Incidence, predictive factors, and outcome of lymphoma transformation in follicular lymphoma patients

PURPOSE: To assess the incidence of lymphoma transformation in the natural history of follicular lymphoma (FL) patients and the factors that are predictive of this event. PATIENTS AND METHODS: Two hundred twenty patients with FL treated in our institution between 1975 and 1990, with a median follow-up duration of 9 years, were included in this retrospective analysis. RESULTS: Transformation was proven by histology in 34 patients or by cytology in 13 patients and was considered as highly probable on clinical arguments in five patients for an overall incidence of 24%. The probability of transformation was 22% at 5 years and 31% at 10 years and tended to plateau after 6 years. Predictive factors for transformation were nonachievement of complete remission (CR) after initial therapy (P < 10(-4), low serum albumin level (< 35 g/L) (P = .001), and beta 2-microglobulin level greater than 3 mg/L (P = .02) at diagnosis. In a multiparametric analysis, only beta 2-microglobulin level retained prognostic significance for freedom-from-transformation (FFT) survival (P = .04). Transformation accounted for 44% of deaths and was associated with a poor outcome, with a median survival time of 7 months. CONCLUSION: Transformation is an early event in the course of the disease and is mainly observed in patients with known adverse prognostic factors or those who do not achieve CR after initial treatment. These findings may be useful to select follicular lymphoma patients for intensive therapeutic approaches.     

A hybrid filter/wrapper approach of feature selection using information theory

We focus on a hybrid approach of feature selection. We begin our analysis with a filter model, exploiting the geometrical information contained in the minimum spanning tree (MST) built on the learning set. This model exploits a statistical test of relative certainty gain, used in a forward selection algorithm. In the second part of the paper, we show that the MST can be replaced by the 1 nearest-neighbor graph without challenging the statistical framework. This leads to a feature selection algorithm belonging to a new category of hybrid models (filter-wrapper). Experimental results on readily available synthetic and natural domains are presented and discussed.     

Progress on collisionally pumped optical-field-ionization soft X-ray lasers

We present the status of optical field ionization soft X-ray lasers. The amplifying medium is generated by focusing a high-energy circularly polarized 30-fs 10-Hz Ti: sapphire laser system in a gaseous medium. Using xenon or krypton, strong laser emission at 41.8 and 32.8 nm, respectively, has been observed. After presenting the basis of the physics, we present recent characterization of the sources as well as dramatic improvement of their performances using the waveguiding technique.     

In bacterial reaction centers rapid delivery of the second proton to QB can be achieved in the absence of L212Glu

In the reaction center (RC) of Rhodobacter capsulatus, residue L212Glu is a component of the pathway for proton transfer to the reduced secondary quinone, QB. We isolated phenotypic revertants of the photosynthetically incompetent (PS-) L212Glu-->Gln mutant; all of them retain the L212Glu-->Gln substitution and carry a second-site mutation: L227Leu-->Phe, L228Gly-->Asp, L231Arg-->Cys, or M231Arg-->Cys. We also characterized the L212Ala strain, which is a phenotypic revertant of the PS- L212Glu-L213Asp-->Ala-Ala mutant. The activities of the RCs of these strains--all of which lack L212Glu--were studied by flash-induced absorption spectroscopy. At pH 7.5, the rate of second electron transfer in the L212Q mutant is comparable to the wild-type rate. However, this mutant shows a marked decrease in the rate of cytochrome oxidation under strong continuous illumination and a very slow phase (0.66 s-1) of the proton transfer kinetics following the second flash, indicating that transfer of the second proton to QB is slowed more than 1000-fold. The levels of recovery of the functional capabilities in the revertant RCs vary widely; their rates of cytochrome oxidation were intermediate between those of the wild-type and the L212Q mutant. The kinetics of proton transfer following the second flash show a significant recovery in the L212Q + M231C and L212A RCs (330-540 s-1), but the L212Q + L227F RCs recover this function only partially. Compensation for the lack of L212Glu in revertant RCs is discussed in terms of (i) conformational changes that could allow water molecules to approach closer to QB and/or (ii) the increase in the negative electrostatic environment and the resultant rise in the free energy level of QB- that is induced by the mutations. The stoichiometries of H+/QB- proton uptake below pH 7.5 in the L212Q mutant, the L212Q + M231C revertant, and the wild-type strains are essentially equivalent, suggesting that L212Glu is protonated at neutral pH in wild-type RCs. This is also supported by the P+QB- charge recombination data. Comparison of H+/QB- proton uptake data with those obtained previously for the stoichiometries of H+/QA- proton uptake [Miksovska, J., Maróti, P., Tandori, J., Schiffer, M., Hanson, D. K., Sebban, P. (1996) Biochemistry 35, 15411-15417] suggests that L212Glu is the key to the electrostatic and perhaps structural interaction between the two quinone sites.     

Sampled Gromov Wasserstein

Machine Learning - Optimal Transport (OT) has proven to be a powerful tool to compare probability distributions in machine learning, but dealing with probability measures lying in different spaces...     

On the analysis of adaptability in multi-source domain adaptation

In many real-world applications, it may be desirable to benefit from a classifier trained on a given source task from some largely annotated dataset in order to address a different but related target task for which only weakly labeled data are available. Domain adaptation (DA) is the framework which aims at leveraging the statistical similarities between the source and target distributions to learn well. Current theoretical results show that the efficiency of DA algorithms depends on (i) their capacity of minimizing the divergence between the source and target domains and (ii) the existence of a good hypothesis that commits few errors in both domains. While most of the work in DA has focused on new divergence measures, the second aspect, often modeled as the capability term, remains surprisingly under-investigated. In this paper, we show that the problem of the best joint hypothesis estimation can be reformulated using a Wasserstein distance-based error function in the context of multi-source DA. Based on this idea, we provide a theoretical analysis of the capability term and derive inequalities allowing us to estimate it from finite samples. We empirically illustrate the proposed idea on different data sets.

     

Learning stochastic edit distance: Application in handwritten character recognition

Many pattern recognition algorithms are based on the nearest-neighbour search and use the well-known edit distance, for which the primitive edit costs are usually fixed in advance. In this article, we aim at learning an unbiased stochastic edit distance in the form of a finite-state transducer from a corpus of (input, output) pairs of strings. Contrary to the other standard methods, which generally use the Expectation Maximisation algorithm, our algorithm learns a transducer independently on the marginal probability distribution of the input strings. Such an unbiased way to proceed requires to optimise the parameters of a conditional transducer instead of a joint one. We apply our new model in the context of handwritten digit recognition. We show, carrying out a large series of experiments, that it always outperforms the standard edit distance.     

Suppression of Mn segregation in Ge/Mn5Ge3 heterostructures induced by interstitial carbon

Mn₅Ge₃ compound, with its room-temperature ferromagnetism and possibility to epitaxially grow on Ge, acts as a potential spin injector into group-IV semiconductors. It is shown that the realization of Ge/Mn₅Ge₃ heterostructures is highly hampered by Mn segregation toward the Ge growing surface. The Mn segregation length can be estimated in-situ and in real time by means of reflection high-energy electron diffraction. We present here an approach allowing to greatly reduce or even to prevent the Mn segregation, whose principle is based on filling the Mn₅Ge₃ lattice with interstitial carbon atoms. In addition, we show that interstitial carbon in Mn₅Ge₃ allows to enhance not only the Curie temperature of Mn₅Ge₃C_x layers but also in the whole Ge/Mn₅Ge₃/Ge heterostructures.