Interpretability methods of machine learning algorithms with applications in breast cancer diagnosis.

Early detection of breast cancer is a powerful tool towards decreasing its socioeconomic burden. Although, artificial intelligence (AI) methods have shown remarkable results towards this goal, their "black box" nature hinders their wide adoption in clinical practice. To address the need for AI guided breast cancer diagnosis, interpretability methods can be utilized. In this study, we used AI methods, i.e., Random Forests (RF), Neural Networks (NN) and Ensembles of Neural Networks (ENN), towards this goal and explained and optimized their performance through interpretability techniques, such as the Global Surrogate (GS) method, the Individual Conditional Expectation (ICE) plots and the Shapley values (SV). The Wisconsin Diagnostic Breast Cancer (WDBC) dataset of the open UCI repository was used for the training and evaluation of the AI algorithms. The best performance for breast cancer diagnosis was achieved by the proposed ENN (96.6% accuracy and 0.96 area under the ROC curve), and its predictions were explained by ICE plots, proving that its decisions were compliant with current medical knowledge and can be further utilized to gain new insights in the pathophysiological mechanisms of breast cancer. Feature selection based on features' importance according to the GS model improved the performance of the RF (leading the accuracy from 96.49% to 97.18% and the area under the ROC curve from 0.96 to 0.97) and feature selection based on features' importance according to SV improved the performance of the NN (leading the accuracy from 94.6% to 95.53% and the area under the ROC curve from 0.94 to 0.95). Compared to other approaches on the same dataset, our proposed models demonstrated state of the art performance while being interpretable.

Optical polarization properties of (11-22) semi-polar InGaN LEDs with a wide spectral range.

Electroluminescence polarization measurements have been performed on a series of semi-polar InGaN light emitting diodes (LEDs) grown on semi-polar (11-22) templates with a high crystal quality. The emission wavelengths of these LEDs cover a wide spectral region from 443 to 555 nm. A systematic study has been carried out in order to investigate the influence of both indium content and injection current on polarization properties, where a clear polarization switching at approximately 470 nm has been observed. The shortest wavelength LED (443 nm) exhibits a positive 0.15 polarization degree, while the longest wavelength LED (555 nm) shows a negative -0.33 polarization degree. All the longer wavelength LEDs with an emission wavelength above 470 nm exhibit negative polarization degrees, and they further demonstrate that the dependence of polarization degree on injection current enhances with increasing emission wavelength. Moreover, the absolute value of the polarization degree decreases with increasing injection current. In contrast, the polarization degree of the 443 nm blue LED remains constant with changing injection current. This discrepancy can be attributed to a significant difference in the density of states (DOS) of the valence subbands.

Monolithically integrated white light LEDs on (11-22) semi-polar GaN templates.

Carrier transport issues in a (11-22) semi-polar GaN based white light emitting diode (consisting of yellow and blue emissions) have been investigated by detailed simulations, demonstrating that the growth order of yellow and blue InGaN quantum wells plays a critically important role in achieving white emission. The growth order needs to be yellow InGaN quantum wells first and then a blue InGaN quantum well after the growth of n-type GaN. The fundamental reason is due to the poor hole concentration distribution across the whole InGaN quantum well region. In order to effectively capture holes in both the yellow InGaN quantum wells and the blue InGaN quantum well, a thin GaN spacer has been introduced prior to the blue InGaN quantum well. The detailed simulations of the band diagram and the hole concentration distribution across the yellow and the blue quantum wells have been conducted, showing that the thin GaN spacer can effectively balance the hole concentration between the yellow and the blue InGaN quantum wells, eventually determining their relative intensity between the yellow and the blue emissions. Based on this simulation, we have demonstrated a monolithically multi-colour LED grown on our high quality semi-polar (11-22) GaN templates.

Monolithically multi-color lasing from an InGaN microdisk on a Si substrate.

An optically pumped multi-color laser has been achieved using an InGaN/GaN based micro-disk with an undercut structure on a silicon substrate. The micro-disk laser has been fabricated by means of a combination of a cost-effective microsphere lithography technique and subsequent dry/wet etching processes. The microdisk laser is approximately 1 µm in diameter. The structure was designed in such a way that the vertical components of the whispering gallery (WG) modes formed can be effectively suppressed. Consequently, three clean lasing peaks at 442 nm, 493 nm and 522 nm have been achieved at room temperature by simply using a continuous-wave diode laser as an optical pumping source. Time-resolved micro photoluminescence (PL) measurements have been performed in order to further confirm the lasing by investigating the excitonic recombination dynamics of these lasing peaks. A three dimensional finite-difference-time-domain (FDTD) simulation has been used for the structure design.

Polarized white light from hybrid organic/III-nitrides grating structures.

Highly polarised white light emission from a hybrid organic/inorganic device has been achieved. The hybrid devices are fabricated by means of combining blue InGaN-based multiple quantum wells (MQWs) with a one-dimensional (1D) grating structure and down-conversion F8BT yellow light emitting polymer. The 1D grating structure converts the blue emission from unpolarised to highly polarised; Highly polarised yellow emission has been achieved from the F8BT polymer filled and aligned along the periodic nano-channels of the grating structure as a result of enhanced nano-confinement. Optical polarization measurements show that our device demonstrates a polarization degree of up to 43% for the smallest nano-channel width. Furthermore, the hybrid device with such a grating structure allows us to achieve an optimum relative orientation between the dipoles in the donor (i.e., InGaN/GaN MQWs) and the diploes in the acceptor (i.e., the F8BT), maximizing the efficiency of non-radiative energy transfer (NRET) between the donor and the acceptor. Time-resolved micro photoluminescence measurements show a 2.5 times enhancement in the NRET efficiency, giving a maximal NRET efficiency of 90%. It is worth highlighting that the approach developed paves the way for the fabrication of highly polarized white light emitters.

Room temperature continuous-wave green lasing from an InGaN microdisk on silicon.

Optically pumped green lasing with an ultra low threshold has been achieved using an InGaN/GaN based micro-disk with an undercut structure on silicon substrates. The micro-disks with a diameter of around 1 µm were fabricated by means of a combination of a cost-effective silica micro-sphere approach, dry-etching and subsequent chemical etching. The combination of these techniques both minimises the roughness of the sidewalls of the micro-disks and also produces excellent circular geometry. Utilizing this fabrication process, lasing has been achieved at room temperature under optical pumping from a continuous-wave laser diode. The threshold for lasing is as low as 1 kW/cm(2). Time-resolved micro photoluminescence (PL) and confocal PL measurements have been performed in order to further confirm the lasing action in whispering gallery modes and also investigate the excitonic recombination dynamics of the lasing.

Influenza surveillance during the post-pandemic influenza 2010/11 season in Greece, 04 October 2010 to 22 May 2011.

In this manuscript, we summarise the experience of Greece during the post-pandemic influenza season 2010/11 from 04 October 2010 to 22 May 2011. The spread of the disease and its impact were monitored using multiple surveillance systems, such as sentinel surveillance, virological surveillance and all-cause mortality surveillance. We also focus on the characteristics of laboratory-confirmed severe influenza cases who required admission to an intensive care unit (ICU) (n=368), and/or with a fatal outcome (n=180). The influenza-like illness rate reported from sentinel surveillance started rising in early January 2011 and peaked between 31 January and 6 February 2011. The total number of ICU admissions was higher in the post-pandemic influenza season than during the pandemic period causing a lot of pressure on ICUs. The overall population mortality rate due to influenza A(H1N1)2009 was higher than during the pandemic period (15.9 vs 13.2 fatal cases per million, p=0.087). Our data suggest that the severity of clinical illness in the first post-pandemic influenza season was comparable or even higher than during the pandemic.

Respiratory health of municipal solid waste workers.

BACKGROUND: There is an increasing evidence that the incidence of work-related pulmonary problems is greater in waste collectors than in the general workforce. AIMS: To evaluate the respiratory health of municipal solid waste workers (MSWWs). METHODS: One hundred and eighty-four municipal employees of Keratsini (104 MSWWs and 80 controls) participated in a cross-sectional study. All participants were asked to fill in a slightly modified version of the Medical Research Council questionnaire. Lung function was evaluated by spirometry. RESULTS: Spirometry revealed reduced mean forced vital capacity (FVC) and forced expiratory volume in 1 s (as a percentage of predicted values) in MSWWs compared with controls. After adjustment for smoking status, only the decline in FVC was statistically significant (P < 0.05). Prevalence of all respiratory symptoms was higher in MSWWs than in controls. After adjustment for confounding factors, the difference reached statistical significance (P < 0.05) for morning cough, cough on exertion and sore throat. CONCLUSIONS: The results of this cross-sectional study indicate a higher prevalence of respiratory symptoms and a greater decrease in lung function in MSWWs. A number of limitations such as the relatively small size of population and the 'healthy worker' effect should be taken into account.

Converging evidence for a pseudoautosomal cytokine receptor gene locus in schizophrenia.

Schizophrenia is a strongly heritable disorder, and identification of potential candidate genes has accelerated in recent years. Genomewide scans have identified multiple large linkage regions across the genome, with fine-mapping studies and other investigations of biologically plausible targets demonstrating several promising candidate genes of modest effect. The recent introduction of technological platforms for whole-genome association (WGA) studies can provide an opportunity to rapidly identify novel targets, although no WGA studies have been reported in the psychiatric literature to date. We report results of a case-control WGA study in schizophrenia, examining approximately 500 000 markers, which revealed a strong effect (P=3.7 x 10(-7)) of a novel locus (rs4129148) near the CSF2RA (colony stimulating factor, receptor 2 alpha) gene in the pseudoautosomal region. Sequencing of CSF2RA and its neighbor, IL3RA (interleukin 3 receptor alpha) in an independent case-control cohort revealed both common intronic haplotypes and several novel, rare missense variants associated with schizophrenia. The presence of cytokine receptor abnormalities in schizophrenia may help explain prior epidemiologic data relating the risk for this illness to altered rates of autoimmune disorders, prenatal infection and familial leukemia.

A patient with Klinefelter's syndrome and thalassemia intermedia.

Klinefelter's syndrome (KS) is associated with a wide spectrum of clinical features, such as tall stature, eunuchoid proportions, testes disproportionately small for the level of pubertal development, gynecomastia and behavioral problems. The association of KS with thalassemia intermedia has not been previously reported. A male patient with thalassemia intermedia was diagnosed with KS at the age of 14 years when endocrine evaluation for delayed puberty showed hypergonadotrophic hypogonadism. Thyroid function was normal; however, basal and GnRH-stimulated gonadotropin concentrations were raised while serum testosterone was low. Karyotype analysis revealed KS (47,XXY). Testosterone replacement therapy started soon after diagnosis and now at the age of 20 years the patient's height is 178.3 cm, the U/L ratio is 0.91. Testicular volume is 12 ml (Prader orchidometer) and his pubic hair is stage 4. To our knowledge this is the first case of a patient suffering from KS and thalassemia intermedia reported in the literature.

Urine biochemical markers of early renal dysfunction are associated with iron overload in beta-thalassaemia.

Renal dysfunction in thalassemia patients can be attributed to chronic anemia, and iron overload as well as to desferioxamine (DFO) toxicity. We analyzed the urine of 91 well-maintained homozygous beta-thalassemia patients, with no evidence of renal disease, for early evidence of kidney dysfunction by means of electrophoresis and quantitative biochemical tests. Measurement of liver magnetic resonance imaging (MRI) T2 values and serum ferritin concentration was used to estimate iron overload. In 55 of the 91 patients, urine analysis indicated signs of tubular dysfunction. The urine concentration of albumin and beta 2-microglobulin, as well as the activity of N-acetyl-beta-D-glucosaminidase (NAG), correlated positively with serum ferritin concentration and liver iron deposition, as detected by MRI T2 values. This suggested that the cause of renal dysfunction in homozygous beta-thalassemia is iron overload. On the other hand, the same urine markers did not correlate with age, indicating that chronic anemia or desferrioxamine (DFO) treatment are not related to renal dysfunction in thalassemia.

Genomic analysis of orthologous mouse and human olfactory receptor loci.

Olfactory receptor (OR) genes represent approximately 1% of genomic coding sequence in mammals, and these genes are clustered on multiple chromosomes in both the mouse and human genomes. We have taken a comparative genomics approach to identify features that may be involved in the dynamic evolution of this gene family and in the transcriptional control that results in a single OR gene expressed per olfactory neuron. We sequenced approximately 350 kb of the murine P2 OR cluster and used synteny, gene linkage, and phylogenetic analysis to identify and sequence approximately 111 kb of an orthologous cluster in the human genome. In total, 18 mouse and 8 human OR genes were identified, including 7 orthologs that appear to be functional in both species. Noncoding homology is evident between orthologs and generally is confined within the transcriptional unit. We find no evidence for common regulatory features shared among paralogs, and promoter regions generally do not contain strong promoter motifs. We discuss these observations, as well as OR clustering, in the context of evolutionary expansion and transcriptional regulation of OR repertoires.

Suppression of the Ewing's sarcoma phenotype by FLI1/ERF repressor hybrids.

Fusion of the 5' half of the Ewing's sarcoma (ES) gene EWS with the DNA-binding domain of several transcription factors has been detected in many human tumors. The t(11;22)(q24;q12) chromosomal translocation is specifically linked to ES and primitive neuroectodermal tumors and results, in the majority of cases, in the fusion of the amino terminus of the EWS gene to the carboxyl-terminal DNA-binding domain of the FLI1 gene. The chimeric protein has been shown to be oncogenic, a potent transcriptional activator, and necessary for the maintenance of the Ewing's phenotype, making it an attractive target for gene therapy. In this study, we demonstrate that the ES transformed phenotype can be suppressed by chimeric transcriptional repressors containing the DNA-binding domain of FLI1 and the regulatory and repressor domain of ERF, a transcription suppressor and member of the ets gene family. The hybrid repressor is expressed at levels comparable with EWS/FLI1, does not affect EWS/FLI1 expression, and exhibits similar DNA-binding specificity but suppresses transcriptional activity. The FLI1/ERF repressor, like the wild-type ERF, is regulated by mitogen-activated protein kinase-dependent subcellular localization. Our data suggest that transformation by EWS/FLI1 may partially be due to activation of specific EWS/FLI1-regulated genes involved in cell proliferation.

Cloning of the murine Drm gene (Cktsf1b1) and characterization of its oncogene suppressible promoter.

The Drm gene, first identified in rat cells in our laboratory, appears to play a significant role in early embryo patterning and limb bud development. We have now isolated mouse Drm (mDrm) cDNA as well as genomic DNA clones and have mapped the Drm gene (Cktsf1b1) to murine chromosome 2. Cktsf1b1 is regulated in a tissue specific fashion and is expressed only in nontransformed mouse cells or primary fibroblasts in culture, but not in established transformed or tumor-derived mouse cell lines. The major transcription start sites map to within 69 bp upstream of the initiating ATG. A promoter was contained in the -214 to +1 bp 5' flanking region, and promoter/reporter constructs showed 10-fold higher activity than control in REF-1 (rat), A31 (mouse) and CHO (hamster) cells. The region contains a TATA sequence and multiple potential transcription factor binding sites. Promoter activity was dose-dependently inhibited by cotransfection with either ras or mos oncogenes, but oncogene inhibition was reversed and the overall activity increased when cells were treated with the MAP kinase kinase (MKK) inhibitor PD98059. An NF-1 and Yi-like site, identified in the minimal promoter region, showed different mobility shift patterns when normal and transformed cell nuclear extracts are compared. Mutation of the NF-1 site reduced Cktsf1b1 promoter activity 25%, while mutation of the Yi-like site destroyed all the activity. Our results indicate that the expression of Cktsf1b1, a gene associated with early development and cell transformation, is sensitive to MKK levels and may be regulated via multiple transcription factor complexes.

FLI-1 is a suppressor of erythroid differentiation in human hematopoietic cells.

The FLI-1 oncogene, a member of the ETS family of transcription factors, is associated with both normal and abnormal hematopoietic cell growth and lineage-specific differentiation. We have previously shown that overexpression of FLI-1 in pluripotent human hematopoietic cells leads to the induction of a megakaryocytic phenotype. In this report we show that FLI-1 also acts as an inhibitor of erythroid differentiation. Following the induction of erythroid differentiation, pluripotent cells express reduced levels of FLI-1. In contrast, when FLI-1 is overexpressed in these cells, the levels of erythroid markers are reduced. The ability of FLI-1 overexpressing cells to respond to erythroid-specific inducers such as hemin and Ara-C is also inhibited, and the uninduced cells show a reduced level of the erythroid-associated GATA-1 transcription factor mRNA. Furthermore, expression of a GATA-1 promoter-driven reporter construct in K562 cells is inhibited by co-transfection with a construct expressing FLI-1. Our results support the hypothesis that FLI-1 can act both positively and negatively in the regulation of hematopoietic cell differentiation, and that inhibition of GATA-1 expression may contribute to FLI-1-mediated inhibition of erythroid differentiation.

Ets and retroviruses - transduction and activation of members of the Ets oncogene family in viral oncogenesis.

Studies of retroviral-induced oncogenesis in animal systems led to the initial discovery of viral oncogenes and their cellular homologs, and provided critical insights into their role in the neoplastic process. V-ets, the founding member of the ETS oncogene family, was originally identified as part of the fusion oncogene encoded by the avian acute leukemia virus E26 and subsequent analysis of virus induced leukemias led to the initial isolation of two other members of the ETS gene family. PU.1 was identified as a target of insertional activation in the majority of tumors induced by the murine Spleen Focus Forming virus (SFFV), while fli-1 proved to be the target of Friend murine leukemia virus (F-MuLV) in F-MuLV induced erythroleukemia, as well as that of the 10A1 and Graffi viruses. The common features of the erythroid and myeloid diseases induced by these viruses provided the initial demonstration that these and other members of the ETS family play important roles in hematopoietic development as well as disease. This review provides an overview of the role of ETS genes in retrovirally induced neoplasia, their possible mechanisms of action, and how these viral studies relate to current knowledge of the functions of these genes in hematopoiesis.