Machine Learning of Functional Connectivity to Biotype Alcohol and Nicotine Use Disorders.

BACKGROUND: Magnetic resonance imaging provides noninvasive tools to investigate alcohol use disorder (AUD) and nicotine use disorder (NUD) and neural phenotypes for genetic studies. A data-driven transdiagnostic approach could provide a new perspective on the neurobiology of AUD and NUD. METHODS: Using samples of individuals with AUD (n = 140), individuals with NUD (n = 249), and healthy control participants (n = 461) from the UK Biobank, we integrated clinical, neuroimaging, and genetic markers to identify biotypes of AUD and NUD. We partitioned participants with AUD and NUD based on resting-state functional connectivity (FC) features associated with clinical metrics. A multitask artificial neural network was trained to evaluate the cluster-defined biotypes and jointly infer AUD and NUD diagnoses. RESULTS: Three biotypes-primary NUD, mixed NUD/AUD with depression and anxiety, and mixed AUD/NUD-were identified. Multitask classifiers incorporating biotype knowledge achieved higher area under the curve (AUD: 0.76, NUD: 0.74) than single-task classifiers without biotype differentiation (AUD: 0.61, NUD: 0.64). Cerebellar FC features were important in distinguishing the 3 biotypes. The biotype of mixed NUD/AUD with depression and anxiety demonstrated the largest number of FC features (n = 5), all related to the visual cortex, that significantly differed from healthy control participants and were validated in a replication sample (p < .05). A polymorphism in TNRC6A was associated with the mixed AUD/NUD biotype in both the discovery (p = 7.3 × 10-5) and replication (p = 4.2 × 10-2) sets. CONCLUSIONS: Biotyping and multitask learning using FC features can characterize the clinical and genetic profiles of AUD and NUD and help identify cerebellar and visual circuit markers to differentiate the AUD/NUD group from the healthy control group. These markers support a new growing body of literature.

Effect of fear of missing out on learning burnout in medical students: a moderated mediation.

Introduction: Learning burnout has a significant negative impact on students' academic performance and professional development, which has been exacerbated by the growing trend of problematic smartphone use, such as smartphone addiction, among young people. Recently, the literature on excessive social media use has revealed a critical role of fear of missing out. Objective The purpose of this study was to examine how fear of missing out affects smartphone addiction and its subsequent effect on learning burnout in college students. Methods: In Study 1, 352 medical students were recruited to complete a cross-sectional survey. In Study 2, 2,948 college students were recruited to complete a cross-sectional survey. Further in Study 3, 30 medical students were recruited into a mindfulness-based intervention program. Results: Study 1 preliminarily confirmed that fear of missing out was positively correlated with learning burnout. Study 2 then revealed a moderated mediation model showing that fear of missing out may increase smartphone addiction, which in turn affects their sleep quality and finally leads to learning burnout. This chain mediation model was moderated by the participants' level of mindfulness. To confirm the promoting role of mindfulness, Study 3 further confirmed that mindfulness training indeed can improve smartphone addiction and reduce learning burnout in medical students. Discussion: Theoretical and practical contributions were discussed, highlighting the effects of fear of missing out on smartphone addiction and a moderating role of mindfulness training.

Identifying alcohol misuse biotypes from neural connectivity markers and concurrent genetic associations.

Alcohol use behaviors are highly heterogeneous, posing significant challenges to etiologic research of alcohol use disorder (AUD). Magnetic resonance imaging (MRI) provides intermediate endophenotypes in characterizing problem alcohol use and assessing the genetic architecture of addictive behavior. We used connectivity features derived from resting state functional MRI to subtype alcohol misuse (AM) behavior. With a machine learning pipeline of feature selection, dimension reduction, clustering, and classification we identified three AM biotypes-mild, comorbid, and moderate AM biotypes (MIA, COA, and MOA)-from a Human Connectome Project (HCP) discovery sample (194 drinkers). The three groups and controls (397 non-drinkers) demonstrated significant differences in alcohol use frequency during the heaviest 12-month drinking period (MOA > MIA; COA > non-drinkers) and were distinguished by connectivity features involving the frontal, parietal, subcortical and default mode networks. Further, COA relative to MIA, MOA and controls endorsed significantly higher scores in antisocial personality. A genetic association study identified that an alcohol use and antisocial behavior related variant rs16930842 from LINC01414 was significantly associated with COA. Using a replication HCP sample (28 drinkers and 46 non-drinkers), we found that subtyping helped in classifying AM from controls (area under the curve or AUC = 0.70, P < 0.005) in comparison to classifiers without subtyping (AUC = 0.60, not significant) and successfully reproduced the genetic association. Together, the results suggest functional connectivities as important features in classifying AM subgroups and the utility of reducing the heterogeneity in connectivity features among AM subgroups in advancing the research of etiological neural markers of AUD.

An Efficient Algorithm for Deep Stochastic Contextual Bandits.

In stochastic contextual bandit (SCB) problems, an agent selects an action based on certain observed context to maximize the cumulative reward over iterations. Recently there have been a few studies using a deep neural network (DNN) to predict the expected reward for an action, and the DNN is trained by a stochastic gradient based method. However, convergence analysis has been greatly ignored to examine whether and where these methods converge. In this work, we formulate the SCB that uses a DNN reward function as a non-convex stochastic optimization problem, and design a stage-wise stochastic gradient descent algorithm to optimize the problem and determine the action policy. We prove that with high probability, the action sequence chosen by this algorithm converges to a greedy action policy respecting a local optimal reward function. Extensive experiments have been performed to demonstrate the effectiveness and efficiency of the proposed algorithm on multiple real-world datasets.

Perceived stress, self-efficacy, and the cerebral morphometric markers in binge-drinking young adults.

Studies have identified cerebral morphometric markers of binge drinking and implicated cortical regions in support of self-efficacy and stress regulation. However, it remains unclear how cortical structures of self-control play a role in ameliorating stress and alcohol consumption or how chronic alcohol exposure alters self-control and leads to emotional distress. We examined the data of 180 binge (131 men) and 256 non-binge (83 men) drinkers from the Human Connectome Project. We obtained data on regional cortical thickness from the HCP and derived gray matter volumes (GMVs) with voxel-based morphometry. At a corrected threshold, binge relative to non-binge drinking men showed diminished posterior cingulate cortex (PCC) thickness and dorsomedial prefrontal cortex (dmPFC) GMV. PCC thickness and dmPFC GMVs were positively and negatively correlated with self-efficacy and perceived stress, respectively, as assessed with the NIH Emotion Toolbox. Mediation and path analyses to query the inter-relationships between the neural markers and clinical variables showed a best fit of the model with daily drinks â†’ lower PCC thickness and dmPFC GMV â†’ lower self-efficacy â†’ higher perceived stress in men. In contrast, binge and non-binge drinking women did not show significant differences in regional cortical thickness or GMVs. These findings suggest a pathway whereby chronic alcohol consumption alters cortical structures and self-efficacy mediates the effects of cortical structural deficits on perceived stress in men. The findings also suggest the need to investigate multimodal neural markers underlying the interplay between stress, self-control and alcohol use behavior in women.

OctSurf: Efficient hierarchical voxel-based molecular surface representation for protein-ligand affinity prediction.

Voxel-based 3D convolutional neural networks (CNNs) have been applied to predict protein-ligand binding affinity. However, the memory usage and computation cost of these voxel-based approaches increase cubically with respect to spatial resolution and sometimes make volumetric CNNs intractable at higher resolutions. Therefore, it is necessary to develop memory-efficient alternatives that can accelerate the convolutional operation on 3D volumetric representations of the protein-ligand interaction. In this study, we implement a novel volumetric representation, OctSurf, to characterize the 3D molecular surface of protein binding pockets and bound ligands. The OctSurf surface representation is built based on the octree data structure, which has been widely used in computer graphics to efficiently represent and store 3D object data. Vanilla 3D-CNN approaches often divide the 3D space of objects into equal-sized voxels. In contrast, OctSurf recursively partitions the 3D space containing the protein-ligand pocket into eight subspaces called octants. Only those octants containing van der Waals surface points of protein or ligand atoms undergo the recursive subdivision process until they reach the predefined octree depth, whereas unoccupied octants are kept intact to reduce the memory cost. Resulting non-empty leaf octants approximate molecular surfaces of the protein pocket and bound ligands. These surface octants, along with their chemical and geometric features, are used as the input to 3D-CNNs. Two kinds of CNN architectures, VGG and ResNet, are applied to the OctSurf representation to predict binding affinity. The OctSurf representation consumes much less memory than the conventional voxel representation at the same resolution. By restricting the convolution operation to only octants of the smallest size, our method also alleviates the overall computational overhead of CNN. A series of experiments are performed to demonstrate the disk storage and computational efficiency of the proposed learning method. Our code is available at the following GitHub repository: https://github.uconn.edu/mldrugdiscovery/OctSurf.

Communication-Optimal Distributed Dynamic Graph Clustering.

We consider the problem of clustering graph nodes over large-scale dynamic graphs, such as citation networks, images and web networks, when graph updates such as node/edge insertions/deletions are observed distributively. We propose communication-efficient algorithms for two well-established communication models namely the message passing and the blackboard models. Given a graph with n nodes that is observed at s remote sites over time [1, t], the two proposed algorithms have communication costs Õ(ns) and Õ(n + s) (Õ hides a polylogarithmic factor), almost matching their lower bounds, Ω(ns) and Ω (n + s), respectively, in the message passing and the blackboard models. More importantly, we prove that at each time point in [1, t] our algorithms generate clustering quality nearly as good as that of centralizing all updates up to that time and then applying a standard centralized clustering algorithm. We conducted extensive experiments on both synthetic and real-life datasets which confirmed the communication efficiency of our approach over baseline algorithms while achieving comparable clustering results.

Characterization and source apportionment of size-segregated atmospheric particulate matter collected at ground level and from the urban canopy in Tianjin.

To investigate the size distributions of chemical compositions and sources of particulate matter (PM) at ground level and from the urban canopy, a study was conducted on a 255 m meteorological tower in Tianjin from December 2013 to January 2014. Thirteen sets of 8 size-segregated particles were collected with cascade impactor at 10 m and 220 m. Twelve components of particles, including water-soluble inorganic ions and carbonaceous species, were analyzed and used to apportion the sources of PM with positive matrix factorization. Our results indicated that the concentrations, size distributions of chemical compositions and sources of PM at the urban canopy were affected by regional transport due to a stable layer approximately 200 m and higher wind speed at 220 m. The concentrations of PM, Cl- and elemental carbon (EC) in fine particles at 10 m were higher than that at 220 m, while the reverse was true for NO3- and SO42-. The concentrations of Na+, Ca2+, Mg2+, Cl- and EC in coarse particles at 10 m were higher than that at 220 m. The size distributions of major primary species, such as Cl-, Na+, Ca2+, Mg2+ and EC, were similar at two different heights, indicating that there were common and dominant sources. The peaks of SO42-, NH4+, NO3- and organic carbon (OC), which were partly secondary generated species, shifted slightly to the smaller particles at 220 m, indicating that there was a different formation mechanism. Industrial pollution and coal combustion, re-suspended dust and marine salt, traffic emissions and transport, and secondary inorganic aerosols were the major sources of PM at both heights. With the increase in vertical height, the influence of traffic emissions, re-suspended dust and biomass burning on PM weakened, but the characteristics of regional transport from Hebei Province and Beijing gradually become obvious.

Vertical characteristics of levels and potential sources of water-soluble ions in PM10 in a Chinese megacity.

To investigate the vertical characteristics of ions in PM10 as well as the contributions and possible locations of their sources, eight water-soluble ions were measured at four heights simultaneously along a meteorological tower in Tianjin, China. The total ion concentrations showed a general decreasing trend with increasing height, ranging from 64.94µgm(-3) at 10m to 44.56µgm(-3) at 220m. NH4(+), SO4(2-) and NO3(-) showed higher height-to-height correlations. In addition, relationships between ions are discussed using Pearson correlation coefficients and hierarchical clustering analysis (HCA), which implied that, for each height, the correlations among NH4(+), SO4(2-) and NO3(-) were higher. Finally, sources were identified qualitatively by the ratio of certain ions and quantitatively by principal component analysis/multiple linear regression (PCA/MLR) and positive matrix factorisation (PMF). Secondary sources played a dominant role for PM10 and water-soluble ions at four heights and became more important at greater heights (the percentage contributions were 43.04-66.41% for four heights by PCA/MLR and 46.93-67.62% by PMF). Then, the redistributed concentration field (RCF) combined with PCA/MLR and PMF was applied, which indicated the high potential source regions. The vertical characteristics of the levels, relationships, source contributions and locations would support the effective management of the water-soluble ions in particulate matter.

Determination of buffering capacity of total suspended particle and its source apportionment using the chemical mass balance approach.

The samples of total suspended particle (TSP) from sources and TSP in the ambient atmosphere were collected in 2006 at Tianjin, People's Republic of China and analyzed for 16 chemical elements, two water-soluble ions, total carbon, and organic carbon. On the basis of the chemical mass balance (CMB) model, the contributions of different TSP sources to the ambient TSP were identified. The results showed that resuspended dust has the biggest contributions to the concentration of ambient TSP. The buffering capacity of each TSP source was also determined by an analytical chemistry method, and the result showed that the constructive dust (the dust emitted from construction work) had the strongest buffering capacity among the measured sources, whereas the coal combustion dust had the weakest buffering capacity. A calculation formula of the source of buffering capacity of ambient TSP was developed based on the result of TSP source apportionment and the identification of the buffering capacity of each TSP source in this study. The results of the source apportionment of the buffering capacity of ambient TSP indicated that open sources (including soil dust, resuspended dust, and constructive dust) were the dominant sources of the buffering capacity of the ambient TSP. Acid rain pollution in cities in Northern China might become serious with a decrease of open source pollution without reducing acidic sources. More efforts must be made to evaluate this potential risk, and countermeasures should be proposed as early as possible.

Application of activated carbon derived from scrap tires for adsorption of Rhodamine B.

Activated carbon derived from solid hazardous waste scrap tires was evaluated as a potential adsorbent for cationic dye removal. The adsorption process with respect to operating parameters was investigated to evaluate the adsorption characteristics of the activated pyrolytic tire char (APTC) for Rhodamine B (RhB). Systematic research including equilibrium, kinetics and thermodynamic studies was performed. The results showed that APTC was a potential adsorbent for RhB with a higher adsorption capacity than most adsorbents. Solution pH and temperature exert significant influence while ionic strength showed little effect on the adsorption process. The adsorption equilibrium data obey Langmuir isotherm and the kinetic data were well described by the pseudo second-order kinetic model. The adsorption process followed intra-particle diffusion model with more than one process affecting the adsorption process. Thermodynamic study confirmed that the adsorption was a physisorption process with spontaneous, endothermic and random characteristics.

Synthesis and the effect of calcination temperature on the physical-chemical properties and photocatalytic activities of Ni,La codoped SrTiO3.

A series of highly vis-light active Ni,La-codoped SrTiO(3) photocatalysts were successfully synthesized with sol-gel process. The characterization results show that the calcination temperature has a strong influence on the physical-chemical properties of as-synthesized photocatalysts. The surface area and porosity, even the initial adsorption rate for malachite green (MG), decreased with increasing calcination temperature. To evaluate the photocatalytic activities, the photodegradation of a water contaminant (MG) was carried out under visible light irradiation. The as-synthesized photocatalysts exhibited a high vis-light activity, and a 100% degradation of MG was observed for the Ni,La-SrTiO(3)-x catalysts calcined at low temperature under visible light irradiation for 1h, during which only 7% and 15% of MG was degraded for self-degrade and commercially available photocatalyst Degussa P25, respectively. The high vis-light activity is a result of the best combination of many properties, such as the intensive visible light response, the large surface area and pore volume and the high initial adsorption rate for substrate.

Use of a nonnegative constrained principal component regression chemical mass balance model to study the contributions of nearly collinear sources.

In this study, a nonnegative constrained principal component regression chemical mass balance (NCPCRCMB) model was used to solve the near collinearity problem among source profiles for source apportionment. The NCPCRCMB model added the principle component regression route into the CMB model iteration. The model was tested with the synthetic data sets, which involved contributions from eleven actual sources, with a serious near collinearity problem among them. The actual source profiles were randomly perturbed and then applied to create the synthetic receptor. The resulting synthetic receptor concentrations were also randomly perturbed to simulate measurement errors. The synthetic receptors were separately apportioned by CMB and NCPCRCMB model. The result showed that source contributions estimated by the NCPCRCMB model were much closer to the true values than those estimated by the CMB model. Next, five real ambient data sets from five cities in China were analyzed using the NCPCRCMB model to test the model practicability. Reasonable results were obtained in all cases. It is shown that the NCPCRCMB model has an advantage over the traditional CMB model when dealing with near collinearity problems in source apportionment studies.

Secondary organic carbon quantification and source apportionment of PM10 in Kaifeng, China.

During 2005, the filter samples of ambient PM10 from five sites and the source samples of particulate matter were collected in Kaifeng, Henan Province of China. Nineteen elements, water-soluble ions, total carbon (TC) and organic carbon (OC) contained in samples were analyzed. Seven contributive source types were identified and their contributions to ambient PM10 were estimated by chemical mass balance (CMB) receptor model. Weak associations between the concentrations of organic carbon and element carbon (EC) were observed during the sampling periods, indicating that there was secondary organic aerosol pollution in the urban atmosphere. An indirect method of "OC/EC minimum ratio" was applied to estimate the concentration of secondary organic carbon (SOC). The results showed that SOC contributed 26.2%, 32.4% and 18.0% of TC in spring, summer-fall and winter, respectively, and the annual average SOC concentration was 7.07 microg/m3, accounting for 5.73% of the total mass in ambient PM10. The carbon species concentrations in ambient PM10 were recalculated by subtracting SOC concentrations from measured concentrations of TC and OC to increase the compatibility of source and receptor measurements for CMB model.

Chemical and stable carbon isotopic characterization for PAHs in aerosol emitted from two indoor sources.

Polycyclic aromatic hydrocarbons (PAHs) are with great environmental concerns due to their toxic, mutagenic, and carcinogenic properties. The harmful effects caused by PAHs emitted from indoor sources may be more direct and serious. In this study, PAHs in aerosol from two typical indoor sources, the cooking fume (CF) and environmental tobacco smoke (ETS) were collected by simulation emission in a chamber. Eighteen PAHs were analyzed by GC/MS and GC/C/IRMS. The chemical profiles of these two sources were described. Results indicated that Fluoranthene, Pyrene and Fluorene in CF, and Fluorene, Phenanthrene, Indeno(1,2,3-cd)pyrene and Benz(a)anthracene in ETS, relative quantity of which were variable in a smaller range, can be regarded as tracers of indoor PAHs sources. There are distinct differences among the ratios of Benz(a)anthracene/Chrysene, Benzo(e)pyrene/Benzo(a)pyrene and Indeno(1,2,3-cd)pyrene/Benzo(g,h,i)perylene between CF and EST. Distribution of delta(13)C of individual PAHs in ETS samples ranged from -21.76 per thousand to -29.32 per thousand, wider than that in CF samples (-22.94 per thousand to -28.39 per thousand). The delta(13)C of Phenanthrene, Benz(a)anthracene, Benzo(b)fluoranthene, Benzo(k)fluoranthene and Indeno(1,2,3-cd)pyrene between these two sources showed great differences. The (13)C was enriched in low molecule weight compounds of CF samples and in high molecule weight compounds of ETS samples.

Using geoaccumulation index to study source profiles of soil dust in China.

Source apportionment studies of TSP (atmospheric particulate matter with aerodynamic diameters < or = 100 microm) and PM10 (atmospheric particulate matter with aerodynamic diameters < or = 10microm) have revealed that soil dust is an important source of these particulates in China. In this study, the contamination of soil dust was assessed through the use of a geoaccumulation index (I(geo)). The mass concentration profiles of 17 elements (Na, Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Br, Ba, and Pb) were established for urban soil dusts. Geochemical compositions of soils from 15 cities were used to represent background urban soil compositions. The results of this study indicated that a number of cities are severely polluted by particulates containing Ca, Cr, Ni, and Cu in both size fractions (TSP and PM10). Contamination with Zn, Pb, Co, and Br was moderate to severe (I(geo) > 2). The Al and Fe concentrations were not high enough for them to be considered contaminants.

Occurrence and behavior of four of the most used sunscreen UV filters in a wastewater reclamation plant.

Four UV filters, benzophenone-3 (BP-3), 4-methylbenzylidence camphor (4-MBC), ethylhexyl methoxycinnamate (EHMC), and octocrylene (OC), have been examined along the different units of a wastewater reclamation plant (WWRP) located in Tianjin, North China. The analytical procedure included solid-phase extraction and gas chromatographic/mass spectrometric analyses. All four UV filters were detected in the influent during the three sampling campaigns (February, July, and September), and the concentrations ranged from 34 to 2128 ng L(-1). The concentrations of the four UV filters were higher in hot weather (July and September) than in cool weather (February). The monthly average removal ranged from 7.6% to 21% for the selected UV filters during coagulation-flocculation (C-F) treatment. The ozonation treatment achieved the maximum removal (16-28%); on the contrary, the continuous microfiltration (CMF) achieved the lowest removal (3.6-8.2%). The total removal efficiencies along the plant varied from 28% to 43%. These results indicate that the UV filters are not completely removed during WWRP treatment and may be carried over into the environment during the reuse applications.

Health risk assessment for traffic policemen exposed to polycyclic aromatic hydrocarbons (PAHs) in Tianjin, China.

In China, traffic policemen have to stand for several hours a day at the road intersections with high vehicle flows. To assess their exposure to airborne carcinogenic polycyclic aromatic hydrocarbons (PAHs) during their working time, a preliminary study was conducted to measure the personal exposure level to PAHs. And a probabilistic incremental lifetime cancer risk (ILCR) model together with the benzo[a]pyrene (BaP) toxic equivalents (BaP(eq)) method was used to conduct health risk assessment. Personal exposure monitors (PEM) were carried by traffic policemen to collect PM10 samples during their daily work in Tianjin, China. Meanwhile, PM100 samples were collected at the roadsides and on campus of Nankai University as comparison. PAHs species were quantitatively analyzed by GC/MS. Experimental results showed that the concentrations of total PAHs, BaP and BaP(eq) were much higher at the road intersections (867.5, 26.2, 82.4 ng m(-3)), where the traffic policemen stand during their work time, than those at the roadsides (46.6, 1.5, 5.7 ng m(-3)), and on campus (19.5, 0.7, 2.4 ng m(-3)). According to the risk assessment results, the occupational risk falls within the range from 10(-6) to 10(-3). On the basis of sensitivity analysis results, further research should be directed to give better characterization of the yearly concentration distribution of PAHs and the cancer slope factor (CSF) of BaP in order to improve the accuracy of the health risk assessment.

Source apportionment of ambient total suspended particulates and coarse particulate matter in urban areas of Jiaozuo, China.

Approximately 750 total suspended particulates (TSPs) and coarse particulate matter (PM10) filter samples from six urban sites and a background site and >210 source samples were collected in Jiaozuo City during January 2002 to April 2003. They were analyzed for mass and abundances of 25 chemical components. Seven contributive sources were identified, and their contributions to ambient TSP/PM10 levels at the seven sites in three seasons (spring, summer, and winter days) and a "whole" year were estimated by a chemical mass balance (CMB) receptor model. The spatial TSP average was high in spring and winter days at a level of approximately 530 microg/m(3) and low in summer days at 456 microg/m(3); however, the spatial PMo0 average exhibited little variation at a level of approximately 325 microg/m(3), and PM10-to-TSP ratios ranged from 0.58 to 0.81, which suggested heavy particulate matter pollution existing in the urban areas. Apportionment results indicated that geological material was the largest contributor to ambient TSP/PM10 concentrations, followed by dust emissions from construction activities, coal combustion, secondary aerosols, vehicle movement, and other industrial sources. In addition, paved road dust and re-entrained dust were also apportioned to the seven source types and found soil, coal combustion, and construction dust to be the major contributors.

Source analysis of particulate-phase polycyclic aromatic hydrocarbons in an urban atmosphere of a northern city in China.

Particle-associated polycyclic aromatic hydrocarbon (PAH) concentrations were investigated at six sampling sites in the heating (February to March 2001) and nonheating (August to September 2001) periods in an industrial city in Northern China. Thirteen PAHs were measured. The total average concentrations (nanograms per meter cubed) of PAHs ranged between 78.93 and 214.63 during the heating period and from 31.48 to 102.26 in the nonheating period. Benzo(a)pyrene occurred at the highest level at a site near an industrial area but occurred at low concentrations far from the city center and industrial areas. In addition, ambient PAH profiles were studied. The five and six-ring species occurred in high fractions at the sampling site. By diagnostic ratio analysis, the major source at each sampling site in the city was coal combustion in the heating period; in the nonheating period, the major sources were relatively complex. Finally, the similarities among the six regions were assessed by principal component analysis, cluster analysis, and coefficient of divergence. These multivariate statistical analyses produced similar results, which agreed with the results from the diagnostic ratio analysis.