. Grip strength was associated with the GMV ratio of the left middle frontal gyrus in mild AD and the GMV ratios of the right superior temporal gyrus and right hippocampus in moderate AD. However, ASMI did not have a relationship to any cerebral GMV ratio. CONCLUSIONS: Among parameters related to sarcopenia, 5-STS time and gait speed were associated with bilateral hippocampus volume ratios at different clinical stages of patients with AD. Five-STS time provide an objective basis for early screening and can help diagnose patients with AD.

Improving MRI-based analysis of brain structural changes in patients with hypertension via a privileged information learning algorithm.

Hypertension can lead to changes in the brain structure and function, and different blood pressure levels (2017ACC/AHA) have different effects on brain structure. It is important to analyze these changes by machine learning methods, and various characteristics can provide rich information for the analysis of these changes. However, multiple feature extraction involves complex data processing. How to make a single feature achieve the same diagnosis effect as multiple features do is worth of study. Kernel ridge regression (KRR) is a kind of machine learning method, which shows faster learning speed and generalization ability in classification tasks. In order to knowledge transfer, we use privileged information (PI) to transfer information of multiple types of feature to single feature. This allows only one feature type to be used during the test stage. In the process of feature fusion, we need to consider all the samples' attribution making the classifier better. In this work, we propose a multi-kernel KRR+ framework based on self-paced learning to analyze the changes of the brain structure in patients with different blood pressure levels. Specifically, one kind of a feature is taken as main feature, and other features are input into the multi-kernel KRR as PI. These two inputs are fed into the final KRR classifier together. In addition, a self-paced learning method is introduced into sample selecting to avoid training the classifier using samples with a large loss value firstly, which improves the generalization performance of the classifier. Experimental results show that the proposed method can make full use of the information of various features and achieve better classification performance. This shows self-paced learning based KRR can help analyze brain structure of patients with different blood pressure levels. The discriminative features may help clinicians to make judgments of hypertension degrees on brain MRI images.

Self-paced learning and privileged information based KRR classification algorithm for diagnosis of Parkinson's disease.

Computer aided diagnosis (CAD) methods for Parkinson's disease (PD) can assist clinicians in diagnosis and treatment. Magnetic resonance imaging (MRI) based CAD methods can help reveal structural changes in brain. Classifier is a key component in CAD system, which directly affects the classification performance. Privileged information (PI) can assist to train the classifier by providing additional information, which makes test samples have less error and improves the classification accuracy. In this paper, we proposed a PI based kernel ridge regression plus (KRR+) in diagnosis of PD. Specifically, morphometric features and brain network features are extracted from MRI. Then, empirical kernel mapping feature expression method is used to make the data separable in high-dimensional space. Besides, we introduce self-paced learning that can adaptively select the sample in training of the model, which can further improve the classification performance. The experimental results show that the proposed method is effective for PD diagnosis, its performance is superior to existing classification model. This method is helpful to assist clinicians to find out possible neuroimaging biomarkers in the diagnosis of PD.

Self-Paced Learning and Privileged Information based Cascaded Multi-column Classification algorithm for ASD diagnosis.

Autism spectrum disorder (ASD) is one of the most serious mental disorder in children. Machine learning based computer aided diagnosis (CAD) on resting-state functional magnetic resonance imaging (rs-fMRI) for ASD has attracted widespread attention. In recent years, learning using privileged information (LUPI), a supervised transfer learning method, has been generally used on multi-modality cases, which can transfer knowledge from source domain to target domain in order to improve the prediction capability on the target domain. However, multi-modality data is difficult to collect in clinical cases. LUPI method without introducing additional imaging modality images is worth further study. Random vector function link network plus (RVFL+) is a LUPI diagnosis algorithm, which has been proven to be effective for classification tasks. In this work, we proposed a self-paced learning based cascaded multi-column RVFL+ algorithm (SPL-cmcRVFL+) for ASD diagnosis. Initial classification model is trained using RVFL on the single-modal data (e.g. rs-fMRI). The output of the initial layer is then sent as privileged information (PI) to train the next layer of classification model. During this process, samples are selected using self-paced learning (SPL), which can adaptively select simple to difficult samples according to the loss value. The procedure is repeated until all samples are included. Experimental results show that our proposed method can accurately identify ASD and normal control, and outperforms other methods by a relatively higher classification accuracy.

Theoretical insight on electronic structure and photophysical properties of a series of cyclometalated iridium(III) complexes bearing the substituted phenylpyrazole with different electron-donating or electron-accepting groups.

By using the density functional theory (DFT) and time-dependent density functional theory (TDDFT), the electronic structure and photophysical properties of a series of cyclometalated iridium(III) complexes bearing the substituted phenylpyrazole have been theoretically investigated. All studied iridium(III) complexes have the distorted octahedral geometry with cis-C,C, cis-O,O, and trans-N,N chelate disposition. The lowest lying singlet → singlet absorptions of all studied iridium(III) complexes are respectively located at 405 nm, 387 nm, 382 nm, 370 nm, and 387 nm. The calculated emission wavelengths for all studied iridium(III) complexes are 654 nm, 513 nm, 506 nm, 505 nm and 499 nm, respectively. The calculated emission wavelength for complex 4 at the CAM-B3LYP level is in good agreement with the experimental value. From the theoretical results, it can be seen that the electron-donating substituent groups have the important effect on the electronic structure and photophysical properties of all studied iridium(III) complexes. We hope that this study can provide valuable guidance for the design of new phosphorescent organic light-emitting diodes (OLEDs) materials.

Generation of Transgenic Cloned Buffalo Embryos Harboring the EGFP Gene in the Y Chromosome Using CRISPR/Cas9-Mediated Targeted Integration.

Sex control technology is of great significance in the production of domestic animals, especially for rapidly breeding water buffalo (bubalus bubalis), which served as a research model in the present study. We have confirmed that a fluorescence protein integrated into the Y chromosome is fit for sexing pre-implantation embryos in the mouse. Firstly, we optimized the efficiency of targeted integration of exogenous gene encoding enhanced green fluorescent protein (eGFP) and mCherry in Neuro-2a cells, mouse embryonic stem cells, mouse embryonic cells (NIH3T3), buffalo fetal fibroblast (BFF) cells. The results showed that a homology arm length of 800 bp on both sides of the target is more efficient that 300 bp or 300 bp/800 bp. Homology-directed repair (HDR)-mediated knock-in in BFF cells was also significantly improved when cells were supplemented with pifithrin-µ, which is a small molecule that inhibits the binding of p53 to mitochondria. Three pulses at 250 V resulted in the most efficient electroporation in BFF cells and 1.5 µg/mL puromycin was found to be the optimal concentration for screening. Moreover, Y-Chr-eGFP transgenic BFF cells and cloned buffalo embryos were successfully generated using CRISPR/Cas9-mediated gene editing combined with the somatic cell nuclear transfer (SCNT) technique. At passage numbers 6-8, the growth rate and cell proliferation rate were significantly lower in Y-Chr-eGFP transgenic than in non-transgenic BFF cells; the expression levels of the methylation-related genes DNMT1 and DNMT3a were similar; however, the expression levels of the acetylation-related genes HDAC1, HDAC2, and HDAC3 were significantly higher (p < 0.05) in Y-Chr-eGFP transgenic BFF cells compared with non-transgenic cells. Y-Chr-eGFP transgenic BFFs were used as donors for SCNT, the results showed that eGFP reporter is suitable for the visualization of the sex of embryos. The blastocyst rates of cloned buffalo embryos were similar; however, the cleavage rates of transgenic cloned embryos were significantly lower compared with control. In summary, we optimized the protocol for generating transgenic BFF cells and successfully generated Y-Chr-eGFP transgenic embryos using these cells as donors.

Maternal transcription profiles at different stages for the development of early embryo in buffalo.

Maternal mRNAs deposited in the egg during oogenesis are critical during the development of early embryo, before the activation of the embryonic genome. However, there is little known about the dynamic expression of maternally expressed genes in mammals. In this study, we made buffalo parthenogenesis as our research model to analyse maternal transcription profiles of pre-implantation embryo in buffalo using RNA sequencing. In total, 3,567 unique genes were detected to be differentially expressed among all constant stages during early embryo development (FPKM > 0). Interestingly, a total of 10,442 new genes were discovered in this study, and gene ontology analysis of the new differentially expressed genes indicated that the new genes have a wide cellular localization and are involved in many molecular functions and biological processes. Moreover, we identified eight clusters that were only highly expressed in a particular developmental stage and enriched a number of GO terms and KEGG pathways that were related to specific stage. Furthermore, we identified 1,530 hub genes (or key members) from the maternally expressed gene networks, and these hub genes were involved in 11 stage-specific modules. After visualization using Cytoscape 3.2.1 software, we obtained complex interaction network of hub genes, indicating the highly efficient cooperation between genes during the development in buffalo embryos. Further research of these genes will greatly deepen our understanding of embryo development in buffalo. Collectively, this research lays the foundation for future studies on the maternal genome function, buffalo nuclear transfer and parthenogenetic embryonic stem cells.

Global transcriptome analysis of different stages of preimplantation embryo development in river buffalo.

BACKGROUND: Water buffalo (Bubalus bubalis) are divided into river buffalo and swamp buffalo subspecies and are essential livestock for agriculture and the local economy. Studies on buffalo reproduction have primarily focused on optimal fertility and embryonic mortality. There is currently limited knowledge on buffalo embryonic development, especially during the preimplantation period. Assembly of the river buffalo genome offers a reference for omics studies and facilitates transcriptomic analysis of preimplantation embryo development (PED). METHODS: We revealed transcriptomic profile of four stages (2-cell, 8-cell, Morula and Blastocyst) of PED via RNA-seq (Illumina HiSeq4000). Each stage comprised three biological replicates. The data were analyzed according to the basic RNA-seq analysis process. Ingenuity analysis of cell lineage control, especially transcription factor (TF) regulatory networks, was also performed. RESULTS: A total of 21,519 expressed genes and 67,298 transcripts were predicted from approximately 81.94 Gb of raw data. Analysis of transcriptome-wide expression, gene coexpression networks, and differentially expressed genes (DEGs) allowed for the characterization of gene-specific expression levels and relationships for each stage. The expression patterns of TFs, such as POU5F1, TEAD4, CDX4 and GATAs, were elucidated across diverse time series; most TF expression levels were increased during the blastocyst stage, during which time cell differentiation is initiated. All of these TFs were involved in the composition of the regulatory networks that precisely specify cell fate. These findings offer a deeper understanding of PED at the transcriptional level in the river buffalo.

Integrative Analysis of Transcriptome and GWAS Data to Identify the Hub Genes Associated With Milk Yield Trait in Buffalo.

The mammary gland is the production organ in mammals that is of great importance for milk production and quality. However, characterization of the buffalo mammary gland transcriptome and identification of the valuable candidate genes that affect milk production is limited. Here, we performed the differential expressed genes (DEGs) analysis of mammary gland tissue on day 7, 50, 140, and 280 after calving and conducted gene-based genome-wide association studies (GWAS) of milk yield in 935 Mediterranean buffaloes. We then employed weighted gene co-expression network analysis (WGCNA) to identify specific modules and hub genes related to milk yield based on gene expression profiles and GWAS data. The results of the DEGs analysis showed that a total of 1,420 DEGs were detected across different lactation points. In the gene-based analysis, 976 genes were found to have genome-wide association (P ≤ 0.05) that could be defined as the nominally significant GWAS geneset (NSGG), 9 of which were suggestively associated with milk yield (P < 10-4). Using the WGCNA analysis, 544 and 225 genes associated with milk yield in the turquoise module were identified from DEGs and NSGG datasets, respectively. Several genes (including BNIPL, TUBA1C, C2CD4B, DCP1B, MAP3K5, PDCD11, SRGAP1, GDPD5, BARX2, SCARA3, CTU2, and RPL27A) were identified and considered as the hub genes because they were involved in multiple pathways related to milk production. Our findings provide an insight into the dynamic characterization of the buffalo mammary gland transcriptome, and these potential candidate genes may be valuable for future functional characterization of the buffalo mammary gland.

Molecular characterisation of the buffalo SCAP gene and its association with milk production traits in water buffaloes.

The study reported in this Research Communication was conducted to investigate the molecular characterisation of buffalo SCAP gene, expression analysis, and the association between single nucleotide polymorphisms and milk production traits in 384 buffaloes. Sequence analysis revealed the SCAP gene had an open reading frame of 3837 bp encoding 1279 amino acids. A ubiquitous expression profile of SCAP gene was detected in various tissues with extreme predominance in the mammary gland during early lactation. Moreover, eleven SNPs in buffalo SCAP gene were identified, six of them (g.1717600A>G, g.1757922C>T, g.1758953G>A, g.1759142C>T, g.1760740G>A, and g.1766036T>C) were found to be significantly associated with 305-day milk yield. Thus, buffalo SCAP could sever as a candidate gene affecting milk production traits in buffalo and the identified SNPs might potentially be genetic markers.

Four novel polymorphisms of buffalo INSIG2 gene are associated with milk production traits in Chinese buffaloes.

Insulin-induced genes (INSIGs), including INSIG1 and INSIG2, are important mediators that play a pivotal role in the lipid metabolism and could cause the retention of the SCAP/SREBP complex. Therefore, the objective of this study is to detect the single nucleotide polymorphisms (SNPs) of buffalo INSIG2 gene and evaluate their associations with milk production traits in Chinese buffaloes. A total of four SNPs (g.621272A > G, g.621364A > C, g.632543G > A, and g.632684C > T) were identified using DNA pooled sequencing, and the SNP genotyping for the identified SNPs was performed by using Matrix-assisted laser desorption/ionization time of flight mass spectrometry method from 264 individuals. The results showed that four SNPs were significantly associated with 305-day milk yield or protein percentage in Murrah and crossbred breeds (P < 0.05), but they had no significant effect on milk production traits in Nili-Ravi buffaloes (P > 0.05). Linkage disequilibrium (LD) analysis revealed that one haplotype block was successfully constructed, of which the diplotype H1H1 showed significant association with 305-day milk yield in Murrah buffaloes (P < 0.05). Our findings provide evidence that polymorphisms in buffalo INSIG2 gene are associated with milk production traits, and could be used as a candidate gene for marker-assisted selection in buffalo breeding program.

De Novo Transcriptome Assembly of the Chinese Swamp Buffalo by RNA Sequencing and SSR Marker Discovery.

The Chinese swamp buffalo (Bubalis bubalis) is vital to the lives of small farmers and has tremendous economic importance. However, a lack of genomic information has hampered research on augmenting marker assisted breeding programs in this species. Thus, a high-throughput transcriptomic sequencing of B. bubalis was conducted to generate transcriptomic sequence dataset for gene discovery and molecular marker development. Illumina paired-end sequencing generated a total of 54,109,173 raw reads. After trimming, de novo assembly was performed, which yielded 86,017 unigenes, with an average length of 972.41 bp, an N50 of 1,505 bp, and an average GC content of 49.92%. A total of 62,337 unigenes were successfully annotated. Among the annotated unigenes, 27,025 (43.35%) and 23,232 (37.27%) unigenes showed significant similarity to known proteins in NCBI non-redundant protein and Swiss-Prot databases (E-value < 1.0E-5), respectively. Of these annotated unigenes, 14,439 and 15,813 unigenes were assigned to the Gene Ontology (GO) categories and EuKaryotic Ortholog Group (KOG) cluster, respectively. In addition, a total of 14,167 unigenes were assigned to 331 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Furthermore, 17,401 simple sequence repeats (SSRs) were identified as potential molecular markers. One hundred and fifteen primer pairs were randomly selected for amplification to detect polymorphisms. The results revealed that 110 primer pairs (95.65%) yielded PCR amplicons and 69 primer pairs (60.00%) presented polymorphisms in 35 individual buffaloes. A phylogenetic analysis showed that the five swamp buffalo populations were clustered together, whereas two river buffalo breeds clustered separately. In the present study, the Illumina RNA-seq technology was utilized to perform transcriptome analysis and SSR marker discovery in the swamp buffalo without using a reference genome. Our findings will enrich the current SSR markers resources and help spearhead molecular genetic research studies on the swamp buffalo.

Molecular cloning and expression analysis of the STAT1 gene in the water buffalo (Bubalus bubalis).

Signal transducer and activator of transcription 1 (STAT1) is a critical component of the transcription factor complex in the interferon (IFN) signaling pathways. Of the seven STAT isoforms, STAT1 is a key mediator of type I and type III IFN signaling, but limited information is available for the STAT genes in the water buffalo. Here, we amplified and identified the complete coding sequence (CDS) of the buffalo STAT1 gene by using reverse transcription polymerase chain reaction (RT-PCR). Sequence analysis indicated that the buffalo STAT1 gene length size was 3437 bp, containing an open reading frame (ORF) of 2244 bp that encoded 747 amino acids for the first time. The buffalo STAT1 CDS showed 99, 98, 89, 93, 86, 85, and 87% identity with that of Bos taurus, Ovis aries, Homo sapiens, Sus scrofa, Rattus norvegicus, Mus musculus, and Capra hircus. The phylogenetic analyses revealed that the nearest relationship existed between the water buffalo and B. taurus. The STAT1 gene was ubiquitously expressed in 11 buffalo tissues by real-time PCR, whereas STAT1 was expressed at higher levels in the lymph. The STAT1 gene contained five targeted microRNA sequences compared with the B. taurus by the miRBase software that provide a fundamental for identifying the STAT1 gene function.

[Design and implementation of pulse instrument based on DSP].

The Traditional Chinese Medical Pulse Instrument uses the HKG-07B infrared pulse sensor to get pulse signal from the body. It makes full use of the TMS320VC5402 chip to realize time-frequency domain parameters extracting, classification and identification of the pulse signal. The system can store a plenty of pulse signal and realize data communication with the PC via the USB interface. According to acquisition and classification of pulse signal experiments of 200 subjects, the results show that the recognition rate of pulse signal can reach to 87.4%. It is applicable to the clinical diagnosis and detection of the pulse signal and home healthcare.

Evaluation of fasting metabolism of growing water buffalo (Bubalus, Bubalis).

The objectives of the present study were to evaluate fasting metabolism (FM) of water buffalo (Bubalus, Bubalis) at three stages of growth (12, 18 and 24 months) in Guangxi, China. Five female water buffalo were used for each age group and their live weight was on average 254, 326 and 338 kg, respectively. All animals were of average body condition, healthy and de-wormed before start of the study. Prior to a 6-day fasting period, buffalo were offered a mixed diet of forage and concentrates (70% and 30%, dry matter basis) on a restricted nutritional level (419 kJ/kg(0.75) of metabolizable energy, ME) for 15 days. Gas exchanges for each animal were determined for 3 days from day 4 of starvation, using open-circuit respiratory head hoods. Fasting body weight was 0.918 of live weight (P < 0.001, r(2) = 0.99). Both fasting heat production (FHP) and FM (MJ/day) increased significantly with increased age of animals (P < 0.05). Linear regression analysis indicated a positive relationship between fasting body weight (kg(0.75)) and FHP (MJ/day, P < 0.01, r(2) = 0.49) or FM (MJ/day P < 0.01, r(2) = 0.52) when using individual animal data across three groups. However, when expressed as kJ/kg(0.75) of fasting body weight, the differences in FHP or FM between three groups of animals were not significant. The present average FHP and FM (322 and 347 kJ/kg(0.75) of fasting body weight) were compatible to those published in the literature for water buffalo, beef and dairy cattle. The present FM data were also used to estimate net energy (NE(m)) and ME (ME(m)) requirements for maintenance for water buffalo. The results for these two parameters were similar to those for FHP and FM. There was no significant difference between three groups of buffalo in NE(m) or ME(m) when expressed as kJ/kg(0.75) of live weight. The present average NE(m) and ME(m) values (347 and 506 kJ/kg(0.75) of live weight) are close to those proposed by the Agricultural and Food Research Council adopted in UK for beef and dairy cattle. The results indicate that the present FM data can be used as a basis for rationing water buffalo in China.