An analysis of entity normalization evaluation biases in specialized domains

BACKGROUND

Entity normalization is an important information extraction task which has recently gained attention, particularly in the clinical/biomedical and life science domains. On several datasets, state-of-the-art methods perform rather well on popular benchmarks. Yet, we argue that the task is far from resolved.

RESULTS

We have selected two gold standard corpora and two state-of-the-art methods to highlight some evaluation biases. We present non-exhaustive initial findings on the existence of evaluation problems of the entity normalization task.

CONCLUSIONS

Our analysis suggests better evaluation practices to support the methodological research in this field.

Experimental investigation of size broadening of a Kα x-ray source produced by high intensity laser pulses

The size of a hard Kα x-ray source ([Formula: see text] = 17.48 keV) produced by a high intensity femtosecond laser interacting with a solid molybdenum target is experimentally investigated for a wide range of laser intensity (I ~ 1017-2.8 × 1019 W/cm2) and for four values of the temporal contrast ratio (6.7 × 107 < CR < 3.3 × 1010). Results point out the size enlargement of the x-ray source with the increase of laser intensity and with the deterioration of temporal contrast. It amounts up to sixteen times the laser spot size at the highest laser intensity and for the lowest temporal contrast ratio. Using hydrodynamic simulations, we evaluate the density scale length of the pre-plasma L/λ just before the main pulse peak. This allows us to show that a direct correlation with the laser absorption mechanisms is not relevant to explain the large size broadening. By varying the thickness of the molybdenum target down to 4 µm, the impact of hot electron scattering inside the solid is also proved irrelevant to explain the evolution of both the x-ray source size and the Kα photon number. We deduce that the most probable mechanism yielding to the broadening of the source size is linked to the creation of surface electromagnetic fields which confine the hot electrons at the solid surface. This assumption is supported by dedicated experiments where the evolution of the size enlargement of the x-ray source is carefully studied as a function of the laser focal spot size for the highest contrast ratio.

[Cervical spondylolysis : a rare fortuitous finding]

A young patient consulted in our physical and rehabilitation medicine department following the onset of pain on the scapula area and at the base of his right upper limb after carrying a heavy load. After a couple of weeks, the patient also developed cervical pain. Fortuitously, the cervical scanner displayed a right C6 spondylolysis. Further evaluation by bone scan confirmed that this lesion was not recent and so probably not the cause of the symptoms. Cervical isthmic spondylolysis is a rare condition, much more common at the lumbar level and often ignored at the cervical one. The etiology, pahophysiology, imaging and treatment options for this cervical pathology are discussed in this article.

Exploring phase contrast imaging with a laser-based Kα x-ray source up to relativistic laser intensity

This study explores the ability of a hard Kα x-ray source (17.48 keV) produced by a 10 TW class laser system operated at high temporal contrast ratio and high repetition rate for phase contrast imaging. For demonstration, a parametric study based on a known object (PET films) shows clear evidence of feasibility of phase contrast imaging over a large range of laser intensity on target (from ~1017 W/cm2 to 7.0 × 1018 W/cm2). To highlight this result, a comparison of raw phase contrast and retrieved phase images of a biological object (a wasp) is done at different laser intensities below the relativistic intensity regime and up to 1.3 × 1019 W/cm2. This brings out attractive imaging strategies by selecting suitable laser intensity for optimizing either high spatial resolution and high quality of image or short acquisition time.

Improving the CONTES method for normalizing biomedical text entities with concepts from an ontology with (almost) no training data

Entity normalization, or entity linking in the general domain, is an information extraction task that aims to annotate/bind multiple words/expressions in raw text with semantic references, such as concepts of an ontology. An ontology consists minimally of a formally organized vocabulary or hierarchy of terms, which captures knowledge of a domain. Presently, machine-learning methods, often coupled with distributional representations, achieve good performance. However, these require large training datasets, which are not always available, especially for tasks in specialized domains. CONTES (CONcept-TErm System) is a supervised method that addresses entity normalization with ontology concepts using small training datasets. CONTES has some limitations, such as it does not scale well with very large ontologies, it tends to overgeneralize predictions, and it lacks valid representations for the out-of-vocabulary words. Here, we propose to assess different methods to reduce the dimensionality in the representation of the ontology. We also propose to calibrate parameters in order to make the predictions more accurate, and to address the problem of out-of-vocabulary words, with a specific method.

[Methotrexit, a HeteroGenious cleaning factory]

iGEM (pour international genetically engineered machine) est un concours international autour de la biologie synthétique réunissant des étudiants de toutes disciplines (mathématiques, physique, biologie, arts, etc.). « L’objectif est de construire un système biologique fonctionnel complexe, en assemblant des composants individuels moléculaires simples et standardisés (fragments d’ADN), appelés « briques biologiques » (biobriques), sorte de « legos » moléculaires, entreposés au MIT (Massachusetts Institute of Technology) (le registry of standard biological parts contient environ 20 000 biobriques). C’est une démarche proche de celle de l’ingénieur qui assemble des circuits électroniques ». En 2004, lors de sa création par le MIT (→), la compétition iGEM regroupait une quarantaine de projets ; 14 ans plus tard, elle accueille 350 équipes (6 000 étudiants, avec leurs instructeurs) issues des universités du monde entier. Elle culmine en un Giant Jamboree de quatre jours à Boston en novembre, au cours duquel les équipes présentent leur projet. Le « wiki » de la compétition (www.igem.org) présente l’ensemble des projets ainsi que le palmarès. Cette année, ont été décernées 114 médailles d’or, 68 d’argent et 107 de bronze. Neuf équipes françaises étaient engagées.

(→) Voir l’article de J. Peccoud et L. Coulombel, dont certains passages sont repris dans ce « chapo », m/s n° 5, mai 2007, page 551

High photon flux Kα Mo x-ray source driven by a multi-terawatt femtosecond laser at 100 Hz

We develop a pulsed hard x-ray K_α source at 17.4 keV produced by the interaction of a multi-terawatt peak power infrared femtosecond laser pulse with a thick molybdenum (Mo) target at a 100 Hz repetition rate. We measure the highest Mo K_α photon production reported to date corresponding to a K_α photon flux of 1×10¹¹  ph/(sr·s) and an estimated peak brightness of ∼2.5×10¹⁷  ph/(s·mm²·mrad²(0.1% bandwidth)) at ∼5×10¹⁸  W/cm² driving laser intensity.

Meet-U: Educating through research immersion

We present a new educational initiative called Meet-U that aims to train students for collaborative work in computational biology and to bridge the gap between education and research. Meet-U mimics the setup of collaborative research projects and takes advantage of the most popular tools for collaborative work and of cloud computing. Students are grouped in teams of 4–5 people and have to realize a project from A to Z that answers a challenging question in biology. Meet-U promotes "coopetition," as the students collaborate within and across the teams and are also in competition with each other to develop the best final product. Meet-U fosters interactions between different actors of education and research through the organization of a meeting day, open to everyone, where the students present their work to a jury of researchers and jury members give research seminars. This very unique combination of education and research is strongly motivating for the students and provides a formidable opportunity for a scientific community to unite and increase its visibility. We report on our experience with Meet-U in two French universities with master’s students in bioinformatics and modeling, with protein–protein docking as the subject of the course. Meet-U is easy to implement and can be straightforwardly transferred to other fields and/or universities. All the information and data are available at www.meet-u.org.

Impact of the pulse contrast ratio on molybdenum Kα generation by ultrahigh intensity femtosecond laser solid interaction

We present an extended experimental study of the absolute yield of Kα x-ray source (17.48 keV) produced by interaction of an ultrahigh intensity femtosecond laser with solid Mo target for temporal contrast ratios in the range of 1.7 × 107-3.3 × 109 and on three decades of intensity 1016-1019  W/cm². We demonstrate that for intensity I ≥ 2 × 1018  W/cm² Kα x-ray emission is independent of the value of contrast ratio. In addition, no saturation of the Kα photon number is measured and a value of ~2 × 1010 photons/sr/s is obtained at 10 Hz and I ~1019  W/cm². Furthermore, Kα energy conversion efficiency reaches the same high plateau equal to ~2 × 10-4 at I = 1019  W/cm² for all the studied contrast ratios. This original result suggests that relativistic J × B heating becomes dominant in these operating conditions which is supposed to be insensitive to the electron density gradient scale length L/λ. Finally, an additional experimental study performed by changing the angle of incidence of the laser beam onto the solid target highlights a clear signature of the interplay between collisionless absorption mechanisms depending on the contrast ratio and intensity.

The bacterial interlocked process ONtology (BiPON): a systemic multi-scale unified representation of biological processes in prokaryotes

BACKGROUND

High-throughput technologies produce huge amounts of heterogeneous biological data at all cellular levels. Structuring these data together with biological knowledge is a critical issue in biology and requires integrative tools and methods such as bio-ontologies to extract and share valuable information. In parallel, the development of recent whole-cell models using a systemic cell description opened alternatives for data integration. Integrating a systemic cell description within a bio-ontology would help to progress in whole-cell data integration and modeling synergistically.

RESULTS

We present BiPON, an ontology integrating a multi-scale systemic representation of bacterial cellular processes. BiPON consists in of two sub-ontologies, bioBiPON and modelBiPON. bioBiPON organizes the systemic description of biological information while modelBiPON describes the mathematical models (including parameters) associated with biological processes. bioBiPON and modelBiPON are related using bridge rules on classes during automatic reasoning. Biological processes are thus automatically related to mathematical models. 37% of BiPON classes stem from different well-established bio-ontologies, while the others have been manually defined and curated. Currently, BiPON integrates the main processes involved in bacterial gene expression processes.

CONCLUSIONS

BiPON is a proof of concept of the way to combine formally systems biology and bio-ontology. The knowledge formalization is highly flexible and generic. Most of the known cellular processes, new participants or new mathematical models could be inserted in BiPON. Altogether, BiPON opens up promising perspectives for knowledge integration and sharing and can be used by biologists, systems and computational biologists, and the emerging community of whole-cell modeling.

Chronic fatigue syndrome and sleep disorders: clinical associations and diagnostic difficulties

INTRODUCTION

Chronic fatigue syndrome (CFS) is characterised by the presence of intractable fatigue and non-restorative sleep, symptoms which are also very prevalent in multiple diseases and appear as side effects of different drugs. Numerous studies have shown a high prevalence of sleep disorders in patients with CFS. However, non-restorative sleep and fatigue are frequently symptoms of the sleep disorders themselves, so primary sleep disorders have to be ruled out in many cases of CFS.

DEVELOPMENT

This review was performed using a structured search of the MeSH terms ([Sleep]+[Chronic fatigue syndrome]) in the PubMed database.

CONCLUSION

Identifying primary sleep disorders in patients meeting diagnostic criteria for CFS will allow for a more comprehensive treatment approach involving new diagnostic and therapeutic strategies that may improve quality of life for these patients.

Two-Dimensional Frequency Resolved Optomolecular Gating of High-Order Harmonic Generation

Probing electronic wave functions of polyatomic molecules is one of the major challenges in high-harmonic spectroscopy. The extremely nonlinear nature of the laser-molecule interaction couples the multiple degrees of freedom of the probed system. We combine two-dimensional control of the electron trajectories and vibrational control of the molecules to disentangle the two main steps in high-harmonic generation-ionization and recombination. We introduce a new measurement scheme, frequency-resolved optomolecular gating, which resolves the temporal amplitude and phase of the harmonic emission from excited molecules. Focusing on the study of vibrational motion in N_{2}O_{4}, we show that such advanced schemes provide a unique insight into the structural and dynamical properties of the underlying mechanism.

Multi-channel electronic and vibrational dynamics in polyatomic resonant high-order harmonic generation

High-order harmonic generation in polyatomic molecules generally involves multiple channels of ionization. Their relative contribution can be strongly influenced by the presence of resonances, whose assignment remains a major challenge for high-harmonic spectroscopy. Here we present a multi-modal approach for the investigation of unaligned polyatomic molecules, using SF6 as an example. We combine methods from extreme-ultraviolet spectroscopy, above-threshold ionization and attosecond metrology. Fragment-resolved above-threshold ionization measurements reveal that strong-field ionization opens at least three channels. A shape resonance in one of them is found to dominate the signal in the 20-26 eV range. This resonance induces a phase jump in the harmonic emission, a switch in the polarization state and different dynamical responses to molecular vibrations. This study demonstrates a method for extending high-harmonic spectroscopy to polyatomic molecules, where complex attosecond dynamics are expected.

Brainstem auditory and somatosensory evoked potentials: a methodological study to evaluate the reproducibility of two devices

We aimed to determine whether statistical significant differences exist between the sets of results obtained from two devices used in our department for measuring brainstem auditory evoked potentials (BAEPs) and somatosensory evoked potentials (SEPs). We obtained BAEP and median and posterior tibial nerve SEP values bilaterally in ten healthy subjects. The tests were performed on the same subject using two devices consecutively. The equipment consisted of a Nicolet Viking-IV (Nicolet, Madison, WI, USA) and a Viking Select (Viasys Healthcare, Madison, WI, USA), and the same recording electrodes and stimulator (auditory and electrical) were used without modifying any postural position of the subject. The stimuli and recording parameters were the same for both devices. We obtained 20 sets of data for each type of test. The Bland–Altman plots as well as the one-sample t-test or Wilcoxon signed rank test were used to compare data between the two groups of data sets. We found no significant differences between the sets of values obtained with the two devices. Our analysis indicates that the two devices are equal in recording all different variables of BAEP and SEP, which allows us to combine the BAEP and SEP data obtained from the two devices for follow-up studies involving quantitative statistical methods. This study received institutional approval (protocol number PRAG-154/2013).

Inhomogeneous high harmonic generation in krypton clusters

High order harmonic generation from clusters is a controversial topic: conflicting theories exist, with different explanations for similar experimental observations. From an experimental point of view, separating the contributions from monomers and clusters is challenging. By performing a spectrally and spatially resolved study in a controlled mixture of clusters and monomers, we are able to isolate a region of the spectrum where the emission purely originates from clusters. Surprisingly, the emission from clusters is depolarized, which is the signature of statistical inhomogeneous emission from a low-density source. The harmonic response to laser ellipticity shows that this generation is produced by a new recollisional mechanism, which opens the way to future theoretical studies.

[The role of atypical microorganisms and viruses in severe acute community-acquired pneumonia]

Usually, intensivists do not focus on atypical bacteria and viruses in severe community-acquired pneumonia (CAP). Only Legionella pneumophila and influenza virus, following the recent H1N1 influenza pandemic, are routinely suggested as responsible agents. However, CAP due to atypical bacteria may represent up to 44% of all CAP. Viral CAP is considered less severe than the usual bacterial ones, although 25% of them warrant hospitalization and 15% result in severe sepsis. Even though L. pneumophila is the most frequently atypical pathogen involved in severe cases, Mycoplasma pneumoniae may be responsible for multiorgan failure. To date, tools including detection of Legionella antigen in urine and Mycoplasma using polymerase chain reaction (PCR) allow rapid and accurate diagnosis. The treatment is based on macrolides and fluoroquinolones that can be associated in severe Legionnaire diseases. The presence of virus in CAP, either alone or in association with bacteria, has been demonstrated using molecular biology tests. These techniques also allowed the identification of several new viruses in CAP. However, the exact role of these detected viruses in CAP as well as the efficiency of antiviral therapy still represent major unsolved concerns.

Prediction of fluid responsiveness by a continuous non-invasive assessment of arterial pressure in critically ill patients: comparison with four other dynamic indices

BACKGROUND

We evaluated the ability of an infrared photoplethysmography arterial waveform (continuous non-invasive arterial pressure, CNAP) to estimate arterial pulse pressure variation (PPV). We compared the ability of non-invasive PPV to predict fluid responsiveness with invasive PPV, respiratory variation of pulse contour-derived stroke volume, and changes in cardiac index induced by passive leg raising (PLR) and end-expiratory occlusion (EEO) tests.

METHODS

We measured the responses of cardiac index (PiCCO) to 500 ml of saline in 47 critically ill patients with haemodynamic failure. Before fluid administration, we recorded non-invasive and invasive PPVs, stroke volume variation, and changes in cardiac index induced by PLR and by 15 s EEO. Logistic regressions were performed to investigate the advantage of combining invasive PPV, stroke volume variation, PLR, and EEO when predicting fluid responsiveness.

RESULTS

In eight patients, CNAP could not record arterial pressure. In the 39 remaining patients, fluid increased cardiac index by ≥15% in 17 'responders'. Considering the 195 pairs of measurements, the bias (sd) between invasive and non-invasive PPVs was -0.6 (2.3)%. The areas under the receiver operating characteristic (ROC) curves for predicting fluid responsiveness were 0.89 (95% confidence interval, 0.78-1.01) for non-invasive PPV compared with 0.89 (0.77-1.01), 0.84 (0.70-0.96), 0.95 (0.88-1.03), and 0.97 (0.91-1.03) for invasive pulse pressure, stroke volume variations, PLR, and EEO tests (no significant difference). Combining multiple tests did not significantly improve the area under the ROC curves.

CONCLUSIONS

Non-invasive assessment of PPV seems valuable in predicting fluid responsiveness.

[Inhalation devices: characteristics, modeling, regulation and use in routine practice. GAT Aerosolstorming, Paris 2011]

Aerosoltherapy is a first-line treatment for chronic obstructive respiratory diseases such as asthma and COPD. Treatment modalities and devices are varied and the choice of the device must be adapted to and optimized for every patient. Spacers can be used for some categories of patients for whom the use of other devices turns out to be complicated. The improvement of these treatments requires the optimization of the lung deposition of inhaled particles; lung modeling plays an essential role in the understanding of the mechanisms of flow in the airways. Regulations must frame prescription of inhaled treatments to optimize its quality and, thus, the care for these chronic diseases. Many generally-accepted ideas concerning these treatments turn out to be false. Inhaled treatments are constantly evolving, both pharmacologically and technologically.

[Arterial and venous thrombosis in lung cancer]

We report the case of a 61-year old man in whom a deep venous thrombosis was the presenting feature of disseminated lung carcinoma. A few days later, an arterial thrombosis occurred necessitating amputation. Within a few weeks, the lung cancer progressed dramatically and the patient died. While the association between venous thrombosis and cancer is well known, the relationship between cancer and arterial thrombosis has been less explored. This observation allows discussion of the pathophysiological and clinical aspects of this association, as well as the implications for patient care.

[Utilization of evoked potentials in intensive care units]

Evoked Potentials (EP) are a neurophysiological tool that makes it possible for us to make an extensive study of the cerebral cortex, the brainstem and the spinal cord. Different techniques can be applied while performing the EPs, such as Visual Evoked Potentials (VEPs), Somatosensory Evoked Potentials (SSEPs), Long Latency Somatosensory Evoked Potentials (LLSEPs), Brainstem Auditory Evoked Potentials (BAEPs), Middle Latency Auditory Evoked Potentials (MLAEPs), Long Latency Auditory Evoked Potentials (LLAEP) and Mismatch Negativity (MMN). The combination of the different techniques of Evoked Potentials (EP) allows us to make a neurofunctional evaluation of comatose patients in the Intensive Care Units (ICU). It is also a useful tool in the diagnosis of the origin of coma, to confirm brain death, and as an evolutive prognosis value of the different central nervous system diseases (SNC). There are also studies that propose using EPs as a monitoring tool of the SNC. We present an up-dated review on the principal aspects of EP neuromonitoring in Intensive Care Unit (ICU) patients.

[Cyclic alternating pattern as a sign of brain instability during sleep]

During the last 20 years, the fact that the role of microarousal arousal is very involved in the pathophysiology of sleep disorders has been widely demonstrated. However, in spite of this, the nature of microarousal is still under debate. The awakening system forms a part of the relationship between activation system and sleep maintenance, thus providing the natural evolution of sleep and defending it from the stimuli of the external world. The awakening system, that includes the concept of microarousal, can appear isolated or periodically, then receiving the name of cyclic alternating pattern (CAP). CAP may be affected by different factors that can alter its periodicity, thus offering quantitative information on sleep instability. It can also serve as another new tool to understand human sleep and as complementary information supplied by the sleep macrostructure and microarousal study. This study aims to explain the concept of CAP and its pathophysiology studied up to the present.

Pseudomonas aeruginosa bronchopulmonary infection in patients with AIDS, with emphasis on relapsing infection

The aim of this study was to delineate the clinical and therapeutic characteristics of Pseudomonas aeruginosa bronchopulmonary infection in acquired immunodeficiency syndrome (AIDS) patients. Eighteen AIDS patients had 39 episodes of P. aeruginosa bronchopulmonary infection. Their mean CD4 cell count was 0.012+/-0.011 cells x 10(9) x L(-1) and two episodes (5.1%) occurred in neutropenic patients. Ten patients (55.5%) had 21 outbreaks of pseudomonal infection. Relapses were more frequent in patients with chronic bronchitis (80 versus 0%, p=0.03) and in those who received initial oral antibiotic therapy (100 versus 55.6%, p=0.25). Three patients died, but death was directly related to pseudomonal infection in only one patient. In a case-control study, patients with bronchopulmonary P. aeruginosa infection had a survival comparable to patients in the control group. Immunoglobulin prophylaxis was administered to three patients with relapses, without success. The two patients who had P. aeruginosa eradicated were those who began triple antiretroviral therapy and had a CD4 cell increase >0.150 cells x 10(9) x L(-1). Relapsing Pseudomonas aeruginosa bronchopulmonary infection affects patients with advanced human immunodeficiency virus infection, prior underlying lung disease, chronic bronchitis and initial oral antibiotic therapy. Immune reconstitution through triple antiretroviral therapy succeeded in eradicating Pseudomonas aeruginosa respiratory infection in two patients.

[Invasive pulmonary aspergillosis: a study of 33 cases]

Invasive pulmonary aspergillosis (IPA) is an infectious complication appearing mainly in immunosuppressed patients, whose diagnosis is often difficult and lately made, and that usually bears a dismal prognosis. Patients diagnosed as having IPA from 1989 to 1994 were retrospectively analyzed. Probable IPA was diagnosed on the basis of a positive culture for Aspergillus together with a consistent radiological image. Confirmed IPA was diagnosed if there was, in addition to the former, a pathological examination showing Aspergillus hifae invading pulmonary parenchyma and/or pulmonary vessels. There were 25 men and 8 women with a mean age of 53.7 +/- 16.9 years (range: 22-86 years). IPA was confirmed in 11 cases and probable in 22. Sixty three percent of the patients had hematologic malignancy or solid cancer, whereas 30.3% did not have prior granulocytopenia or immunosuppressive therapy. The mean (SD) interval between admission and diagnosis was 40.2 (37.1) days (range: 1-180 days), and the diagnosis was made while the patient was still alive in 75% of the cases. Fifteen percent of the patients had extrapulmonary aspergillosis. The most frequent finding both on X-ray film of the chest and pulmonary computed tomography were bilateral multiple pulmonary nodules. Thirteen patients were treated with itraconazole, 6 with amphotericin B, 5 received both drugs, and 2 received fluconazole. Nineteen patients (57.6%) died and the case-fatality rate among treated patients was 46.1%. IPA presents mainly in immunosuppressed patients, but there was a not negligible proportion of patients lacking the classical risk factors. IPA is often a lately made diagnosis and in a quarter of the patients it is not made when the patient is alive. The most frequent radiological presentation are multiple bilateral nodules. The case-fatality rate of IPA is exceedingly high, even when if the patient has been adequately treated.

Receptors for anti-müllerian hormone on Leydig cells are responsible for its effects on steroidogenesis and cell differentiation

Strong overexpression of anti-Müllerian hormone (AMH) in transgenic mice leads to incomplete fetal virilization and decreased serum testosterone in the adult. Conversely, AMH-deficient mice exhibit Leydig cell hyperplasia. To probe the mechanism of action of AMH on Leydig cell steroidogenesis, we have studied the expression of mRNA for steroidogenic proteins in vivo and in vitro and performed a morphometric analysis of testicular tissue in mice overexpressing the hormone. We show that overexpression of AMH in male transgenic mice blocks the differentiation of Leydig cell precursors. Expression of steroidogenic protein mRNAs, mainly cytochrome P450 17 alpha-hydroxylase/C17-20 lyase (P450c17), is decreased in transgenic mice overexpressing AMH and in AMH-treated purified Leydig cells. In contrast, transgenic mice in whom the AMH locus has been disrupted show increase expression of P450c17. In vitro, but not in vivo, AMH also decreases the expression of the luteinizing hormone receptor. The effect of AMH is explained by the presence of its receptor on Leydig cells. Our results provide insight into the action of AMH as a negative modulator of Leydig cell differentiation and function.