Influences of porcine circo-, arteriviruses and cathelicidins in plasmacytoid dendritic cell-derived interferon-alpha responses

Type I interferons (IFNs), mainly IFN-α/β play a crucial role in innate defense against viruses. In addition to their direct antiviral activity, type I IFNs have antitumoral and immunomodulatory effects. Although all cells are virtually able to induce IFN-α, the plasmacytoid dendritic cell (pDC) subset represents the ultimate producers of IFN-α as well as other proinflammatory cytokines. Due to the specific expression of TLR7 and TLR9 recognizing single-stranded (ss) RNA and unmethylated CpG motifs respectively, pDCs can secrete up to 1000 times more IFN-α than any cellular types. Additionally, it is well known that several cytokines including type I and II IFNs, Flt3-L, IL-4 and GM-CSF favor pDC-derived IFN-α responses to unmethylated CpG motifs. In a first step, we aimed to characterize and clarify the interactions of two porcine viruses with pDCs. The double-stranded DNA replicative forms of porcine circovirus type 2 (PCV2) were demonstrated to inhibit CpG-induced IFN- α by pDCs. Our study showed that none of the cytokines known to enhance pDC responsiveness can counter-regulate the PCV2-mediated inhibition of IFN-α induced by CpG, albeit IFN-γ significantly reduced the level of inhibition. Interestingly, the presence of IFN-γ enabled pDCs to induce IFN-α to low doses of PCV2. We also noted that after DNase treatment, PCV2 preparations were still able to stimulate pDCs. These data suggest that encapsulated viral ssDNA promotes the induction of IFN-α in pDCs treated with IFN-γ whereas free DNA, presumably as double-stranded forms, was responsible for inhibiting pDC responses. Regarding PRRSV, it has been reported that North American isolates did not induce and even inhibited IFN-α response in pDCs. However, PRRSV infection was also shown to lead to an induction of IFN-α in the serum and in the lungs suggesting that certain cells are responsive to the virus. Contrasting to previous reports we found that numerous PRRSV isolates directly induced IFN-α in pDCs. This response was still observed after UV-inactivation of viruses and required TLR7 signaling. The inhibition of CpG-induced IFN-α was weak and strain dependent, again contrasting with a previous report. We also observed that IFN-γ and IL-4 enhanced IFN-α response to two prototype strains, VR-2332 and LVP23. In summary, we demonstrated that both PCV2 and PRRSV promote IFN-α secretion in pDCs in vitro suggesting that IFN-α detected in PCV2- or PRRSV-infected animal might originate from pDCs. On the other hand, PRRSV replication is restricted to the macrophage (MΦ) lineage. These innate immune cells represent a heterogeneous population which can be induce to “classical” (M1) and “alternative” (M2) activated MΦ acquiring inflammatory or “wound-healing” functional properties, respectively. Nonetheless, little is known about the effect of polarization into M1 or M2 and the susceptibility of these cells to PRRSV. Thus, we examined the impact of cytokine on MΦ polarization into M1 or M2. Infections of these cells by several PRRSV isolates enabled the discrimination of PRRSV isolate in a genotype- and irulencedependent manner in M1 and IFN-β-activated MΦ. In contrast, the expression of PRRSV nucleocapsid in M2 or inactivated MΦ was indistinguishable among the PRRSV isolates tested. In the last part of my Thesis, we investigated the influence of three synthetic porcine cathelicidin peptides for their ability to deliver nucleic acid to pDCs. We reported that all cathelicidins tested can complex and quickly deliver nucleic acids resulting in IFN-α induction. Moreover, we show that the typical α- helical amphipathic conformation is required to mediate killing of bacteria but not for inducing IFN-α secretion by pDCs. Furthermore, we found that E.coli treated with one of these cathelicidins is able to induce significantly higher levels of IFN-α compared to a non-sense version of the peptide. These data suggest that cathelicidins could influence the immune response in a two-step process. First, these peptides target bacteria leading to cell lysis. In turn, cathelicidins form complexes and deliver extracellular microbial nucleic acids released into pDCs. These pDC-derived IFN-α responses could be of particular relevance in driving the adaptive immune responses against microbial infections.

Virulence and genotype-associated infectivity of interferon-treated macrophages by porcine reproductive and respiratory syndrome viruses.

The polarization into M1 and M2 macrophages (MΦ) is essential to understand MΦ function. Consequently, the aim of this study was to determine the impact of IFN-γ (M1), IL-4 (M2) and IFN-β activation of MΦ on the susceptibility to genotype 1 and 2 porcine reproductive respiratory syndrome (PRRS) virus (PRRSV) strains varying in virulence. To this end, monocyte-derived MΦ were generated by culture during 72h and polarization was induced for another 24h by addition of IFN-γ, IL-4 or IFN-β. MΦ were infected with a collection of PRRSV isolates belonging to genotype 1 and genotype 2. Undifferentiated and M2 MΦ were highly susceptible to all PRRSV isolates. In contrast, M1 and IFN-β activated MΦ were resistant to low pathogenic genotype 1 PRRSV but not or only partially to genotype 2 PRRSV strains. Interestingly, highly virulent PRRSV isolates of both genotypes showed particularly high levels of infection compared with the prototype viruses in both M1 and IFN-β-treated MΦ (P<0.05). This was seen at the level of nucleocapsid expression, viral titres and virus-induced cell death. In conclusion, by using IFN-γ and IFN-β stimulated MΦ it is possible to discriminate between PRRSV varying in genotype and virulence. Genotype 2 PRRSV strains are more efficient at escaping the intrinsic antiviral effects induced by type I and II IFNs. Our in vitro model will help to identify viral genetic elements responsible for virulence, an information important not only to understand PRRS pathogenesis but also for a rational vaccine design. Our results also suggest that monocyte-derived MΦ can be used as a PRRSV infection model instead of alveolar MΦ, avoiding the killing of pigs.

Prolonged activity of the pestiviral RNase Erns as an interferon antagonist after uptake by clathrin-mediated endocytosis

The RNase activity of the envelope glycoprotein E(rns) of the pestivirus bovine viral diarrhea virus (BVDV) is required to block type I interferon (IFN) synthesis induced by single-stranded RNA (ssRNA) and double-stranded RNA (dsRNA) in bovine cells. Due to the presence of an unusual membrane anchor at its C terminus, a significant portion of E(rns) is also secreted. In addition, a binding site for cell surface glycosaminoglycans is located within the C-terminal region of E(rns). Here, we show that the activity of soluble E(rns) as an IFN antagonist is not restricted to bovine cells. Extracellularly applied E(rns) protein bound to cell surface glycosaminoglycans and was internalized into the cells within 1 h of incubation by an energy-dependent mechanism that could be blocked by inhibitors of clathrin-dependent endocytosis. E(rns) mutants that lacked the C-terminal membrane anchor retained RNase activity but lost most of their intracellular activity as an IFN antagonist. Surprisingly, once taken up into the cells, E(rns) remained active and blocked dsRNA-induced IFN synthesis for several days. Thus, we propose that E(rns) acts as an enzymatically active decoy receptor that degrades extracellularly added viral RNA mainly in endolysosomal compartments that might otherwise activate intracellular pattern recognition receptors (PRRs) in order to maintain a state of innate immunotolerance. IMPORTANCE The pestiviral RNase E(rns) was previously shown to inhibit viral ssRNA- and dsRNA-induced interferon (IFN) synthesis. However, the localization of E(rns) at or inside the cells, its species specificity, and its mechanism of interaction with cell membranes in order to block the host's innate immune response are still largely unknown. Here, we provide strong evidence that the pestiviral RNase E(rns) is taken up within minutes by clathrin-mediated endocytosis and that this uptake is mostly dependent on the glycosaminoglycan binding site located within the C-terminal end of the protein. Remarkably, the inhibitory activity of E(rns) remains for several days, indicating the very potent and prolonged effect of a viral IFN antagonist. This novel mechanism of an enzymatically active decoy receptor that degrades a major viral pathogen-associated molecular pattern (PAMP) might be required to efficiently maintain innate and, thus, also adaptive immunotolerance, and it might well be relevant beyond the bovine species.

Pestiviral E(rns) blocks TLR-3-dependent IFN synthesis by LL37 complexed RNA

The ribonuclease activity of the soluble glycoprotein E(rns) of pestiviruses represents a unique mechanism to circumvent the host's innate immune system by blocking interferon type-I synthesis in response to extracellularly added single- (ss) and double-stranded (ds) RNA. However, the reason why pestiviruses encode a ribonuclease in addition to the abundant serum RNases remained elusive. Here, we show that the 5' UTR and NS5B regions of various strains of the RNA genome of the pestivirus bovine viral diarrhea virus (BVDV) are resistant to serum RNases and are potent TLR-3 agonists. Inhibitory activity of E(rns) was restricted to cleavable RNA products, and did not extend to the synthetic TLR-7/8 agonist R-848. RNA complexed with the antimicrobial peptide LL37 was protected from degradation by E(rns)in vitro but was fully inhibited by E(rns) in its ability to induce IFN in cell cultures, suggesting that the viral protein is mainly active in cleaving RNA in an intracellular compartment. We propose that secreted E(rns) represents a potent IFN antagonist, which degrades viral RNA that is resistant to the ubiquitous host RNases in the extracellular space. Thus, the viral RNase prevents its own pathogen-associated molecular pattern (PAMP) to inadvertently activate the IFN response that might break innate immunotolerance required for persistent pestivirus infections.

Is The Allergen Really Needed in Allergy Immunotherapy?

Immunotherapy for type I allergies is well established and is regarded to be the most efficient treatment option besides allergen avoidance. As of today, different forms of allergen preparations are used in this regard, as well as different routes of application. Virus-like particles (VLPs) represent a potent vaccine platform with proven immunogenicity and clinical efficacy. The addition of toll-like receptor ligands and/or depot-forming adjuvants further enhances activation of innate as well as adaptive immune responses. CpG motifs represent intensively investigated and potent direct stimulators of plasmacytoid dendritic cells and B cells, while T cell responses are enhanced indirectly through increased antigen presentation and cytokine release. This article will focus on the function of VLPs loaded with DNA rich in nonmethylated CG motifs (CpGs) and the clinical experience gained in the treatment of allergic rhinitis, demonstrating clinical efficacy also if administered without allergens. Several published studies have demonstrated a beneficial impact on allergic symptoms by treatment with CpG-loaded VLPs. Subcutaneous injection of VLPs loaded with CpGs was tested with or without the adjuvant alum in the presence or absence of an allergen. The results encourage further investigation of VLPs and CpG motifs in immunotherapy, either as a stand-alone product or as adjuvants for allergen-specific immunotherapy.

Retraction

Vareille M, Kieninger E, Alves MP, et al. Impaired type I and type III interferon induction and rhinovirus control in human cystic fibrosis airway epithelial cells. Thorax 2012;67:517-25. This article has been retracted. In our article recently published in Thorax, we described a novel mechanism explaining the increased susceptibility of patients with cystic fibrosis (CF) to rhinovirus infections, namely defective interferon type I and III production by CF airway epithelial cells. In experiments performed after publication of the article we were unable to consistently replicate our findings of deficient interferon type I and III production by CF airway epithelial cells upon rhinovirus infection. In the light of these results, we carried out detailed investigations of the data reported in the above manuscript and regrettably found evidence of deliberate manipulation of experimental data by the first author Dr M. Vareille. This manipulation was accompanied in some instances by absence of original data files. The manipulation/original data absence involved data presented in most, if not all of the figures, thus we wish to fully retract the paper and apologise to the readers of Thorax and to the scientific community for the inconvenience this has caused. We also checked data published by our group in manuscripts on which Dr Vareille was a co-author and found that data published in these manuscripts had not been manipulated. These two manuscripts, whose data and conclusions we stand by are: Edwards MR, Regamey N, Vareille M, et al. Impaired innate interferon induction in severe therapy resistant atopic asthmatic children. Mucosal Immunol 2013;6:797–806. doi: 10.1038/mi.2012.118. and Kieninger E, Vareille M, Kopf BS, et al. Lack of an exaggerated inflammatory response on virus infection in cystic fibrosis. Eur Respir J 2012;39:297–304. doi: 10.1183/09031936.00054511. Dr. Vareille has received a letter from the Secretary General of the University of Bern condemning her scientific misconduct as a severe offence against the rules of scientific integrity. Her current employers have also been informed. All co-authors of the publication including Dr. Vareille concur with the retraction statement.

Tolerance of allografts across MHC barriers by allochimeric class I MHC proteins

Class I MHC proteins have been shown to induce accelerated rejection or prolong survival of allografts in various experimental models. These immunological effects have been attributed to the highly polymorphic alpha helical regions of the extracellular portions of the class I MHC molecule. The present experiments were designed to elucidate the immunomodulatory effects of these polymorphic regions and delineate the mechanisms involved. Soluble allochimeric class I MHC proteins were produced by substituting the PVG class I MHC RT1.Ac amino acid residues within the a 1 helical region with those of the donor BN ( a 1hn-RT1.Ac), the a 2 helical region of BN ( a 2hn-RT1.Ac), and both the a 1 and a 2 helical regions (RT1.An). The class I MHC proteins were produced in an E. coli protein expression system. The a 2hn-RT1.Ac and RT1.An proteins, when administered subcutaneously into PVG hosts 7 days prior to transplantation, resulted in accelerated rejection of BN cardiac allografts. The a 1hn-RT1.Ac construct did not demonstrate such immunogenic effects. Intra-portal administration of a 1hn-RT1.Ac or RT1.An, in combination with perioperative CsA, induced tolerance to BN cardiac allografts. The a 1hn-RT1.Ac protein was able to induce tolerance in a larger majority of the PVG recipients and at a lower dose of protein when compared to the RT1.An protein. RT1.An administered orally to PVG recipients also induced long term survival of cardiac allografts. In vitro analysis revealed that lymphocytes from tolerant hosts were hyporesponsive to donor splenocytes, but responsive to 3rd party splenocytes. Evaluation of T cell cytokine expression patterns revealed that rejector PVG hosts displayed a Type I T-cell response when re-challenged with donor splenocytes, in contrast to tolerant animals that displayed a Type II T-cell response. FACS analysis of the T cells revealed that the ratio of CD4 to CD8 cells was 3:1 and was consistent in the groups tested suggesting a complex interaction between the subsets of T cells, yielding the observed results. Histologic analysis of the cardiac allografts revealed that tolerant PVG hosts maintained BN cardiac allografts without any evidence of acute or chronic rejection after 300 days post transplant. This body of work has demonstrated that the use of soluble donor/recipient allochimeric class I MHC proteins with a short peri-operative course of CsA resulted in transplant tolerance. This treatment regimen proffers a clinically relevant approach to the induction of tolerance across MHC barriers. ^

Collagen I modulates growth and gene expression in prostate cancer cells

Prostate cancer is the second leading cause of male cancer-related deaths in the United States. Interestingly, prostate cancer preferentially metastasizes to skeletal tissue. Once in the bone microenvironment, advanced prostate cancer becomes highly resistant to therapeutic modalities. Several factors, such as extracellular matrix (ECM) components, have been implicated in the spread and propagation of prostatic carcinoma. In these studies, we have utilized the PC3 cell line, derived from a human bone metastasis, to investigate the influence of the predominant bone ECM protein, type I collagen, on prostate cancer cell proliferation and gene expression. We have also initiated the design and production of ribozymes to specific gene targets that may influence prostate cancer bone metastasis. ^ Our results demonstrate that PC3 cells rapidly adhere and spread on collagen I to a greater degree than on fibronectin (FN) or poly-L-lysine (PLL). Flow cytometry analysis reveals the presence of the α1, α2 and α3 collagen binding integrin subunits. The use of antibody function blocking studies reveals that PC3 cells can utilize α2β 1 and α3β1 integrins to adhere to collagen I. Once plated on collagen I, the cells exhibit increased rates of proliferation compared with cells plated on FN or tissue culture plastic. Additionally, cells plated on collagen I show increased expression of proteins associated with progression through G1 phase of the cell cycle. Inhibitor studies point to a role for phosphatidylinositol 3-kinase (PI3K), MAP kinase (MAPK), and p70 S6 kinase in collagen I-mediated PC3 cell proliferation and cyclin D1 expression. To further characterize the effect of type I collagen on prostate cancer bone metastasis, we utilized a cDNA microarray strategy to monitor type I collagen-mediated changes in gene expression. Results of this analysis revealed a gene expression profile reflecting the increased proliferation occurring on type I collagen. Microarray analysis also revealed differences in the expression of specific gene targets that may impact on prostate cancer metastasis to bone. ^ As a result of our studies on the interaction of prostate cancer cells and the skeletal ECM, we sought to develop novel molecular tools for future gene therapy of functional knockdown experiments. To this end, we developed a series of ribozymes directed against the α2 integrin and at osteopontin, a protein implicated in the metastasis of various cancers, including prostate. These ribozymes should facilitate the future study of the mechanism of prostate cancer cell proliferation, and disease progression occurring at sites of skeletal metastasis where a type I collagen-based environment predominates. ^ Together these studies demonstrate the involvement of bone ECM proteins on prostate cancer cell proliferation and suggest that they may play a significant role on the growth of prostate metastases once in the bone microenvironment. ^

The role of plasmacytoid dendritic cells in the regulation of adaptive immunity

Plasmacytoid dendritic cells (pDCs) selectively express TLR7 which allows them to respond to RNA viruses and TLR9 which allows them to respond to DNA viruses and CpG oligonucleotides. Upon exposure to virus pDCs produce vast amounts of type I interferon (IFN) directly inhibiting viral replication and contributing to the activation of other immune cells. The ability of pDCs to promote B and T cell differentiation through type I IFN has been well documented although the role of additional factors including tumor necrosis factor (TNF) family members has not been thoroughly addressed. Here the expression of selected TNF family members in pDCs was examined and the role of TNF receptor-ligand interactions in the regulation of B and T lymphocyte growth and differentiation by pDCs was investigated. Upon stimulation with CpG-B, pDCs exhibit strong and stable expression of CD70, a TNF family ligand that binds to its receptor CD27 on memory B cells and promotes plasma cell differentiation and Ig secretion. Using an in vitro pDC/B cell co-culture system, it was determined that CpG-B-stimulated pDCs induce the proliferation of CD40L-activated human peripheral B cells and Ig secretion. This occurs independently of IFN and residual CpG, and requires physical contact between pDCs and B cells. CpG-stimulated pDCs induce the proliferation of both naive and memory B cells although Ig secretion is restricted to the memory subset. Blocking the interaction of CD70 with CD27 using an antagonist anti-CD70 antibody reduces the induction of B cell proliferation and IgG secretion by CpG-B-stimulated pDCs. Published studies have also indicated an important role for CD70 in promoting the expansion of CD4+ and CD8+ T cells and the development of effector function. CpG-B-stimulated pDCs induce naïve CD4+ T cell proliferation and production of multiple cytokines including IFN-γ, TNF-α, IL-10, IL-4, IL-5 and IL-13. Blocking the function of CD70 with an antagonist anti-CD70 antibody significantly reduced the induction of naïve CD4+ T cell proliferation by CpG-B-stimulated pDCs and the production of IL-4 and IL-13. Collectively these data indicate an important role for CD70 in the regulation of B and T lymphocyte growth and differentiation by pDCs. ^

Delayed Thrombus Resolution and Fibroproliferative Vascular Wound Healing From Deficiency of Type III Collagen: a Paradoxical Mechanism for Tissue Fragility

Vascular Ehlers-Danlos syndrome is a heritable disease of connective tissue caused by mutations in COL3A1, conferring a tissue deficiency of type III collagen. Cutaneous wounds heal poorly in these patients, and they are susceptible to spontaneous and catastrophic rupture of expansible hollow organs like the gut, uterus, and medium-sized to large arteries, which leads to premature death. Although the predisposition for organ rupture is often attributed to inherent tissue fragility, investigation of arteries from a haploinsufficient Col3a1 mouse model (Col3a1+/-) demonstrates that mutant arteries withstand even supraphysiologic pressures comparably to wild-type vessels. We hypothesize that injury that elicits occlusive thrombi instead unmasks defective thrombus resolution resulting from impaired production of type III collagen, which causes deranged remodeling of matrix, persistent inflammation, and dysregulated behavior by resident myofibroblasts, culminating in the development of penetrating neovascular channels that disrupt the mechanical integrity of the arterial wall. Vascular injury and thrombus formation following ligation of the carotid artery reveals an abnormal persistence and elevated burden of occlusive thrombi at 21 post-operative days in vessels from Col3a1+/- mice, as opposed to near complete resolution and formation of a patent and mature neointima in wild-type mice. At only 14 days, both groups harbor comparable burdens of resolving thrombi, but wild-type mice increase production of type III collagen in actively resolving tissues, while mutant mice do not. Rather, thrombi in mutant mice contain higher burdens of macrophages and proliferative myofibroblasts, which persist through 21 days while wild-type thrombi, inflammatory cells, and proliferation all regress. At the same time that increased macrophage burdens were observed at 14 and 21 days post ligation, the medial layer of mutant arterial walls concurrently harbored a significantly higher incidence of penetrating neovessels compared with those in wild-type mice. To assess whether limited type III collagen production alters myofibroblast behavior, fibroblasts from vEDS patients with COL3A1 missense mutations were seeded into three-dimensional fibrin gel constructs and stimulated with transforming growth factor-β1 to initiate myofibroblast differentiation. Although early signaling events occur similarly in all cell lines, late extracellular matrix- and mechanically-regulated events like transcriptional upregulation of type I and type III collagen secretion are delayed in mutant cultures, while transcription of genes encoding intracellular contractile machinery is increased. Sophisticated imaging of collagen synthesized de novo by resident myofibroblasts visualizes complex matrix reorganization by control cells but only meager remodeling by COL3A1 mutant cells, concordant with their compensatory contraction to maintain tension in the matrix. Finally, administration of immunosuppressive rapamycin to mice following carotid ligation sufficiently halts the initial inflammatory phase of thrombus resolution and fully prevents both myofibroblast migration into the thrombus and the differential development of neovessels between mutant and wild-type mice, suggesting that pathological defects in mutant arteries develop secondarily to myofibroblast dysfunction and chronic inflammatory stimulation, rather than as a manifestation of tissue fragility. Together these data establish evidence that pathological defects in the vessel wall architecture develop in mutant arteries as sequelae to abnormal healing and remodeling responses activated by arterial injury. Thus, these data support the hypothesis that events threatening the integrity of type III collagen-deficient vessels develop not as a result of inherent tissue weakness and fragility at baseline but instead as an episodic byproduct of abnormally persistent granulation tissue and fibroproliferative intravascular remodeling.

Characterizing Overall Response to Two-shot Vaccination for BVDV Type 2 in Angus Weanling Calves

Vaccination is a management strategy utilized to help reduce prevalence of bovine respiratory disease in feedlots. However, not all animals respond similarly to vaccinations. It is believed that an animal’s genetics control part of the ability to respond to a vaccination protocol. In order to evaluate the genetic control of a new trait such as response to vaccination, it is important to understand the non-genetic factors that affect an animal’s response to vaccination. The objective of this study was to characterize the non-genetic factors affecting overall response to a two-shot vaccination for bovine viral diarrhea virus type 2 (BVDV2) in Angus weanling calves.

Practice advisory: interim guidance for care of obstetric patients during a zika virus outbreak

Zika during pregnancy has been associated with birth defects, specifically significant microcephaly. Transmission of Zika to the fetus has been documented in all trimesters; Zika virus RNA has been detected in fetal tissue from early missed abortions, amniotic fluid, term neonates and the placenta. However, much is not yet known about Zika virus in pregnancy. Uncertainties include the incidence of Zika virus infection among pregnant women in areas of Zika virus transmission, the rate of vertical transmission and the rate with which infected fetuses manifest complications such as microcephaly or demise. The absence of this important information makes management and decision making in the setting of potential Zika virus exposure (i.e. travel to endemic areas) or maternal infection, difficult. Currently, there is no vaccine or treatment for this infection.

PeECE I - Pelagic Ecosystem CO2 Enrichment Study, Raunefjord, Bergen, Norway, 2001

The effects of CO2-induced seawater acidification on plankton communities were also addressed in a series of 3 mesocosm experiments, called the Pelagic Ecosystem CO2 Enrichment (PeECE I-III) studies, which were conducted in the Large-Scale Mesocosm Facilities of the University of Bergen, Norway in 2001, 2003 and 2005, respectively. Each experiment consisted of 9 mesocosms, in which CO2 was manipulated to initial concentrations of 190, 350 and 750 µatm in 2001 and 2003, and 350, 700 and 1050 µatm in 2005. The present dataset concerns PeECE I.

High efficient luminescence in type-II GaAsSb-capped InAs quantum dots upon annealing

The photoluminescence efficiency of GaAsSb-capped InAs/GaAs type II quantum dots (QDs) can be greatly enhanced by rapid thermal annealing while preserving long radiative lifetimes which are ∼20 times larger than in standard GaAs-capped InAs/GaAs QDs. Despite the reduced electron-hole wavefunction overlap, the type-II samples are more efficient than the type-I counterparts in terms of luminescence, showing a great potential for device applications. Strain-driven In-Ga intermixing during annealing is found to modify the QD shape and composition, while As-Sb exchange is inhibited, allowing to keep the type-II structure. Sb is only redistributed within the capping layer giving rise to a more homogeneous composition.

Multi-dimensional classification using Bayesian networks for stationary and evolving streaming data

Hoy en día, con la evolución continua y rápida de las tecnologías de la información y los dispositivos de computación, se recogen y almacenan continuamente grandes volúmenes de datos en distintos dominios y a través de diversas aplicaciones del mundo real. La extracción de conocimiento útil de una cantidad tan enorme de datos no se puede realizar habitualmente de forma manual, y requiere el uso de técnicas adecuadas de aprendizaje automático y de minería de datos. La clasificación es una de las técnicas más importantes que ha sido aplicada con éxito a varias áreas. En general, la clasificación se compone de dos pasos principales: en primer lugar, aprender un modelo de clasificación o clasificador a partir de un conjunto de datos de entrenamiento, y en segundo lugar, clasificar las nuevas instancias de datos utilizando el clasificador aprendido. La clasificación es supervisada cuando todas las etiquetas están presentes en los datos de entrenamiento (es decir, datos completamente etiquetados), semi-supervisada cuando sólo algunas etiquetas son conocidas (es decir, datos parcialmente etiquetados), y no supervisada cuando todas las etiquetas están ausentes en los datos de entrenamiento (es decir, datos no etiquetados). Además, aparte de esta taxonomía, el problema de clasificación se puede categorizar en unidimensional o multidimensional en función del número de variables clase, una o más, respectivamente; o también puede ser categorizado en estacionario o cambiante con el tiempo en función de las características de los datos y de la tasa de cambio subyacente. A lo largo de esta tesis, tratamos el problema de clasificación desde tres perspectivas diferentes, a saber, clasificación supervisada multidimensional estacionaria, clasificación semisupervisada unidimensional cambiante con el tiempo, y clasificación supervisada multidimensional cambiante con el tiempo. Para llevar a cabo esta tarea, hemos usado básicamente los clasificadores Bayesianos como modelos. La primera contribución, dirigiéndose al problema de clasificación supervisada multidimensional estacionaria, se compone de dos nuevos métodos de aprendizaje de clasificadores Bayesianos multidimensionales a partir de datos estacionarios. Los métodos se proponen desde dos puntos de vista diferentes. El primer método, denominado CB-MBC, se basa en una estrategia de envoltura de selección de variables que es voraz y hacia delante, mientras que el segundo, denominado MB-MBC, es una estrategia de filtrado de variables con una aproximación basada en restricciones y en el manto de Markov. Ambos métodos han sido aplicados a dos problemas reales importantes, a saber, la predicción de los inhibidores de la transcriptasa inversa y de la proteasa para el problema de infección por el virus de la inmunodeficiencia humana tipo 1 (HIV-1), y la predicción del European Quality of Life-5 Dimensions (EQ-5D) a partir de los cuestionarios de la enfermedad de Parkinson con 39 ítems (PDQ-39). El estudio experimental incluye comparaciones de CB-MBC y MB-MBC con los métodos del estado del arte de la clasificación multidimensional, así como con métodos comúnmente utilizados para resolver el problema de predicción de la enfermedad de Parkinson, a saber, la regresión logística multinomial, mínimos cuadrados ordinarios, y mínimas desviaciones absolutas censuradas. En ambas aplicaciones, los resultados han sido prometedores con respecto a la precisión de la clasificación, así como en relación al análisis de las estructuras gráficas que identifican interacciones conocidas y novedosas entre las variables. La segunda contribución, referida al problema de clasificación semi-supervisada unidimensional cambiante con el tiempo, consiste en un método nuevo (CPL-DS) para clasificar flujos de datos parcialmente etiquetados. Los flujos de datos difieren de los conjuntos de datos estacionarios en su proceso de generación muy rápido y en su aspecto de cambio de concepto. Es decir, los conceptos aprendidos y/o la distribución subyacente están probablemente cambiando y evolucionando en el tiempo, lo que hace que el modelo de clasificación actual sea obsoleto y deba ser actualizado. CPL-DS utiliza la divergencia de Kullback-Leibler y el método de bootstrapping para cuantificar y detectar tres tipos posibles de cambio: en las predictoras, en la a posteriori de la clase o en ambas. Después, si se detecta cualquier cambio, un nuevo modelo de clasificación se aprende usando el algoritmo EM; si no, el modelo de clasificación actual se mantiene sin modificaciones. CPL-DS es general, ya que puede ser aplicado a varios modelos de clasificación. Usando dos modelos diferentes, el clasificador naive Bayes y la regresión logística, CPL-DS se ha probado con flujos de datos sintéticos y también se ha aplicado al problema real de la detección de código malware, en el cual los nuevos ficheros recibidos deben ser continuamente clasificados en malware o goodware. Los resultados experimentales muestran que nuestro método es efectivo para la detección de diferentes tipos de cambio a partir de los flujos de datos parcialmente etiquetados y también tiene una buena precisión de la clasificación. Finalmente, la tercera contribución, sobre el problema de clasificación supervisada multidimensional cambiante con el tiempo, consiste en dos métodos adaptativos, a saber, Locally Adpative-MB-MBC (LA-MB-MBC) y Globally Adpative-MB-MBC (GA-MB-MBC). Ambos métodos monitorizan el cambio de concepto a lo largo del tiempo utilizando la log-verosimilitud media como métrica y el test de Page-Hinkley. Luego, si se detecta un cambio de concepto, LA-MB-MBC adapta el actual clasificador Bayesiano multidimensional localmente alrededor de cada nodo cambiado, mientras que GA-MB-MBC aprende un nuevo clasificador Bayesiano multidimensional. El estudio experimental realizado usando flujos de datos sintéticos multidimensionales indica los méritos de los métodos adaptativos propuestos. ABSTRACT Nowadays, with the ongoing and rapid evolution of information technology and computing devices, large volumes of data are continuously collected and stored in different domains and through various real-world applications. Extracting useful knowledge from such a huge amount of data usually cannot be performed manually, and requires the use of adequate machine learning and data mining techniques. Classification is one of the most important techniques that has been successfully applied to several areas. Roughly speaking, classification consists of two main steps: first, learn a classification model or classifier from an available training data, and secondly, classify the new incoming unseen data instances using the learned classifier. Classification is supervised when the whole class values are present in the training data (i.e., fully labeled data), semi-supervised when only some class values are known (i.e., partially labeled data), and unsupervised when the whole class values are missing in the training data (i.e., unlabeled data). In addition, besides this taxonomy, the classification problem can be categorized into uni-dimensional or multi-dimensional depending on the number of class variables, one or more, respectively; or can be also categorized into stationary or streaming depending on the characteristics of the data and the rate of change underlying it. Through this thesis, we deal with the classification problem under three different settings, namely, supervised multi-dimensional stationary classification, semi-supervised unidimensional streaming classification, and supervised multi-dimensional streaming classification. To accomplish this task, we basically used Bayesian network classifiers as models. The first contribution, addressing the supervised multi-dimensional stationary classification problem, consists of two new methods for learning multi-dimensional Bayesian network classifiers from stationary data. They are proposed from two different points of view. The first method, named CB-MBC, is based on a wrapper greedy forward selection approach, while the second one, named MB-MBC, is a filter constraint-based approach based on Markov blankets. Both methods are applied to two important real-world problems, namely, the prediction of the human immunodeficiency virus type 1 (HIV-1) reverse transcriptase and protease inhibitors, and the prediction of the European Quality of Life-5 Dimensions (EQ-5D) from 39-item Parkinson’s Disease Questionnaire (PDQ-39). The experimental study includes comparisons of CB-MBC and MB-MBC against state-of-the-art multi-dimensional classification methods, as well as against commonly used methods for solving the Parkinson’s disease prediction problem, namely, multinomial logistic regression, ordinary least squares, and censored least absolute deviations. For both considered case studies, results are promising in terms of classification accuracy as well as regarding the analysis of the learned MBC graphical structures identifying known and novel interactions among variables. The second contribution, addressing the semi-supervised uni-dimensional streaming classification problem, consists of a novel method (CPL-DS) for classifying partially labeled data streams. Data streams differ from the stationary data sets by their highly rapid generation process and their concept-drifting aspect. That is, the learned concepts and/or the underlying distribution are likely changing and evolving over time, which makes the current classification model out-of-date requiring to be updated. CPL-DS uses the Kullback-Leibler divergence and bootstrapping method to quantify and detect three possible kinds of drift: feature, conditional or dual. Then, if any occurs, a new classification model is learned using the expectation-maximization algorithm; otherwise, the current classification model is kept unchanged. CPL-DS is general as it can be applied to several classification models. Using two different models, namely, naive Bayes classifier and logistic regression, CPL-DS is tested with synthetic data streams and applied to the real-world problem of malware detection, where the new received files should be continuously classified into malware or goodware. Experimental results show that our approach is effective for detecting different kinds of drift from partially labeled data streams, as well as having a good classification performance. Finally, the third contribution, addressing the supervised multi-dimensional streaming classification problem, consists of two adaptive methods, namely, Locally Adaptive-MB-MBC (LA-MB-MBC) and Globally Adaptive-MB-MBC (GA-MB-MBC). Both methods monitor the concept drift over time using the average log-likelihood score and the Page-Hinkley test. Then, if a drift is detected, LA-MB-MBC adapts the current multi-dimensional Bayesian network classifier locally around each changed node, whereas GA-MB-MBC learns a new multi-dimensional Bayesian network classifier from scratch. Experimental study carried out using synthetic multi-dimensional data streams shows the merits of both proposed adaptive methods.