9616 research outputs found
Sort by
Detecting changes in product quality by statistical divergence
No full textPri spremljanju velike količine podatkov v proizvodnih procesih je pomembno sprotno zaznavati, ali je v njihovem vzorcu prišlo do morebitne spremembe. Ker so podatki običajno rezultat procesov z naključnimi vplivi, jih je treba obravnavati s pomočjo statističnih orodij. Za zaznavanje razlike med podatkovnimi nizi se pogosto uporablja pristop z divergenco. Medtem ko je v teoriji divergenca močno orodje, pri računanju s histogrami v praksi naletimo na numerične probleme. O tem je v literaturi napisanega sorazmerno malo. Namen magistrskega dela je analizirati numerične omejitve pri računanju Kullback-Leiblerjeve divergence na simuliranih podatkih ter predlagati numerični postopek, ki premosti omenjene omejitve. Ker je kakovost histogramov pomembna za računanje divergence, smo najprej analizirali metode za realizacijo histogramov s stališča izbora stolpcev. Potem smo analizirali probleme, ki se pojavljajo v praktičnih aplikacijah, kadar sta dva histograma močno narazen, zlasti na področjih, kjer ni prekrivanja med njima. En možen način za izognitev težavam je aproksimacija histograma z jedrnimi funkcijami Gaussovega tipa. Na ta način se izognemo težavam pri računanju Kullback-Leiblerjeve divergence. Pristop smo najprej preizkusili na simuliranih numeričnih podatkih in potem na realnih podatkih iz končnega nadzora kakovosti v serijski izdelavi elektromotorjev.When monitoring large amounts of data in manufacturing processes, it is important to detect changes in the process in real time. Since data result from processes subjected by random disturbances, it is necessary to treat them by means of the statistical tools. The divergence approach is often used to detect the difference between data sets. In theory, divergence is a powerful tool, however, numerical problems are encountered when calculating the divergence with histograms in practice. Relatively little can be found about this in the literature. The purpose of this thesis is to analyze numerical limitations in calculating the Kullback-Leibler divergence on simulated data and to propose a numerical procedure that bridges these limitations. Since the quality of histograms is important for divergence calculation, we first analyzed the methods for realization of histograms from the point of view of bins selection. Then we analyzed the problems that arise in practical applications when two histograms are strongly apart, especially in the areas where there is no overlap between them. One possible way to avoid problems is to approximate the histogram with the Gaussian kernel functions. This avoids the difficulty of calculating the Kullback-Leibler divergence. The approach was first tested on simulated numerical data and then on real data from final quality control in serial manufacturing of electric motors
Probing protoneutron stars with gamma-ray axionscopes
No full textAxion-like particles (ALPs) coupled to nucleons can be efficiently produced in the
interior of protoneutron stars (PNS) during supernova (SN) explosions. If these ALPs are also
coupled to photons they can convert into gamma rays in the Galactic magnetic field. This
SN-induced gamma-ray burst can be observable by gamma-ray telescopes like Fermi-LAT if the
SN is in the field of view of the detector. We show that the observable gamma-ray spectrum is
sensitive to the production processes in the SN core. In particular, if the nucleon-nucleon
bremsstrahlung is the dominant axion production channel, one expects a thermal spectrum with
average energy E
a
≃ 50 MeV. In this case the gamma-ray spectrum observation allows for
the reconstruction of the PNS temperature. In case of a sizable pion abundance in the SN core, one
expects a second spectral component peaked at E
a
≃ 200 MeV due to axion pionic
processes. We demonstrate that, through a dedicated LAT analysis, we can detect the presence of
this pionic contribution, showing that the detection of the spectral shape of the gamma-ray signal
represents a unique probe of the pion abundance in the PNS
Contribution of APOBEC3 proteins to the oncogenicity of HPV viruses
No full textHuman papillomaviruses (HPV) cause almost 5% of all human malignancies, including
cervical cancer and head and neck cancer. The DNA-editing activity of a group of cytosine
deaminases, APOBEC 3 (A3), is part of the innate immune response to viral infections.
However, in persistent HPV infections, A3 proteins are involved in the accumulation of
specific mutational signatures in the host cell genome, leading to high levels of DNA damage
and oncogenesis. Two members of this family, A3A and A3B, have been associated with a
high mutational burden in HPV-related cancers.
Our recent research focusses on the editing-independent role of A3 proteins in HPV
infection and cell transformation, which is still largely unknown. First, we analysed
the expression profile of head and neck cancer patients from The Cancer Genome Atlas
(TCGA) and identified genes that correlated with either A3A or A3B. The in silico results
were complemented by expression analyses of HPV host cells HFK (human foreskin
keratinocytes) and HFK-16, which lack the A3A or A3B protein. The experimental data
confirmed an altered expression of host genes that correlates with the expression of A3A or
A3B proteins and HPV oncoproteins. Genes whose expression correlates with A3A appeared
to be generally downregulated in HPV-positive patients. GO analysis of these genes revealed
an enrichment of genes associated with epidermal differentiation and innate immunity.
Genes correlated with A3B are associated with the cell cycle, chromosome organisation,
DNA replication and DNA repair and were overexpressed in HPV-positive cancer patients.
This indicates a different role for the two enzymes. A3A- and A3B-silenced cell lines also
showed altered patterns of cell metabolic activity, attachment and migration. Finally, some
selected oncogenes identified in the expression analysis and the time frame of altered A3Aand
A3B-dependent gene expression during HPV infection will be discussed. Overall, the
presentation will provide insight into the role of editing-independent activity of A3 proteins
in cell transformation, particularly in the context of HPV infection
