A Knowledge-Graph-Based Multimodal Deep Learning Framework for Identifying Drug-Drug Interactions.

The identification of drug-drug interactions (DDIs) plays a crucial role in various areas of drug development. In this study, a deep learning framework (KGCN_NFM) is presented to recognize DDIs using coupling knowledge graph convolutional networks (KGCNs) with neural factorization machines (NFMs). A KGCN is used to learn the embedding representation containing high-order structural information and semantic information in the knowledge graph (KG). The embedding and the Morgan molecular fingerprint of drugs are then used as input of NFMs to predict DDIs. The performance and effectiveness of the current method have been evaluated and confirmed based on the two real-world datasets with different sizes, and the results demonstrate that KGCN_NFM outperforms the state-of-the-art algorithms. Moreover, the identified interactions between topotecan and dantron by KGCN_NFM were validated through MTT assays, apoptosis experiments, cell cycle analysis, and molecular docking. Our study shows that the combination therapy of the two drugs exerts a synergistic anticancer effect, which provides an effective treatment strategy against lung carcinoma. These results reveal that KGCN_NFM is a valuable tool for integrating heterogeneous information to identify potential DDIs.

Identification of MiRNA-Disease Associations Based on Information of Multi-Module and Meta-Path.

Cumulative research reveals that microRNAs (miRNAs) are involved in many critical biological processes including cell proliferation, differentiation and apoptosis. It is of great significance to figure out the associations between miRNAs and human diseases that are the basis for finding biomarkers for diagnosis and targets for treatment. To overcome the time-consuming and labor-intensive problems faced by traditional experiments, a computational method was developed to identify potential associations between miRNAs and diseases based on the graph attention network (GAT) with different meta-path mode and support vector (SVM). Firstly, we constructed a multi-module heterogeneous network based on the meta-path and learned the latent features of different modules by GAT. Secondly, we found the average of the latent features with weight to obtain a final node representation. Finally, we characterized miRNA-disease-association pairs with the node representation and trained an SVM to recognize potential associations. Based on the five-fold cross-validation and benchmark datasets, the proposed method achieved an area under the precision-recall curve (AUPR) of 0.9379 and an area under the receiver-operating characteristic curve (AUC) of 0.9472. The results demonstrate that our method has an outstanding practical application performance and can provide a reference for the discovery of new biomarkers and therapeutic targets.

Identification of Potential Parkinson's Disease Drugs Based on Multi-Source Data Fusion and Convolutional Neural Network.

Parkinson's disease (PD) is a serious neurodegenerative disease. Most of the current treatment can only alleviate symptoms, but not stop the progress of the disease. Therefore, it is crucial to find medicines to completely cure PD. Finding new indications of existing drugs through drug repositioning can not only reduce risk and cost, but also improve research and development efficiently. A drug repurposing method was proposed to identify potential Parkinson's disease-related drugs based on multi-source data integration and convolutional neural network. Multi-source data were used to construct similarity networks, and topology information were utilized to characterize drugs and PD-associated proteins. Then, diffusion component analysis method was employed to reduce the feature dimension. Finally, a convolutional neural network model was constructed to identify potential associations between existing drugs and LProts (PD-associated proteins). Based on 10-fold cross-validation, the developed method achieved an accuracy of 91.57%, specificity of 87.24%, sensitivity of 95.27%, Matthews correlation coefficient of 0.8304, area under the receiver operating characteristic curve of 0.9731 and area under the precision-recall curve of 0.9727, respectively. Compared with the state-of-the-art approaches, the current method demonstrates superiority in some aspects, such as sensitivity, accuracy, robustness, etc. In addition, some of the predicted potential PD therapeutics through molecular docking further proved that they can exert their efficacy by acting on the known targets of PD, and may be potential PD therapeutic drugs for further experimental research. It is anticipated that the current method may be considered as a powerful tool for drug repurposing and pathological mechanism studies.

Performance improvement for a 2D convolutional neural network by using SSC encoding on protein-protein interaction tasks.

BACKGROUND: The interactions of proteins are determined by their sequences and affect the regulation of the cell cycle, signal transduction and metabolism, which is of extraordinary significance to modern proteomics research. Despite advances in experimental technology, it is still expensive, laborious, and time-consuming to determine protein-protein interactions (PPIs), and there is a strong demand for effective bioinformatics approaches to identify potential PPIs. Considering the large amount of PPI data, a high-performance processor can be utilized to enhance the capability of the deep learning method and directly predict protein sequences. RESULTS: We propose the Sequence-Statistics-Content protein sequence encoding format (SSC) based on information extraction from the original sequence for further performance improvement of the convolutional neural network. The original protein sequences are encoded in the three-channel format by introducing statistical information (the second channel) and bigram encoding information (the third channel), which can increase the unique sequence features to enhance the performance of the deep learning model. On predicting protein-protein interaction tasks, the results using the 2D convolutional neural network (2D CNN) with the SSC encoding method are better than those of the 1D CNN with one hot encoding. The independent validation of new interactions from the HIPPIE database (version 2.1 published on July 18, 2017) and the validation of directly predicted results by applying a molecular docking tool indicate the effectiveness of the proposed protein encoding improvement in the CNN model. CONCLUSION: The proposed protein sequence encoding method is efficient at improving the capability of the CNN model on protein sequence-related tasks and may also be effective at enhancing the capability of other machine learning or deep learning methods. Prediction accuracy and molecular docking validation showed considerable improvement compared to the existing hot encoding method, indicating that the SSC encoding method may be useful for analyzing protein sequence-related tasks. The source code of the proposed methods is freely available for academic research at https://github.com/wangy496/SSC-format/ .

Novel Potential Small Molecule-MiRNA-Cancer Associations Prediction Model Based on Fingerprint, Sequence, and Clinical Symptoms.

As an important biomarker in organisms, miRNA is closely related to various small molecules and diseases. Research on small molecule-miRNA-cancer associations is helpful for the development of cancer treatment drugs and the discovery of pathogenesis. It is very urgent to develop theoretical methods for identifying potential small molecular-miRNA-cancer associations, because experimental approaches are usually time-consuming, laborious, and expensive. To overcome this problem, we developed a new computational method, in which features derived from structure, sequence, and symptoms were utilized to characterize small molecule, miRNA, and cancer, respectively. A feature vector was construct to characterize small molecule-miRNA-cancer association by concatenating these features, and a random forest algorithm was utilized to construct a model for recognizing potential association. Based on the 5-fold cross-validation and benchmark data set, the model achieved an accuracy of 93.20 ± 0.52%, a precision of 93.22 ± 0.51%, a recall of 93.20 ± 0.53%, and an F1-measure of 93.20 ± 0.52%. The areas under the receiver operating characteristic curve and precision recall curve were 0.9873 and 0.9870. The real prediction ability and application performance of the developed method have also been further evaluated and verified through an independent data set test and case study. Some potential small molecules and miRNAs related to cancer have been identified and are worthy of further experimental research. It is anticipated that our model could be regarded as a useful high-throughput virtual screening tool for drug research and development. All source codes can be downloaded from https://github.com/LeeKamlong/Multi-class-SMMCA.

Continuous-flow microwave-assisted extraction coupled with online single drop microextraction prior to GC-MS for determination of amide herbicides in rice samples.

A rapid method based on continuous-flow microwave-assisted extraction and online single drop microextraction was first developed and applied to the determination of amide herbicides in rice. The present method has the advantages of both continuous-flow microwave-assisted extraction and online single drop microextraction, which combines extraction, separation, preconcentration, and sample introduction in one step. By continuous-flow microwave-assisted extraction, analytes were first extracted from the rice samples using 15% methanol-water, and then concentrated into single drop. The microdrop was retracted into microsyringe and directly analyzed by gas chromatography with mass spectrometry without any filtration or clean-up process. The method greatly simplifies the sample treatment procedure, reduces consumption of toxic organic solvent, and extends the application of single drop microextraction to complex solid samples. Several parameters were optimized by Box-Behnken design. Under optimal experimental conditions, good linearity was observed in the range of 2.0-500.0 µg/kg. The limits of detection and quantification were in the range of 0.3-1.5 and 1.1-5.1 µg/kg, respectively. The intra- and inter-day precisions were between 1.9 and 4.8%. The present method was used to the analysis of real rice samples, and the recoveries of analytes were between 80.3 and 102.3% with the relative standard deviations ranging from 1.1 to 6.9%.

Microwave absorption medium-assisted extraction coupled with reversed-phase dispersive liquid-liquid microextraction of triazine herbicides in corn and soybean samples.

A new extraction method, microwave absorption medium-assisted extraction coupled with reversed-phase dispersive liquid-liquid microextraction, was developed for the determination of triazine herbicides in corn and soybean samples. Triazine herbicides were extracted with hexane and then directly enriched into the ionic liquid phase. The purification of sample and concentration of target analytes were performed simultaneously. The method combines the advantages of nonpolar solvent dynamic microwave extraction and reversed-phase dispersive liquid-liquid microextraction, which could greatly simplify the operation and reduce the whole pretreatment time. The Box-Behnken design was used to the optimization of experimental factors involved in the dynamic microwave-assisted extraction. In the present study, good linearity in the range of 5.00-500.00 µg/kg was obtained. The limits of detection and quantification varying from 1.3 to 4.2 and 4.1 to 13.9 µg/kg were achieved, respectively. The intra- and interday precisions were between 2.7 and 6.9%. The present method was applied to the analysis of corn and soybean samples, and the recoveries of analytes ranged from 80.7 to 106.9% with the relative standard deviations of 2.1-7.8%. The present method shows the potentials of practical applications in the treatment of the complex fatty solid samples.

Structural Elucidation and Total Synthesis of Three 9-Norlignans from Curculigo capitulata.

Capitulactones A-C, three unprecedented 9-norlignans featuring a unique 3,5-dihydrofuro[2,3- d]oxepin-7(2 H)-one scaffold, were isolated from the roots of Curculigo capitulata. Their structures with absolute configurations were unambiguously established by a combination of spectroscopic data, ECD analysis, and total synthesis. Biomimetic total syntheses of three pairs of the corresponding enantiomers were achieved in 9-10 steps with overall yields of 14.8, 12.7, and 10.3%, respectively. Notably, the unique scaffold of the common western hemisphere of the molecules was constructed by using the oxidation-reduction strategy from benzodihydrofuran.

Disinfection and removal performance for Escherichia coli, toxic heavy metals and arsenic by wood vinegar-modified zeolite.

Zeolite has traditionally been used to remediate wastewater and soil. The present study shows a new method for natural zeolite (NZ) modification with wood vinegar (WV). The optimal conditions for NZ modification with WV were determined, and the adsorption capacities towards lead (Pb), cadmium (Cd) and arsenic (As), antimicrobial activities against Escherichia coli, heavy metal(loid) fraction and characterizations of selected modified zeolites (MZs) were also investigated. The results indicate that 50-fold dilution of WV, 5 g of NZ dosage, 105 °C of drying temperature, 4 h and 95 °C of water bath are preferred for NZ modification with WV. The WV+NaOH-MZ exhibited the best performance in heavy metal removal and the most powerful antimicrobial activity among all the zeolites. The sequence of WV+NaOH-MZ for the maximum single metal(loid) adsorption capacities was Pb (48.67 mg/g) >Cd (23.67 mg/g) > As (0.024 mg/g). The WV+NaOH and WV modifications also can increase the stabilities of heavy metals in the MZs. The residual fractions of single Pb and Cd in WV+NaOH-MZ and WV-MZ were 50%, 55%, 34% and 30%, respectively. The pore size of WV+NaOH-MZ (11.73 nm) was bigger than that of NZ or WV-MZ. Additionally, the proportion of clinoptilolite in WV+NaOH-MZ was also higher than other zeolites. The surfaces of WV+alkali-MZs were rougher than that of NZ. Considering the low cost and environmental risk of WV, this work provides some useful information for management of agricultural and industrial residues, environment and food safety.

Deficit irrigation effectively reduces soil carbon dioxide emissions from wheat fields in Northwest China.

BACKGROUND: An irrigation regime is an important factor in regulating soil CO2 emissions from wheat fields. Deficit irrigation can be applied easily in the fields and has been implemented in northwest China. Previous studies have mainly focused on the effects of deficit irrigation on crop yield and quality. Studies on its environmental impacts are sparse. RESULTS: Soil CO2 fluxes from deficit-irrigated fields were lower than those from full irrigation (CK) during most of the growing season. Cumulative soil CO2 emissions from deficit-irrigated fields were reduced by 10.2-25.5%, compared with the CK. Peaks of soil CO2 fluxes were observed 3-7 days after irrigation in the water-filled pore space (WFPS) range of 65.7-80.4%. Under different irrigation regimes, significant positive correlations were observed between soil CO2 fluxes and WFPS (P < 0.01), but no significant correlations were found between soil CO2 fluxes and soil temperature. Compared to CK, yields for the T1, T2, and T4 were significantly reduced (P < 0.05) but the yield for T3 was only reduced by 2.3% (P > 0.05); T3 significantly reduced soil CO2 emissions by 10.2% (P < 0.05) and reduced the irrigation water amount by 5.7%. CONCLUSION: Deficit irrigation effectively reduced CO2 emissions from winter wheat field soils. T3 may be a water-saving, CO2 emission-reducing and high-yield irrigation regime for winter wheat fields in northwest China. The research laid a preliminary theoretical foundation for formulating winter wheat irrigation systems that are water saving, emission reducing, and that produce high yields. © 2019 Society of Chemical Industry.

Hamiltonian replica exchange simulations of glucose oxidase adsorption on charged surfaces.

The Hamiltonian replica exchange Monte Carlo (H-REMC) algorithm was applied to study protein adsorption and its performance was compared with that of the temperature replica exchange Monte Carlo (T-REMC). Comparisons indicate that the simulation results are consistent but the computational efficiency is improved for H-REMC. H-REMC could accurately and efficiently identify the preferred orientations of glucose oxidase (GOx) on charged surfaces; different preferential GOx orientations on different surfaces and solution conditions could be spotted with a much fewer number of simulation runs. On positively charged surfaces, when electrostatic interactions dominate, the negatively charged GOx can be easily adsorbed with the "standing" orientation for which the substrate-binding domain is accessible to substrates. As the surface charge densities decrease and ionic strengths increase, there is an increasing contribution from the van der Waals (vdW) interactions, and thus more possible orientations appear. When the vdW interactions dominate, the unfavorable "front-lying" becomes the preferred orientation for which the substrate-binding domain is blocked by the surface. On negatively charged surfaces, though GOx has a net charge of -30 e under physiological conditions, the charged groups are unevenly distributed over the protein surface; the positive potential regions in the "back" of GOx enable the protein to be adsorbed on negatively charged surfaces with the "back-lying" orientation. The H-REMC provides an alternative method to accurately and efficiently probe the lowest-energy orientation of proteins adsorbed on surfaces for biotechnological applications.

Identification of drug-target interaction from interactome network with 'guilt-by-association' principle and topology features.

MOTIVATION: Identifying drug-target protein interaction is a crucial step in the process of drug research and development. Wet-lab experiment are laborious, time-consuming and expensive. Hence, there is a strong demand for the development of a novel theoretical method to identify potential interaction between drug and target protein. RESULTS: We use all known proteins and drugs to construct a nodes- and edges-weighted biological relevant interactome network. On the basis of the 'guilt-by-association' principle, novel network topology features are proposed to characterize interaction pairs and random forest algorithm is employed to identify potential drug-protein interaction. Accuracy of 92.53% derived from the 10-fold cross-validation is about 10% higher than that of the existing method. We identify 2272 potential drug-target interactions, some of which are associated with diseases, such as Torg-Winchester syndrome and rhabdomyosarcoma. The proposed method can not only accurately predict the interaction between drug molecule and target protein, but also help disease treatment and drug discovery. CONTACTS: zhanchao8052@gmail.com or ceszxy@mail.sysu.edu.cn SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Identifying and prioritizing disease-related genes based on the network topological features.

Identifying and prioritizing disease-related genes are the most important steps for understanding the pathogenesis and discovering the therapeutic targets. The experimental examination of these genes is very expensive and laborious, and usually has a higher false positive rate. Therefore, it is highly desirable to develop computational methods for the identification and prioritization of disease-related genes. In this study, we develop a powerful method to identify and prioritize candidate disease genes. The novel network topological features with local and global information are proposed and adopted to characterize genes. The performance of these novel features is verified based on the 10-fold cross-validation test and leave-one-out cross-validation test. The proposed features are compared with the published features, and fused strategy is investigated by combining the current features with the published features. And, these combination features are also utilized to identify and prioritize Parkinson's disease-related genes. The results indicate that identified genes are highly related to some molecular process and biological function, which provides new clues for researching pathogenesis of Parkinson's disease. The source code of Matlab is freely available on request from the authors.

TMSI-promoted vinylogous Michael addition of siloxyfuran to 2-substituted chromones: a general approach for the total synthesis of chromanone lactone natural products.

A concise and facile synthetic protocol for the construction of the 2-γ-lactone chromanone skeleton has been achieved through a TMSI-promoted diastereoselective vinylogous Michael addition of siloxyfuran to 2-substituted chromones. The applicability of this method is demonstrated through the rapid access to the total syntheses of (±)-microdiplodiasone, (±)-lachnone C, and (±)-gonytolides C and G.

Determination of triazine herbicides in fresh vegetables by dynamic microwave-assisted extraction coupled with homogeneous ionic liquid microextraction high performance liquid chromatography.

A novel extraction method, dynamic microwave-assisted extraction coupled with homogeneous ionic liquid microextraction, was developed for the determination of triazine herbicides, including desmetryn, terbumeton, propazine, terbuthylazine, dimethametryn, and dipropetryn in fresh vegetable samples by high performance liquid chromatography (HPLC). In the developed method, 120 µL of 1-butyl-3-methylimidazolium tetrafluoroborate ([C4MIM][BF4]) was added to 10 mL of aqueous solution containing 0.3 g of NaCl to obtained the extraction solvent. Six triazines could be extracted completely within 4 min by the present method. Then, [NH4][PF6] was added into the extract to form a water-insoluble ionic liquid [C4MIM][PF6] via a simple metathesis reaction, and the analytes were enriched into the ionic liquid phase. After centrifugation and dilution with acetonitrile, the resulting solution was analyzed directly by HPLC. The effects of some experimental parameters, including type and volume of ionic liquid, volume of extraction solvent, amount of ion-pairing agent [NH4][PF6], salt concentration, microwave power, and flow rate of extraction solvent on the extraction efficiency were investigated and optimized. Under the optimum experimental conditions, the linearity for determining the analytes was in the range of 2.50-250.00 µg kg(-1), with the correlation coefficients ranging from 0.9989 to 0.9999. When the present method was applied to the analysis of vegetable samples, satisfactory recoveries were obtained in the range of 76.8%-106.9%, and relative standard deviations were lower than 9.8%.

Identifying Protein Phosphorylation Site-Disease Associations Based on Multi-Similarity Fusion and Negative Sample Selection by Convolutional Neural Network.

As one of the most important post-translational modifications (PTMs), protein phosphorylation plays a key role in a variety of biological processes. Many studies have shown that protein phosphorylation is associated with various human diseases. Therefore, identifying protein phosphorylation site-disease associations can help to elucidate the pathogenesis of disease and discover new drug targets. Networks of sequence similarity and Gaussian interaction profile kernel similarity were constructed for phosphorylation sites, as well as networks of disease semantic similarity, disease symptom similarity and Gaussian interaction profile kernel similarity were constructed for diseases. To effectively combine different phosphorylation sites and disease similarity information, random walk with restart algorithm was used to obtain the topology information of the network. Then, the diffusion component analysis method was utilized to obtain the comprehensive phosphorylation site similarity and disease similarity. Meanwhile, the reliable negative samples were screened based on the Euclidean distance method. Finally, a convolutional neural network (CNN) model was constructed to identify potential associations between phosphorylation sites and diseases. Based on tenfold cross-validation, the evaluation indicators were obtained including accuracy of 93.48%, specificity of 96.82%, sensitivity of 90.15%, precision of 96.62%, Matthew's correlation coefficient of 0.8719, area under the receiver operating characteristic curve of 0.9786 and area under the precision-recall curve of 0.9836. Additionally, most of the top 20 predicted disease-related phosphorylation sites (19/20 for Alzheimer's disease; 20/16 for neuroblastoma) were verified by literatures and databases. These results show that the proposed method has an outstanding prediction performance and a high practical value.

Exposure to O3 and NO2 on the interfacial chemistry of the pulmonary surfactant and the mechanism of lung oxidative damage.

Exposure to ozone (O3) and nitrogen dioxide (NO2) are related to pulmonary dysfunctions and various lung diseases, but the underlying biochemical mechanisms remain uncertain. Herein, the effect of inhalable oxidizing gas pollutants on the pulmonary surfactant (PS, extracted from porcine lungs), a mixture of active lipids and proteins that plays an important role in maintaining normal respiratory mechanics, is investigated in terms of the interfacial chemistry using in-vitro experiments; and the oxidative stress induced by oxidizing gases in the simulated lung fluid (SLF) supplemented with the PS is explored. The results showed that O3 and NO2 individually increased the surface tension of the PS and reduced its foaming ability; this was accompanied by the surface pressure-area isotherms of the PS monolayers shifting toward lower molecular areas, with O3 exhibiting more severe effects than NO2. Moreover, both O3 and NO2 produced reactive oxygen species (ROS) resulting in lipid peroxidation and protein damage to the PS. The formation of superoxide radicals (O2•-) was correlated with the decomposition of O3 and the reactions of O3 and NO2 with antioxidants in the SLF. These radicals, in the presence of antioxidants, led to the formation of hydrogen peroxide and hydroxyl radicals (•OH). Additionally, the direct oxidation of unsaturated lipids by O3 and NO2 further caused an increase in the ROS content. This change in the ROS chemistry and increased •OH production tentatively explain how inhalable oxidizing gases lead to oxidative stress and adverse health effects. In summary, our results indicated that inhaled O3 and NO2 exposure can significantly alter the interfacial properties of the PS, oxidize its active ingredients, and induce ROS formation in the SLF. The results of this study provide a basis for the elucidation of the potential hazards of inhaled oxidizing gas pollutants in the human respiratory system.

Dye-doped cellulose nanocrystals as novel dusting powders for visualizing latent fingerprints.

Powder-dusting method based on the visual contrast between the background surface and powder-covered ridges of a fingerprint is widely used to develop the invisible latent fingerprints (LFPs) left at crime scenes. Recently, the development of nano-sized powders with excellent optical performances has been extensively explored. In this work, we employed environmentally friendly and low-toxicity cellulose nanocrystals as the novel support. Using dye-doped cellulose nanocrystals as novel dusting powders, two dyes (phenylfluorone and curcumin) were adsorbed on the cellulose nanocrystals by a simple batch adsorption method. The dye-doped cellulose nanocrystals (namely, phenylfluorone-doped cellulose nanocrystals (PDCN) and curcumin-doped cellulose nanocrystals (CDCN)) containing 2% of the loaded mass of both the dyes with bright green fluorescence were developed to visualize LFPs on the surfaces of various substrates (such as glass slide, printing paper, orange plastic card, tile, stainless steel, compact disc, red plastic packing, copper foil and aluminum foil). Images of the LFPs can been obtained by both the dye-doped cellulose nanocrystals with sufficient affinity to the ridges of LFPs. High-quality ridge details with features at the second and third level can be detected by CDCN, whereas PDCN only display the secondary-level features of ridge details. Compared with PDCN, CDCN illustrate higher sensitivity, higher selectivity, and better contrast, especially for detecting fresh and non-fresh LFPs on porous and non-porous substrates, and has the potential for practical use in forensic science.

Green fabrication of modified lignin/zeolite/chitosan-based composite membranes for preservation of perishable foods.

Chitosan, as a kind of naturally occurring green and degradable material for the preservation of perishable foods, was investigated in this study with the objective of enhancing its preservation performances. Herein, lignin was modified using the solvent fractionation method (modified lignin, ML, including ML1-ML3), while natural clinoptilolite zeolite was modified using the alkali modification method (modified clinoptilolite zeolite, MCZ, including MCZ1-MCZ5). After optimizing the conditions, it was discovered that incorporating both ML3 and MCZ3 into pure chitosan-based membranes might be conducive to fabricate chitosan-based composite membranes for the preservation of perishable foods. As-prepared composite membranes possessed better visible light transmittance, antioxidant activity, and carbon dioxide/oxygen selectivity, resulting in improved preservation effects on the model perishable foods such as bananas, cherry tomatoes, and cheeses. These findings might indicate promising applications for chitosan-based composite membranes with modified lignin and zeolite in the field of eco-friendly degradable materials for the preservation of perishable foods.

Predicting promoters by pseudo-trinucleotide compositions based on discrete wavelets transform.

In this paper, the discrete wavelet transform was introduced into the trinucleotide compositions and a novel DNA sequence representation technique, namely pseudo-trinucleotide compositions was proposed. The pseudo-trinucleotide compositions based on discrete wavelets transform were then employed to model support vector machines (SVM) for the prediction of promoters. The model was evaluated on the genie dataset, and the overall prediction accuracy (ACC) by jackknife validation for the classification of promoters, introns and exons was 82.46%, while the ACC for the classification of promoters and unpromoters was 82.18%, which was far better than the previous results. The satisfied prediction result revealed that the pseudo-trinucleotide composition based on discrete wavelet transform was an effective representation method for DNA sequence, and plays a very important role in the prediction of DNA function.