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.     

Plant litter effects on soil nutrient availability and vegetation dynamics: changes that occur when annual grasses invade shrub-steppe communities

Changes in the quantity and quality of plant litter occur in many ecosystems as they are invaded by exotic species, which impact soil nutrient cycling and plant community composition. Such changes in sagebrush-steppe communities are occurring with invasion of annual grasses (AG) into a perennial grass (PG) dominated system. We conducted a 5-year litter manipulation study located in the northern Great Basin, USA. Springtime litter was partially or completely removed in three communities with differing levels of invasion (invaded, mixed, and native) to determine how litter removal and litter biomass affected plant-available soil N and plant community composition. Litter biomass (prior to the removal treatment) was negatively correlated with plant-available N in the invaded community, but was positively correlated in the native community. Plant-available N had greater intra- and inter-annual fluctuations in the invaded compared to the mixed or native communities, but was not generally affected by removal treatments. Litter removal had negative effects on AG cover during a warm/dry year and negative effects on PG cover during a cool/wet year in the mixed community. Overall, the effectiveness of springtime litter manipulations on plant-available N were limited and weather dependent, and only removal treatments >75 % had effects on the plant community. Our study demonstrates how communities invaded by AGs have significantly increased temporal variability in nutrient cycling, which may decrease ecosystem stability. Further, we found that the ecological impacts from litter manipulation on sagebrush communities were dependent on the extent of AG invasion, the timing of removal, and seasonal precipitation.     

TLR4 and TLR21 expression,MIF, IFN-β, MD-2, CD14 activation,and sIgA production in chickens administered with EFAL41 strain challenged with <Emphasis Type="Italic">Campylobacter jejuni</Emphasis>

The protective effect of Enterococcus faecium EFAL41 on chicken’s caecum in relation to the TLR (TLR4 and TLR21) activation and production of luminal IgA challenged with Campylobacter jejuni CCM6191 was assessed. The activation of MIF, IFN-β, MD-2 and CD14 was followed-up after bacterial infection. Day-old chicks (40) were divided into four groups (n = 10): control (C), E. faecium AL41 (EFAL41), C. jejuni (CJ) and combined E. faecium AL41+C. jejuni (EFAL41+CJ). Relative mRNA expression of TLR4, TLR21 and CD14 was upregulated in the probiotic strain and infected (combined) group on day 4 and 7 post infection (p.i.). The caecal relative MD-2 mRNA expression was upregulated on day 4 p.i. in the EFAL41+CJ and CJ groups. MIF and IFN-β reached the highest levels in the combined groups on day 7 p.i. The concentration of the sIgA in intestinal flush was upregulated in EFAL41+CJ group on day 4 p.i. The results demonstrated that E. faecium EFAL41 probiotic strain can modulate the TLRs expression and modify the activation of MIF, IFN-β, MD-2 and CD14 molecules in the chickens caecum challenged with C. jejuni CCM 6191. The counts of EFAL41 were sufficient and high, similarly the counts of enterococci in both, caecum and faeces but without reduction of Campylobacter counts.     

Cognitive network interactions and beta 2 coherence in processing non-target stimuli in visual oddball task

Spatiotemporal dynamics of event-related potentials (ERP) evoked by non-target stimuli in a visual oddball experiment and the presence of coherent oscillations in beta 2 frequency band of decomposed EEG records from peristimulus period were investigated by means of intracranial electrodes in humans. Twenty-one patients with medically intractable epilepsy participated in the study. The EEG signal was recorded using platinum electrodes implanted in several cortical and subcortical sites. Averaged 2 s EEG records were analyzed. Task-specific EEG changes were found in each patient, ERPs were derived from 92 electrodes used (96 % of possible cases). In the majority of analysed cases, ERPs were composed of several distinct components, and their duration was mostly longer than 1 s. The mean onset of the first ERP component was 158+/-132 ms after the stimulus (median 112 ms, minimum value 42 ms, maximum value 755 ms), and large variability of these onset times was found in all the investigated structures. Possible coherence between neural activities of remote brain sites was investigated by calculating running correlations between pairs of decomposed EEG records (alpha, beta 1, beta 2 frequency bands were used, total number of correlated pairs was 662 in each frequency band). The record pairs exhibiting highly correlated time segments represented 23 % of all the investigated pairs in alpha band, 7 % in beta 1 band, and 59 % in beta 2 band. In investigated 2 s record windows, such segments were distributed evenly, i.e. they were also found before the stimulus onset. In conclusion, the results have implicated the idea that a lot of recorded ERPs was more or less by-products of chance in spreading a signal within the neuronal network, and that their functional relevance was somewhat linked with the phenomenon of activity synchronization.     

Improving Structural Order for a High‐Performance Diketopyrrolopyrrole‐Based Polymer Solar Cell with a Thick Active Layer

A series of low‐bandgap polymers based on thienyl benzodithiophene (BDTT) and furan‐substituted diketopyrrolopyrrole (FDPP) units with different side chains are presented. By replacing the branched ethylhexyl group on the FDPP unit with linear side chains, the structural order and carrier mobility of the modified polymers are enhanced accordingly. The power conversion efficiencies (PCE) of single‐junction devices based on polymers of this series are improved from ≈5% to ≈7%. More importantly, devices made from the modified polymers show excellent photovoltaic performance with a thick active layer of up to 360 nm, a very promising characteristic for the industrial implementation of polymer solar cell devices.