Bayesian Hierarchical Clustering Research Articles

Abstract PO5-09-02: Breast Tissue Proteomic Profile of Breast Cancer in Premenopausal Women and Association with Mammographic Breast Density

Abstract Introduction: Breast cancer incidence is rising in premenopausal, hence, there is a critical need to understand factors underlying premenopausal breast cancer development in order to guide targeted prevention. Mammographic breast density is a strong risk factor for, as well as an intermediate phenotype for premenopausal breast cancer. Yet, the molecular mechanisms underlying the associations of dense breasts with breast cancer are not well understood. Our objectives in this study are to perform proteomic analysis in breast tissues to (i) identify proteins that are associated with breast cancer development in premenopausal women; (ii) determine which of these proteins are also associated with dense breasts. Methods: We performed proteomic analysis on tumor and adjacent normal tissues from 50 premenopausal women with breast cancer who had breast tissue samples archived at the St. Louis Breast Tumor Registry. Samples were analyzed on Orbitrap Eclipse Tribrid coupled with Vanquish Neo LC system (Thermo Fisher Scientific). Database search was performed using MaxQuant and MS1 LFQ intensities were used for further data analysis. We performed multiple imputation using Bayesian Principal Component Analysis and hierarchical clustering analysis using ‘ConsensusClusterPlus’ package in R. Pathway enrichment analysis was performed using Reactome Pathway database. 10 times repeated 5-fold cross validation using Logistic Regression (LR) was performed for model development. Top 10 proteins were selected by variable importance in logistic regression, and recursive feature elimination was performed for feature selection. We corrected for multiple testing by setting the false discovery rate, FDR, p-value < 0.05. Results: After imputing for missing values, we identified 3,571 proteins that were expressed in tumor and normal breast tissues. We selected 1,640 proteins for further analysis using median absolute deviation analysis. In multivariable-adjusted analysis, 451 proteins were up-regulated, and 180 proteins were down-regulated in tumor tissues compared to adjacent normal tissues (FDR < 0.05). 53 of the 451 proteins that were upregulated in tumor tissues were also upregulated in dense breasts. Some of the top proteins that were upregulated in both tumor tissues and dense breasts were AIMP2, ALDH18A1, BZW1, CKAP5, COPG1, ERGIC1, GSPT1, ILF2, IPO7, LRBA, and PSMD12. Mammographic breast density protein clusters were highly correlated with hormone receptor positivity clusters and were enriched in ESR-mediated signaling, IGF1R signaling, NOTCH4, PTEN regulation, and NFkB pathway. Discussion: Using untargeted proteomics, we identified proteomic signatures of breast cancer development in premenopausal women. We further identified protein biomarkers and pathways that are shared between breast tumors and dense breast tissues in premenopausal women. Our novel findings highlight mammographic breast density as a strong intermediate phenotype that can be targeted in breast cancer prevention in premenopausal women. Larger studies are needed to validate our findings. Citation Format: Shaili Tapiavala, Minsoo Son, Graham Colditz, Ah Young Goo, Adetunji Toriola. Breast Tissue Proteomic Profile of Breast Cancer in Premenopausal Women and Association with Mammographic Breast Density [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-09-02.

Evolutionary dynamics of the accessory genomes of Staphylococcus aureus.

Staphylococcus aureus is a ubiquitous commensal and opportunistic bacterial pathogen that can cause a wide gamut of infections, which are exacerbated by the presence of multidrug-resistant and methicillin-resistant S. aureus. S. aureus is genetically heterogeneous and consists of numerous distinct lineages. Using 558 complete genomes of S. aureus, we aim to determine how the accessory genome content among phylogenetic lineages of S. aureus is structured and has evolved. Bayesian hierarchical clustering identified 10 sequence clusters, of which seven contained major sequence types (ST 1, 5, 8, 30, 59, 239, and 398). The seven sequence clusters differed in their accessory gene content, including genes associated with antimicrobial resistance and virulence. Focusing on the two largest clusters, BAPS8 and BAPS10, and each consisting mostly of ST5 and ST8, respectively, we found that the structure and connected components in the co-occurrence networks of accessory genomes varied between them. These differences are explained, in part, by the variation in the rates at which the two sequence clusters gained and lost accessory genes, with the highest rate of gene accumulation occurring recently in their evolutionary histories. We also identified a divergent group within BAPS10 that has experienced high gene gain and loss early in its history. Together, our results show highly variable and dynamic accessory genomes in S. aureus that are structured by the history of the specific lineages that carry them.IMPORTANCEStaphylococcus aureus is an opportunistic, multi-host pathogen that can cause a variety of benign and life-threatening infections. Our results revealed considerable differences in the structure and evolution of the accessory genomes of major lineages within S. aureus. Such genomic variation within a species can have important implications on disease epidemiology, pathogenesis of infection, and interactions with the vertebrate host. Our findings provide important insights into the underlying genetic basis for the success of S. aureus as a highly adaptable and resistant pathogen, which will inform current efforts to control and treat staphylococcal diseases.

Data Fusion-Based Approach for the Investigation of Reaction Networks in Hydrous Pyrolysis of Biomass

In this work, we present and validate a methodology for generating reaction networks from spectroscopic data using data-driven methods by applying it to the hydrothermal liquefaction (HTL) of Monterrey pine biomass and its constituents, viz., cellulose and lignin. This work is presented as a step toward automated inference of chemistry of the hydrothermal liquefaction process, thus limiting the need for human expertise. Bayesian hierarchical clustering of spectra and self-modeling multivariate spectral curve resolution are used to generate groups of chemically similar species, the reaction networks among which have been developed using Bayesian networks. Fourier transform infrared spectroscopy and proton nuclear magnetic resonance spectroscopy-based measurements are used as input data. The data-driven reaction network includes pathways representing decomposition of the biomass components, large molecule hydrolysis, and reformation of produced molecules and is consistent with the literature. Furthermore, the comparison of the networks generated for biomass and its components (levoglucosan, representing cellulose, and 2-phenoxy-ethyl benzene, representing lignin) reveals the relationship between the biomass HTL reaction network and the reaction networks of the components. The data-driven approach provides a diagnostic tool to identify the most probable reaction chemistry for complex biomass feedstocks and can be used for process understanding, design, and control.

Single Cell RNA-Seq Reveals Intra-Tumoral Heterogeneity Relevant to Differentiation States and Outcomes Among Newly Diagnosed Acute Myeloid Leukemia Patients

Background: Intra-tumoral heterogeneity emerging from aberrant differentiation and the accumulation of blasts at different maturation stages underscores AML biology. Importantly, the heterogeneity of AML cells contributes to therapy resistance and poor outcomes. Revealing the underlying biological programs that govern AML differentiation and are shared among patients can translate into effective therapeutic strategies. Methods: We leveraged single cell gene expression profiling (scRNA) of bone marrow mononulear cells from 20 newly diagnosed adult AML patients (median age 73 years; range 52-87 years). After doublet removal and quality assessment, a total of 111,130 single cells passed the filtering criteria. Of those, 56,168 (50.5%) cells (patient range: 293- 9781 cells) were identified as AML cells by integrating flow cytometry, immunohistochemistry, conventional cytogenetic and FISH at time of diagnosis with the scRNA expression profiles. Results: To determine the shared cellular states across different AML patients, we first applied an unsupervised Bayesian hierarchical clustering and identified four AML cell clusters. The four clusters had distinct expression of the lineage and maturation markers: CD34 (primitive), CD33 (myeloid), and CD14 and CD68 (monocytic), suggesting different AML maturation states in these clusters. To further explore the degree of differentiation in each cluster, we applied diffusion mapping, Slingshot pseudotime analysis and CytoTRACE profiling, all of which supported cluster-specific differentiation states. Specifically, Cluster 2 (CD34 high, CD14 negative) was early in the pseudotemporal trajectory, while Cluster 1 (CD14 high, CD34 negative) was towards the end of the trajectory. Clusters 3 and 4 were intermediary. Remarkably, each of the four clusters had contributions from at least 16 patients. These results support our approach to unmask intra-tumoral heterogeneity patterns associated with maturation and shared across different patients, independent of patient characteristics, molecular profiles and cytogenetics. To further explore the biologic pathways correlating with the four clusters that are surrogates of maturation states, we calculated single cell enrichment scores across a comprehensive gene signature collections from curated Molecular Signatures Database and other previously published studies. Cluster 1 was mainly enriched in immune-related pathways including IL6, JAK-STAT3 signaling, and interferon gamma response suggestive of an inflammatory component, and associated with the CD14+ CD68+ CD34- monocytic pattern of differentiation. Cluster 2 had the highest LSC104 stemness score and CD34 expression. Hedgehog signaling pathway was also highly enriched in this cluster and correlated with LSC104 score (r=0.28, p<0.01). This is of interest as hedgehog plays a fundamental role in the leukemic stem cell quiescence. Cluster 3 and 4, harboring lower CD34 expression compared to Cluster 2, but are CD14- CD68- were mainly enriched in metabolic processes including fatty acid metabolism, heme metabolism and erythroid lineage and oxidative phosphorylation, in addition to DNA repair and interferon alpha response. To understand how these defined populations and hierarchies relate to clinical outcomes, we deconvolved bulk RNA signatures in 3 AML cohorts: TCGA, BEAT AML and Abbas et al 2021, using the four clusters as reference. We then assigned each AML patient a cluster membership based on a deconvolution cut-off for cluster abundance. This translated into four distinct clinical groups with significantly different outcomes based on our cluster annotation. Remarkably, AML patients in Cluster 1 (monocytic enriched) had the worst outcomes (Panel B demonstrating TCGA cohort results, p=0.0084). Conclusions: We conducted unsupervised clustering of AML patients and characterized gene and pathway profiles of each cluster revealing a distinct spectrum of differentiation states from primitive to monocytic. Interestingly, AML patients with enrichment to monocytic phenotype characterized by higher inflammatory states demonstrated the worst outcome. Non-monocytic clusters are less variable in marker genes and stemness scores. However, the cluster with the highest CD34 expression are enriched in hedgehog signaling pathway, suggesting a potentially biological programs distinguishing non-monocytic AML cells. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Understanding the nature of face processing in early autism: A prospective study.

Dimensional approaches to psychopathology interrogate the core neurocognitive domains interacting at the individual level to shape diagnostic symptoms. Embedding this approach in prospective longitudinal studies could transform our understanding of the mechanisms underlying neurodevelopmental disorders. Such designs require us to move beyond traditional group comparisons and determine which domain-specific alterations apply at the level of the individual, and whether they vary across distinct phenotypic subgroups. As a proof of principle, this study examines how the domain of face processing contributes to the emergence of autism spectrum disorder (ASD). We used an event-related potentials (ERPs) task in a cohort of 8-month-old infants with (n = 148) and without (n = 68) an older sibling with ASD, and combined traditional case-control comparisons with machine-learning techniques for prediction of social traits and ASD diagnosis at 36 months, and Bayesian hierarchical clustering for stratification into subgroups. A broad profile of alterations in the time-course of neural processing of faces in infancy was predictive of later ASD, with a strong convergence in ERP features predicting social traits and diagnosis. We identified two main subgroups in ASD, defined by distinct patterns of neural responses to faces, which differed on later sensory sensitivity. Taken together, our findings suggest that individual differences between infants contribute to the diffuse pattern of alterations predictive of ASD in the first year of life. Moving from group-level comparisons to pattern recognition and stratification can help to understand and reduce heterogeneity in clinical cohorts, and improve our understanding of the mechanisms that lead to later neurodevelopmental outcomes.

Agglomerative and divisive hierarchical Bayesian clustering

Cluster analysis methods are designed to discover groups of subjects or objects in datasets by uncovering latent patterns in data. Two model-based Bayesian hierarchical clustering algorithms are presented—divisive and agglomerative—that return nested clustering configurations and provide guidance on the plausible number of clusters in a principled way. These algorithms outperform many existing clustering methods on benchmark data. The methods are applied to identify subpopulations among Parkinson's disease subjects using only baseline data, and differing patterns of progression in the few years following diagnosis are demonstrated in the identified clusters.

Advanced consistency management of highly-distributed transactional database in a hybrid cloud environment using novel R-TBC/RTA approach

This paper examines possibilities for improving the existing strategies of consistency management for highly-distributed transactional database in a hybrid cloud environment. With a detailed analysis of the existing consistency models for distributed database and standard strategies including Classic, Quorum and Tree Based Consistency (TBC), it is concluded that an improved advanced model of so-called visible adaptive consistency needs to be applied in a highly-distributed cloud environment, as necessary and sufficient degree of synchronization of all replicas. Along with the proposed model, research and development of an advanced novel strategy for consistency management Rose TBC (R-TBC) approach has been conducted, by improving standard TBC approach. Regarding implementation, a specific agglomerative Rose Tree Algorithm (RTA) has been developed, based on Bayesian hierarchical clustering and Graph Partitioning Algorithm - Multidimensional Data Clustering (GPA-MDC) intelligent partitioning of transactional Cloud Database Management System (CDBMS). The final result is constructed R-TBC model that changes in accordance with dynamic changes of entire heterogeneous CDBMS environment.

Data‐driven hypotheses of reaction networks for thermochemical conversion of a physical mixture of levoglucosan and 2‐phenoxyethyl benzene

AbstractIn this work, we analyze the hydrous pyrolysis of a physical mixture of the model components representing cellulose (levoglucosan) and lignin (2‐phenoxyethyl benzene). Fourier transform infrared (FTIR) and proton nuclear magnetic resonance (1H‐NMR) spectroscopy was used to characterize the products of the reaction. The main objective of the work was to use data‐driven methods to develop a reaction network for this system based on the spectroscopic data. This was achieved using Bayesian hierarchical clustering to identify pseudocomponents and Bayesian networks to develop a reaction network between these pseudocomponents. The data‐driven reaction network was shown to be consistent with the known chemistry of the pyrolysis of cellulose and lignin, and the chemistry of the physical mixture incorporated/combined elements of the reaction mechanisms of cellulose and lignin.

Hierarchical Bayesian clustering design of multiple biomarker subgroups (HCOMBS).

Given the Food and Drug Administration's (FDA's) acceptance of master protocol designs in recent guidance documents, the oncology field is rapidly moving to address the paradigm shift to molecular subtype focused studies. Identifying new "marker-based" treatments requires new methodologies to address the growing demand to conduct clinical trials in smaller molecular subpopulations, identify effective treatment and marker interactions, and control for false positives. We introduce our methodology, Hierarchical Bayesian Clustering Design of Multiple Biomarker Subgroups (HCOMBS), a two-stage umbrella Phase II design with effect size clustering and information borrowing across multiple biomarker-treatment pairs. HCOMBS was designed to reduce required sample size, differentiate between varying effect sizes, and control for operating characteristics in the multi-arm setting. When compared to independently applied Simon's Optimal two-stage design, we showed through simulations that HCOMBS required less participants per treatment arm with a well-controlled family-wise error rate and desirable marginal power. Additionally, HCOMBS features a statistical approach that simultaneously conducts clustering and hypothesis testing in one step. We also applied the proposed design on the alliance brain metastases umbrella trial.

Factors of Mobilization to Protest, Their Impact and Variability: Hierarchical Bayesian Approach

Mobilization to protest is a category of political action that has been observed quite often over the past two decades - for example, The Arab Spring, The Yellow Vests, and BLM movements can be named here. Indeed, mass protests are the voice that a crowd can use to deliver a directive message to political elites. But what can be called a driver of mobilization for protests is a debatable question. First, in different countries, the same indicator may give a different response. Secondly, the situation can change from year to year. Therefore, in this work, an attempt was made to identify what factors can be called drivers of the growth of mobilization to protests in 48 different countries of the world over an extended period of 15 years (from 2004 to 2018). The paper used methods of hierarchical Bayesian clustering, with the help of which it became possible to find out how the influence of factors of mobilization to protests differs when countries are grouped according to geographic (continental) and regime (level of democratization) characteristics. Also, the work managed to establish the fact that the most significant trigger (from the pool of those tested) to the mobilization of protests is the level of economic discontent of citizens - grievances.

Whole genomic comparative analysis of Streptococcus pneumoniae serotype 1 isolates causing invasive and non-invasive infections among children under 5 years in Casablanca, Morocco

BackgroundStreptococcus pneumoniae serotype 1 remains a leading cause of invasive pneumococcal diseases, even in countries with PCV-10/PCV-13 vaccine implementation. The main objective of this study, which is part of the Pneumococcal African Genome project (PAGe), was to determine the phylogenetic relationships of serotype 1 isolates recovered from children patients in Casablanca (Morocco), compared to these from other African countries; and to investigate the contribution of accessory genes and recombination events to the genetic diversity of this serotype.ResultsThe genome average size of the six-pneumococcus serotype 1 from Casablanca was 2,227,119 bp, and the average content of coding sequences was 2113, ranging from 2041 to 2161. Pangenome analysis of the 80 genomes used in this study revealed 1685 core genes and 1805 accessory genes. The phylogenetic tree based on core genes and the hierarchical bayesian clustering analysis revealed five sublineages with a phylogeographic structure by country. The Moroccan strains cluster in two different lineages, the five invasive strains clusters altogether in a divergent clade distantly related to the non-invasive strain, that cluster with all the serotype 1 genomes from Africa.ConclusionsThe whole genome sequencing provides increased resolution analysis of the highly virulent serotype 1 in Casablanca, Morocco. Our results are concordant with previous works, showing that the phylogeography of S. pneumoniae serotype 1 is structured by country, and despite the small size (six isolates) of the Moroccan sample, our analysis shows the genetic cohesion of the Moroccan invasive isolates.

Modeling of Value Indicators and Extraction of Key Management Points in Product Development using Bayesian Network and Hierarchical Clustering

Modeling of Value Indicators and Extraction of Key Management Points in Product Development using Bayesian Network and Hierarchical Clustering

Modeling of Value Indicators and Extraction of Key Management Points in Product Development using Bayesian Network and Hierarchical Clustering

Modeling of Value Indicators and Extraction of Key Management Points in Product Development using Bayesian Network and Hierarchical Clustering

Strong genetic structure among populations of the tick Ixodes ricinus across its range

Ixodes ricinus is the most common and widely distributed tick species in Europe, responsible for several zoonotic diseases, including Lyme borreliosis. Population genetics of disease vectors is a useful tool for understanding the spread of pathogens and infection risks. Despite the threat to the public health due to the climate-driven distribution changes of I. ricinus, the genetic structure of tick populations, though essential for understanding epidemiology, remains unclear. Previous studies have demonstrated weak to no apparent spatial pattern of genetic differentiation between European populations. Here, we analysed the population genetic structure of 497 individuals from 28 tick populations sampled from 20 countries across Europe, the Middle-East, and northern Africa. We analysed 125 SNPs loci after quality control. We ran Bayesian and multivariate hierarchical clustering analyses to identify and describe clusters of genetically related individuals. Both clustering methods support the identification of three spatially-structured clusters. Individuals from the south and north-western parts of Eurasia form a separated cluster from northern European populations, while central European populations are a mix between the two groups. Our findings have important implications for understanding the dispersal processes that shape the spread of zoonotic diseases under anthropogenic global changes.

A novel learning and prediction Bayesian hierarchical clustering-Dirichlet mixture model for effective data mining

Decision making and business support is an important process in data mining and this can be achieved by means of pattern classification and extraction. Since the huge volume of data needs starving knowledge to process and organisation faces many issues in solving those issues. Clustering is an effective technology available to analyse and convert the datasets into meaningful patterns. Clustering in data mining uses various attributes to compute large dataset and meet out the real time issues. The proposed model uses Bayesian hierarchical clustering model with Dirichlet model to resolve the issues in large dataset analysis. Experimental results prove that proposed model experience better clustering efficiency than conventional complete link agglomerative clustering by achieving 92% of clustering accuracy.

Genetic diversity analysis of sesame-A bayesian clustering approach

Diversity in plant genetic resources (PGR) provides opportunity for plant breeders to develop new and improved cultivars with desirable characteristics viz., high yield, pest and disease resistance, photosensitivity and high oil quality. Genetic diversity is a ubiquitous feature of all species in nature. Therefore, different genotypes of sesame were used for diversity analysis. Different clustering techniques were widely used for the analysis of diversity. In this paper, Bayesian hierarchical clustering algorithm is applied which can be interpreted as a novel fast bottom-up approximate inference method. Finally, this method clusters the genotypes into various groups with their corresponding genotypes in respective clusters.

Population genomic data reveal extreme geographic subdivision and novel conservation actions for the declining foothill yellow-legged frog.

Genomic data have the potential to inform high resolution landscape genetic and biological conservation studies that go far beyond recent mitochondrial and microsatellite analyses. We characterize the relationships of populations of the foothill yellow-legged frog, Rana boylii, a declining, "sentinel" species for stream ecosystems throughout its range in California and Oregon. We generated RADseq data and applied phylogenetic methods, hierarchical Bayesian clustering, PCA and population differentiation with admixture analyses to characterize spatial genetic structure across the species range. To facilitate direct comparison with previous analyses, we included many localities and individuals from our earlier work based on mitochondrial DNA. The results are striking, and emphasize the power of our landscape genomic approach. We recovered five extremely differentiated primary clades that indicate that R. boylii may be the most genetically differentiated anuran yet studied. Our results provide better resolution and more spatially consistent patterns than our earlier work, confirming the increased resolving power of genomic data compared to single-locus studies. Genomic structure is not equal across the species distribution. Approximately half the range of R. boylii consists of a single, relatively uniform population, while Sierra Nevada and coastal California clades are deeply, hierarchically substructured with biogeographic breaks observed in other codistributed taxa. Our results indicate that clades should serve as management units for R. boylii rather than previously suggested watershed boundaries, and that the near-extinct population from southwestern California is particularly diverged, exhibits the lowest genetic diversity, and is a critical conservation target for species recovery.

Population structure of caribou in an ice‐bound archipelago

AbstractAimArchipelagos provide ideal natural systems for testing the effects of isolation and fragmentation of habitats on the genetic makeup of populations—an important consideration, given that many insular species are of conservation concern. Two theories predominate: Island Biogeography Theory (IBT) posits that proximity to the mainland drives the potential for migrants and gene flow. The Central Marginal Hypothesis (CMH) predicts that island populations at the periphery of a species range may experience low gene flow, small population size and high rates of genetic drift. We investigated population genetic structure, genetic diversity and key drivers of diversity for Arctic island‐dwelling caribou (Rangifer tarandus). Our aim was to inform intraspecific units for conservation and decipher how IBT and CMH could act in an archipelago where isolation is highly variable due to sea ice and open water.LocationCanadian Arctic Archipelago, Canada (Latitude, 55–82°N; Longitude, 61–123°W).MethodsWe genotyped 447 caribou at 16 microsatellite loci; these caribou represented two subspecies (R. t. groenlandicus, R. t. pearyi) and three designatable units. We used hierarchical Bayesian clustering and ordination to determine genetic groups. We evaluated the influence of ecological and geographic variables on genetic diversity using linear mixed‐effects models and compared diversity among mainland and island herds.ResultsBayesian clustering revealed nine genetic clusters with differentiation among and within caribou subspecies. Genetic differentiation was explained predominantly by isolation‐by‐distance across all caribou, even at the scale of subspecies. Island caribou were less genetically diverse than mainland herds; individual heterozygosity was negatively correlated with distance‐to‐mainland and the extent of autumn ice‐free coastline and positively correlated with unglaciated island size.Main conclusionsOur findings underscore the importance of hierarchical analysis when investigating genetic population structure. Genetic diversity and its key drivers lend support to both IBT and CMH and highlight the pending threat of climate change for Arctic island caribou.

Range-Wide Snow Leopard Phylogeography Supports Three Subspecies.

The snow leopard, Panthera uncia, is an elusive high-altitude specialist that inhabits vast, inaccessible habitat across Asia. We conducted the first range-wide genetic assessment of snow leopards based on noninvasive scat surveys. Thirty-three microsatellites were genotyped and a total of 683 bp of mitochondrial DNA sequenced in 70 individuals. Snow leopards exhibited low genetic diversity at microsatellites (AN = 5.8, HO = 0.433, HE = 0.568), virtually no mtDNA variation, and underwent a bottleneck in the Holocene (∼8000 years ago) coinciding with increased temperatures, precipitation, and upward treeline shift in the Tibetan Plateau. Multiple analyses supported 3 primary genetic clusters: (1) Northern (the Altai region), (2) Central (core Himalaya and Tibetan Plateau), and (3) Western (Tian Shan, Pamir, trans-Himalaya regions). Accordingly, we recognize 3 subspecies, Panthera uncia irbis (Northern group), Panthera uncia uncia (Western group), and Panthera uncia uncioides (Central group) based upon genetic distinctness, low levels of admixture, unambiguous population assignment, and geographic separation. The patterns of variation were consistent with desert-basin "barrier effects" of the Gobi isolating the northern subspecies (Mongolia), and the trans-Himalaya dividing the central (Qinghai, Tibet, Bhutan, and Nepal) and western subspecies (India, Pakistan, Tajikistan, and Kyrgyzstan). Hierarchical Bayesian clustering analysis revealed additional subdivision into a minimum of 6 proposed management units: western Mongolia, southern Mongolia, Tian Shan, Pamir-Himalaya, Tibet-Himalaya, and Qinghai, with spatial autocorrelation suggesting potential connectivity by dispersing individuals up to ∼400 km. We provide a foundation for global conservation of snow leopard subspecies, and set the stage for in-depth landscape genetics and genomic studies.

Calculating the required cash in bank branches: a Bayesian-data mining approach

The issue of sufficiency of cash in bank branches is considered as an important issue especially for branch managers; because, not only the insufficiency of daily cash results in lack of response to needs of customers, but also may its excess result in increase in costs for banks. Hence, banks are always attempting to determine their required cash based on their daily operation. For this purpose, in this paper, 18 branches of a certain bank in a period of five months, due to diversity of the branches, have been classified by two methods of hierarchical clustering and Bayesian hierarchical clustering in similar clusters, and then by considering the results obtained from clustering, amounts of entered and consumed branch cash have been estimated by neural network (via classic and Bayesian approach), so that the cash required by branches can be calculated. The error criteria of the estimates show that calculations by applying Bayesian neural network method with considering Bayesian clustering have the least error compared to other methods.