Clustering more than two million biomedical publications: comparing the accuracies of nine text-based similarity approaches

Background

We investigate the accuracy of different similarity approaches for clustering over two million biomedical documents. Clustering large sets of text documents is important for a variety of information needs and applications such as collection management and navigation, summary and analysis. The few comparisons of clustering results from different similarity approaches have focused on small literature sets and have given conflicting results. Our study was designed to seek a robust answer to the question of which similarity approach would generate the most coherent clusters of a biomedical literature set of over two million documents.

Methodology

We used a corpus of 2.15 million recent (2004-2008) records from MEDLINE, and generated nine different document-document similarity matrices from information extracted from their bibliographic records, including titles, abstracts and subject headings. The nine approaches were comprised of five different analytical techniques with two data sources. The five analytical techniques are cosine similarity using term frequency-inverse document frequency vectors (tf-idf cosine), latent semantic analysis (LSA), topic modeling, and two Poisson-based language models – BM25 and PMRA (PubMed Related Articles). The two data sources were a) MeSH subject headings, and b) words from titles and abstracts. Each similarity matrix was filtered to keep the top-n highest similarities per document and then clustered using a combination of graph layout and average-link clustering. Cluster results from the nine similarity approaches were compared using (1) within-cluster textual coherence based on the Jensen-Shannon divergence, and (2) two concentration measures based on grant-to-article linkages indexed in MEDLINE.

Conclusions

PubMed''s own related article approach (PMRA) generated the most coherent and most concentrated cluster solution of the nine text-based similarity approaches tested, followed closely by the BM25 approach using titles and abstracts. Approaches using only MeSH subject headings were not competitive with those based on titles and abstracts.     

Spatial Patterns of Primate Electrocutions in Diani,Kenya

Electrocution from power infrastructure threatens many primate species, yet knowledge of effective evidence-based mitigation strategies is limited. Mitigation planning requires an understanding of the spatial distribution of electrocutions to prioritize high-risk areas. In Diani, a coastal Kenyan town, electrocution is an important cause of death for five primate species. In this study we aim to describe the spatial patterns of electrocutions and electric shock incidents (collectively referred to as electrocutions hereafter) and identify electrocution hotspots to guide an effective primate conservation approach in Diani. Colobus Conservation, a not-for-profit organization, has recorded electrocutions and annual primate census data since 1998. We georeferenced 329 electrocution data points and analyzed them using QGIS. We identified and compared hotspots across species, seasons, and time using kernel density estimation and Getis-Ord-Gi*. We employed spatial regression models to test whether primate population density and power line density predicted the location of electrocution hotspots. Electrocutions occurred in hotspots that showed little variation in location between species and seasons. The limited variation in hotspot location over time likely occurred as a result of new building development in Diani and variability in primate detection rates by community members. Primate density and power line density were significant predictors of electrocution density for Angolan black-and-white colobus (Colobus angolensis palliatus) and Sykes monkeys (Cercopithecus mitis albogularis), but the relationship was weak, suggesting the presence of additional risk factors. This study provides a framework for systematic spatial prioritization of power lines that can be used to reduce primate electrocutions in Diani, and can be adopted in other areas of the world where primates are at risk from electrocution.     

Conformational behavior and aggregation of alpha-synuclein in organic solvents: modeling the effects of membranes

Intracellular proteinaceous inclusions (Lewy bodies and Lewy neurites) of alpha-synuclein are pathological hallmarks of neurodegenerative diseases such as Parkinson's disease, dementia with Lewy bodies (DLB), and multiple systemic atrophy. The molecular mechanisms underlying the aggregation of alpha-synuclein into such filamentous inclusions remain unknown, although many factors have been implicated, including interactions with lipid membranes. To model the effects of membrane fields on alpha-synuclein, we analyzed the structural and fibrillation properties of this protein in mixtures of water with simple and fluorinated alcohols. All solvents that were studied induced folding of alpha-synuclein, with the common first stage being formation of a partially folded intermediate with an enhanced propensity to fibrillate. Protein fibrillation was completely inhibited due to formation of beta-structure-enriched oligomers with high concentrations of methanol, ethanol, and propanol and moderate concentrations of trifluoroethanol (TFE), or because of the appearance of a highly alpha-helical conformation at high TFE and hexafluoro-2-propanol concentrations. At least to some extent, these conformational effects mimic those observed in the presence of phospholipid vesicles, and can explain some of the observed effects of membranes on alpha-synuclein fibrillation.     

Primary infection of mice with high titer inoculum respiratory syncytial virus: characterization and response to antiviral therapy

Intranasal infection of BALB/c mice with respiratory syncytial virus (RSV)-A2 (0.5 x 10(8) - 2.0 x 10(8) plaque-forming units, PFU) produced disease characterized by weight loss (2-3 g) and mortality (60%-100%) with the mean day of death ranging from 6-7 d after infection. The extent of RSV disease was inoculum titer-dependent and required a replication competent virus. Lung titers of virus peaked at 0.5-1 x 10(6) PFU/g wet weight. Bronchoalveolar lavage fluid (BALF) levels of IL-1beta, TNF-alpha, INF-gamma IL-12, IL-6, MIP-1alpha, RANTES, and protein were elevated, whereas IL-2, IL-4, IL-5, IL-13, and IL-10 were unchanged. Histological assessment of lungs revealed marked inflammatory pathology characterized by bronchiolitis, vasculitis, and interstitial pneumonia. Whole-body plethysmography revealed significant disease-associated deficits of respiratory function. Therapy with ribavirin administered either by the intranasal, subcutaneous, or oral route significantly reduced disease in a dose-dependent manner. Delaying the initiation of therapy resulted in a loss of activity for ribavirin. Synagis administered either intramuscularly as a single dose in prophylaxis or intranasally in prophylaxis, followed by therapy, also significantly reduced disease in a dose-dependent manner. Infection of mice with a high titer inoculum of RSV-A2 resulted in severe and fatal pulmonary disease that was responsive to treatment. This model may be useful to characterize the in vivo activity of experimental therapies for RSV infection.     

rac regulates its effector phospholipase Cgamma2 through interaction with a split pleckstrin homology domain

Several isoforms of phospholipase C (PLC) are regulated through interactions with Ras superfamily GTPases, including Rac proteins. Interestingly, of two closely related PLCgamma isoforms, only PLCgamma(2) has previously been shown to be activated by Rac. Here, we explore the molecular basis of this interaction as well as the structural properties of PLCgamma(2) required for activation. Based on reconstitution experiments with isolated PLCgamma variants and Rac2, we show that an unusual pleckstrin homology (PH) domain, designated as the split PH domain (spPH), is both necessary and sufficient to effect activation of PLCgamma(2) by Rac2. We also demonstrate that Rac2 directly binds to PLCgamma(2) as well as to the isolated spPH of this isoform. Furthermore, through the use of NMR spectroscopy and mutational analysis, we determine the structure of spPH, define the structural features of spPH required for Rac interaction, and identify critical amino acid residues at the interaction interface. We further discuss parallels and differences between PLCgamma(1) and PLCgamma(2) and the implications of our findings for their respective signaling roles.