Distributed Feature Selection in Large n and Large p Regression Problems

Fitting statistical models is computationally challenging when the sample size or the dimension of the dataset is huge. An attractive approach for down-scaling the problem size is to first partition the dataset into subsets and then fit using distributed algorithms. The dataset can be partitioned either horizontally (in the sample space) or vertically (in the feature space), and the challenge arise in defining an algorithm with low communication, theoretical guarantees and excellent practical performance in general settings. For sample space partitioning, I propose a MEdian Selection Subset AGgregation Estimator ({\em message}) algorithm for solving these issues. The algorithm applies feature selection in parallel for each subset using regularized regression or Bayesian variable selection method, calculates the `median' feature inclusion index, estimates coefficients for the selected features in parallel for each subset, and then averages these estimates. The algorithm is simple, involves very minimal communication, scales efficiently in sample size, and has theoretical guarantees. I provide extensive experiments to show excellent performance in feature selection, estimation, prediction, and computation time relative to usual competitors.

While sample space partitioning is useful in handling datasets with large sample size, feature space partitioning is more effective when the data dimension is high. Existing methods for partitioning features, however, are either vulnerable to high correlations or inefficient in reducing the model dimension. In the thesis, I propose a new embarrassingly parallel framework named {\em DECO} for distributed variable selection and parameter estimation. In {\em DECO}, variables are first partitioned and allocated to m distributed workers. The decorrelated subset data within each worker are then fitted via any algorithm designed for high-dimensional problems. We show that by incorporating the decorrelation step, DECO can achieve consistent variable selection and parameter estimation on each subset with (almost) no assumptions. In addition, the convergence rate is nearly minimax optimal for both sparse and weakly sparse models and does NOT depend on the partition number m. Extensive numerical experiments are provided to illustrate the performance of the new framework.

For datasets with both large sample sizes and high dimensionality, I propose a new "divided-and-conquer" framework {\em DEME} (DECO-message) by leveraging both the {\em DECO} and the {\em message} algorithm. The new framework first partitions the dataset in the sample space into row cubes using {\em message} and then partition the feature space of the cubes using {\em DECO}. This procedure is equivalent to partitioning the original data matrix into multiple small blocks, each with a feasible size that can be stored and fitted in a computer in parallel. The results are then synthezied via the {\em DECO} and {\em message} algorithm in a reverse order to produce the final output. The whole framework is extremely scalable.

Comparison of nitrogen isotopic values from different preparation methods

We are writing to comment on the work of Tamburini et al. (2003, doi:10.1029/2000PA000616). During the course of subsequent discussions between the authors and ourselves, it has become clear that the published sedimentary nitrogen isotopic values for Ocean Drilling Program (ODP) Site 724 are in error. Our reanalysis of sediment samples from the same intervals has revealed a significant offset from the original d15N data, requiring a revised assessment of their initial interpretation. The purposes of this comment are to (1) address the origin of these errors; (2) outline a protocol for future validation of nitrogen isotopic analyses; and (3) provide revised interpretations of the sedimentary d15N data in terms of the regional relative contributions of denitrification and nitrogen fixation and mean state of the southwest monsoon. (2) Nitrogen isotopic values measured on late Quaternary sediments at Arabian Sea ODP Site 724 by Tamburini et al. (2003, doi:10.1029/2000PA000616) are inexplicably different from a number of published records of d15N from very nearby on the Oman margin (Altabet et al., 1995, doi:10.1038/373506a0; 1999, doi:10.1029/1999PA900035; 2002, doi:10.1038/415159a; Higginson et al., 2004, doi:10.1016/j.gca.2004.03.015) and elsewhere in the Arabian Sea (Reichart et al., 1998, doi:10.1029/98PA02203). These data were generated using similar instrumentation (elemental analyzer coupled with an isotope ratio mass spectrometer) and analytical methodology to those already published. Concerned by this clear discrepancy, we analyzed aliquots of sediment from the same depth intervals for nitrogen abundance and bulk sedimentary nitrogen isotopes. We have been unable to duplicate the values published by Tamburini et al. (2003, doi:10.1029/2000PA000616 ), even after analysis of multiple replicates and due consideration of natural sediment heterogeneities and postrecovery sample storage.

Acceptance of relapse fears in breast cancer patients: effects of an act-based abridged intervention

Objective: Relapse fear is a common psychological scar in cancer survivors. The aim of this study is to assess the effects of an abridged version of Acceptance and Commitment Therapy (ACT) in breast cancer patients.Method: An open trial was developed with 12 non-metastatic breast cancer patients assigned to 2 conditions, ACT and waiting list. Interventions were applied in just one session and focused on the acceptance of relapse fears through a ‘defusion’ exercise. Interference and intensity of fear measured through subjective scales were collected after each intervention and again 3 months later. Distress, hypochondria and ‘anxious preocupation’ were also evaluated through standardized questionnaires.Results: The analysis revealed that ‘defusion’ contributed to decrease the interference of the fear of recurrence, and these changes were maintained three months after intervention in most subjects. 87% of participants showed clinically significant decreases in interference at follow-up sessions whereas no patient in the waiting list showed such changes. Statistical analysis revealed that the changes in interference were significant when comparing pre, post and follow-up treatment, and also when comparing ACT and waiting list groups. Changes in intensity of fear, distress, anxious preoccupation and hypochondria were also observed.Conclusions: Exposure through ‘defusion’ techniques might be considered a useful option for treatment of persistent fears in cancer patients. This study provides evidence for therapies focusing on psychological acceptance in cancer patients through short, simple and feasible therapeutic methods.

Aerodynamic optimization using Adjoint methods and parametric CAD models

This paper describes an implementation of a method capable of integrating parametric, feature based, CAD models based on commercial software (CATIA) with the SU2 software framework. To exploit the adjoint based methods for aerodynamic optimisation within the SU2, a formulation to obtain geometric sensitivities directly from the commercial CAD parameterisation is introduced, enabling the calculation of gradients with respect to CAD based design variables. To assess the accuracy and efficiency of the alternative approach, two aerodynamic optimisation problems are investigated: an inviscid, 3D, problem with multiple constraints, and a 2D high-lift aerofoil, viscous problem without any constraints. Initial results show the new parameterisation obtaining reliable optimums, with similar levels of performance of the software native parameterisations. In the final paper, details of computing CAD sensitivities will be provided, including accuracy as well as linking geometric sensitivities to aerodynamic objective functions and constraints; the impact in the robustness of the overall method will be assessed and alternative parameterisations will be included.

An exploration of men’s experiences of undergoing active surveillance for favourable-risk prostate cancer: A mixed methods study protocol

Background
Prostate cancer is one of the most common male cancers worldwide. Active Surveillance (AS) has been developed to allow men with lower risk disease to postpone or avoid the adverse side effects associated with curative treatments until the disease progresses. Despite the medical benefits of AS, it is reported that living with untreated cancer can create a significant emotional burden for patients.

Methods/design
The aim of this study is to gain insight into the experiences of men eligible to undergo AS for favourable-risk PCa.

This study has a mixed-methods sequential explanatory design consisting of two phases: quantitative followed by qualitative. Phase 1 has a multiple point, prospective, longitudinal exploratory design. Ninety men diagnosed with favourable-risk prostate cancer will be assessed immediately post-diagnosis (baseline) and followed over a period of 12 months, in intervals of 3 month. Ninety age-matched men with no cancer diagnosis will also be recruited using peer nomination and followed up in the same 3 month intervals. Following completion of Phase 1, 10–15 AS participants who have reported both the best and worst psychological functioning will be invited to participate in semi-structured qualitative interviews. Phase 2 will facilitate further exploration of the quantitative results and obtain a richer understanding of participants’ personal interpretations of their illness and psychological wellbeing.

Discussion
To our knowledge, this is the first study to utilise early baseline measures; include a healthy comparison group; calculate sample size through power calculations; and use a mixed methods approach to gain a deeper more holistic insight into the experiences of men diagnosed with favourable-risk prostate cancer.

Application of self-quenched JH consensus primers for real-time quantitative PCR of IGH gene to minimal residual disease evaluation in multiple myeloma.

Monitoring multiple myeloma patients for relapse requires sensitive methods to measure minimal residual disease and to establish a more precise prognosis. The present study aimed to standardize a real-time quantitative polymerase chain reaction (PCR) test for the IgH gene with a JH consensus self-quenched fluorescence reverse primer and a VDJH or DJH allele-specific sense primer (self-quenched PCR). This method was compared with allele-specific real-time quantitative PCR test for the IgH gene using a TaqMan probe and a JH consensus primer (TaqMan PCR). We studied nine multiple myeloma patients from the Spanish group treated with the MM2000 therapeutic protocol. Self-quenched PCR demonstrated sensitivity of >or=10(-4) or 16 genomes in most cases, efficiency was 1.71 to 2.14, and intra-assay and interassay reproducibilities were 1.18 and 0.75%, respectively. Sensitivity, efficiency, and residual disease detection were similar with both PCR methods. TaqMan PCR failed in one case because of a mutation in the JH primer binding site, and self-quenched PCR worked well in this case. In conclusion, self-quenched PCR is a sensitive and reproducible method for quantifying residual disease in multiple myeloma patients; it yields similar results to TaqMan PCR and may be more effective than the latter when somatic mutations are present in the JH intronic primer binding site.

Minimal residual disease monitoring in multiple myeloma: a comparison between allelic-specific oligonucleotide real-time quantitative polymerase chain reaction and flow cytometry.

BACKGROUND AND OBJECTIVES: Minimal residual disease (MRD) studies are useful in multiple myeloma (MM). However, the definition of the best technique and clinical utility are still unresolved issues. The aim of this study was to analyze and compare the clinical utility of MRD studies in MM with two different techniques: allelic-specific oligonucleotide real-time quantitative PCR (ASO-RQ-PCR), and flow cytometry (FCM). DESIGN AND METHODS: Bone marrow samples from 32 MM patients who had achieved complete response after transplantation were evaluated by ASO-RQ-PCR, using TaqMan technology, and multiparametric FCM. RESULTS: ASO-RQ-PCR was only applicable in 75% of patients for a variety of technical reasons, while FCM was applicable in up to 90%. Therefore, simultaneous PCR/FCM analysis was possible in only 24 patients. The number of residual tumor cells identified by both techniques was very similar (mean=0.29%, range=0.001-1.61%, correlation coefficient=0.861). However, RQ-PCR was able to detect residual myelomatous cells in 17 patients while FCM only did so in 11; thus, 6 cases were FCM negative but PCR positive, all of them displaying a very low number of clonal cells (median=0.014%, range=0.001-0.11). Using an MRD threshold of 0.01% (10(-4)) two risk groups with significantly different progression-free survival could be identified by either PCR (34 vs. 15m, p=0.04) or FCM (27 vs. 10m, p=0.05). INTERPRETATION AND CONCLUSIONS: Although MRD evaluation by ASO-RQ-PCR is slightly more sensitive and specific than FCM, it is applicable in a lower proportion of MM patients and is more time-consuming, while both techniques provide similar prognostic information.

Molecular characteristics and gene segment usage in IGH gene rearrangements in multiple myeloma.

BACKGROUND AND OBJECTIVES: Analysis of IgH rearrangements in B-cell malignancies has provided clinical researchers with a wide range of information during the last few years. However, only a few studies have contributed to the characterization of these features in multiple myeloma (MM), and they have been focused on the analysis of the expressed IgH allele only. Comparison between the expressed and the non-functional IgH alleles allows further characterizion of the selection processes to which pre-myeloma cells are submitted. DESIGN AND METHODS: We analyzed a cohort of 84 untreated MM patients in order to characterize their functional VDJH and non-functional DJH rearrangements. The pattern of mutations and gene segment usage for both types of rearrangements was analyzed by polymerase chain reaction and sequencing. RESULTS: VH3 and VH1 family members were over- and under-represented, respectively. VH3-30 and VH3-15 segments were the most frequently used, whereas VH4-34 was found only in non-functional or heavily mutated VDJH rearrangements. DH2 and DH3 family members were over-represented in both VDJH and DJH repertoires, while the DH1 family was under-represented only in the productive VDJH rearrangements. Finally, DH3-22 and DH2-21 gene segments were found to be over-represented in the functional repertoire while segments commonly used by less mature B-cell malignancies, such as DH6-19 or DH3-3, were under-represented. INTERPRETATION AND CONCLUSIONS: Data reported here help to identify the clonogenic MM cell as a post-germinal center B cell that has undergone selection processes during the germinal center reaction.

Heteroduplex PCR analysis of rearranged immunoglobulin genes for clonality assessment in multiple myeloma.

BACKGROUND AND OBJECTIVE: Molecular analysis by PCR of monoclonally rearranged immunoglobulin (Ig) genes can be used for diagnosis in B-cell lymphoproliferative disorders (LPD), as well as for monitoring minimal residual disease (MRD) after treatment. This technique has the risk of false-positive results due to the "background" amplification of similar rearrangements derived from polyclonal B-cells. This problem can be resolved in advance by additional analyses that discern between polyclonal and monoclonal PCR products, such as the heteroduplex analysis. A second problem is that PCR frequently fails to amplify the junction regions, mainly due to somatic mutations frequently present in mature (post-follicular) B-cell lymphoproliferations. The use of additional targets (e.g. Ig light chain genes) can avoid this problem. DESIGN AND METHODS: We studied the specificity of heteroduplex PCR analysis of several Ig junction regions to detect monoclonal products in samples from 84 MM patients and 24 patients with B cell polyclonal disorders. RESULTS: Using two distinct VH consensus primers (FR3 and FR2) in combination with one JH primer, 79% of the MM displayed monoclonal products. The percentage of positive cases was increased by amplification of the Vlamda-Jlamda junction regions or kappa(de) rearrangements, using two or five pairs of consensus primers, respectively. After including these targets in the heteroduplex PCR analysis, 93% of MM cases displayed monoclonal products. None of the polyclonal samples analyzed resulted in monoclonal products. Dilution experiments showed that monoclonal rearrangements could be detected with a sensitivity of at least 10(-2) in a background with >30% polyclonal B-cells, the sensitivity increasing up to 10(-3) when the polyclonal background was

Joint Analysis of Multiple Algorithms and Performance Measures

There has been an increasing interest in the development of new methods using Pareto optimality to deal with multi-objective criteria (for example, accuracy and time complexity). Once one has developed an approach to a problem of interest, the problem is then how to compare it with the state of art. In machine learning, algorithms are typically evaluated by comparing their performance on different data sets by means of statistical tests. Standard tests used for this purpose are able to consider jointly neither performance measures nor multiple competitors at once. The aim of this paper is to resolve these issues by developing statistical procedures that are able to account for multiple competing measures at the same time and to compare multiple algorithms altogether. In particular, we develop two tests: a frequentist procedure based on the generalized likelihood-ratio test and a Bayesian procedure based on a multinomial-Dirichlet conjugate model. We further extend them by discovering conditional independences among measures to reduce the number of parameters of such models, as usually the number of studied cases is very reduced in such comparisons. Data from a comparison among general purpose classifiers is used to show a practical application of our tests.

Multiple training interventions significantly improve reproducibility of PET/CT-based lung cancer radiotherapy target volume delineation using an IAEA study protocol

BACKGROUND AND PURPOSE: To assess the impact of a standardized delineation protocol and training interventions on PET/CT-based target volume delineation (TVD) in NSCLC in a multicenter setting.

MATERIAL AND METHODS: Over a one-year period, 11 pairs, comprised each of a radiation oncologist and nuclear medicine physician with limited experience in PET/CT-based TVD for NSCLC from nine different countries took part in a training program through an International Atomic Energy Agency (IAEA) study (NCT02247713). Teams delineated gross tumor volume of the primary tumor, during and after training interventions, according to a provided delineation protocol. In-house developed software recorded the performed delineations, to allow visual inspection of strategies and to assess delineation accuracy.

RESULTS: Following the first training, overall concordance indices for 3 repetitive cases increased from 0.57±0.07 to 0.66±0.07. The overall mean surface distance between observer and expert contours decreased from -0.40±0.03 cm to -0.01±0.33 cm. After further training overall concordance indices for another 3 repetitive cases further increased from 0.64±0.06 to 0.80±0.05 (p=0.01). Mean surface distances decreased from -0.34±0.16 cm to -0.05±0.20 cm (p=0.01).

CONCLUSION: Multiple training interventions improve PET/CT-based TVD delineation accuracy in NSCLC and reduces interobserver variation.

Prevalence and risk of Down syndrome in monozygotic and dizygotic multiple pregnancies in Europe: implications for prenatal screening

OBJECTIVE: To determine risk of Down syndrome (DS) in multiple relative to singleton pregnancies, and compare prenatal diagnosis rates and pregnancy outcome.

DESIGN: Population-based prevalence study based on EUROCAT congenital anomaly registries.

SETTING: Eight European countries.

POPULATION: 14.8 million births 1990-2009; 2.89% multiple births.

METHODS: DS cases included livebirths, fetal deaths from 20 weeks, and terminations of pregnancy for fetal anomaly (TOPFA). Zygosity is inferred from like/unlike sex for birth denominators, and from concordance for DS cases.

MAIN OUTCOME MEASURES: Relative risk (RR) of DS per fetus/baby from multiple versus singleton pregnancies and per pregnancy in monozygotic/dizygotic versus singleton pregnancies. Proportion of prenatally diagnosed and pregnancy outcome.

STATISTICAL ANALYSIS: Poisson and logistic regression stratified for maternal age, country and time.

RESULTS: Overall, the adjusted (adj) RR of DS for fetus/babies from multiple versus singleton pregnancies was 0.58 (95% CI 0.53-0.62), similar for all maternal ages except for mothers over 44, for whom it was considerably lower. In 8.7% of twin pairs affected by DS, both co-twins were diagnosed with the condition. The adjRR of DS for monozygotic versus singleton pregnancies was 0.34 (95% CI 0.25-0.44) and for dizygotic versus singleton pregnancies 1.34 (95% CI 1.23-1.46). DS fetuses from multiple births were less likely to be prenatally diagnosed than singletons (adjOR 0.62 [95% CI 0.50-0.78]) and following diagnosis less likely to be TOPFA (adjOR 0.40 [95% CI 0.27-0.59]).

CONCLUSIONS: The risk of DS per fetus/baby is lower in multiple than singleton pregnancies. These estimates can be used for genetic counselling and prenatal screening.

Designing low-emission aero-engines using adjoint methods

The aim of this work is to investigate an efficient CAD based adjoint process chain for calculating sensitivities of the objective function to the CAD parameter in time scales acceptable for industrial design processes.

The Precipitationshed : Concepts, Methods, and Applications

Human societies are reliant on the functioning of the hydrologic cycle. The atmospheric branch of this cycle, often referred to as moisture recycling in the context of land-to-land exchange, refers to water evaporating, traveling through the atmosphere, and falling out as precipitation. Similar to the surface water cycle that uses the watershed as the unit of analysis, it is also possible to consider a ‘watershed of the sky’ for the atmospheric water cycle. Thus, I explore the precipitationshed - defined as the upwind surface of the Earth that provides evaporation that later falls as precipitation in a specific place. The primary contributions of this dissertation are to (a) introduce the precipitationshed concept, (b) provide a quantitative basis for the study of the precipitationshed, and (c) demonstrate its use in the fields of hydrometeorology, land-use change, social-ecological systems, ecosystem services, and environmental governance. In Paper I, the concept of the precipitationshed is introduced and explored for the first time. The quantification of precipitationshed variability is described in Paper II, and the key finding is that the precipitationsheds for multiple regions are persistent in time and space. Moisture recycling is further described as an ecosystem service in Paper III, to integrate the concept into the existing language of environmental sustainability and management. That is, I identify regions where vegetation more strongly regulates the provision of atmospheric water, as well as the regions that more strongly benefit from this regulation. In Paper IV, the precipitationshed is further explored through the lens of urban reliance on moisture recycling. Using a novel method, I quantify the vulnerability of urban areas to social-ecological changes within their precipitationsheds. In Paper V, I argue that successful moisture recycling governance will require flexible, transboundary institutions that are capable of operating within complex social-ecological systems. I conclude that, in the future, the precipitationshed can be a key tool in addressing the complexity of social-ecological systems. 

Experiências de Combate e a sua Relação com a Sintomatologia Associada à Perturbação Pós-Stress Traumático em Veteranos da Guerra Colonial Portuguesa

Objectivos: O presente estudo avaliou a relação entre diversos aspectos objectivos e subjectivos inerentes às experiências de combate/guerra e sintomatologia associada à Perturbação Pós-Stress Traumático. Os instrumentos de auto-resposta utilizados na colheita dos dados foram analisados quanto às suas principais propriedades psicométricas. Métodos: A amostra foi composta por 105 combatentes da Guerra Colonial Portuguesa. Foram utilizados os seguintes instrumentos de avaliação: um questionário sóciodemográfico de auto-resposta para a caracterização da amostra, o Posttraumatic Stress Disorder Checklist (para militares), o Peritraumatic Dissociative Experiences Questionnaire, o Questionário de Experiências de Combate e duas escalas do Deployment Risk and Resilience Inventory, nomeadamente a Difficult Livings and Working Environment Scale e a Perceived Threat Scale. Resultados: O estudo das qualidades psicométricas dos instrumentos apresenta uma consistência interna superior a .86 e uma estabilidade temporal superior a .79. A análise factorial do Peritraumatic Dissociative Experiences Questionnaire apresenta bons valores na medida de Kaiser-Meyer-Olkin e no teste de esfericidade de Bartlett. Na rotação de Varimax é extraído um factor, predizendo a sua unidimensionalidade, que explica cerca de 70% da variância total do questionário. Os resultados relativos à validade divergentediscriminante indicam, por si só, que esta se apresenta de forma clara entre os instrumentos estudados (PCL-M, PDEQ, DLWES e PTS) no subgrupo com PTSD, com excepção da verificada entre o PCL-M e o PDEQ.. Quanto à validade de critério, constatam-se diferenças entre os subgrupos com e sem PTSD. Através do método da regressão linear múltipla, duas das seis variáveis têm, isoladamente, um peso preditivo significativo para a PTSD, nomeadamente a frequência de exposição a situações típicas e objectivas de combate/ guerra e os desconfortos diários durante a(s) comissão(ões). Ao considerarmos as variáveis no seu conjunto, aquelas que mais contribuem são a dissociação peritraumática e os desconfortos diários durante a(s) comissão(ões). Conclusão: Constata-se que os factores peritraumáticos em estudo desempenham um importante papel no desenvolvimento da sintomatologia inerente à PTSD. /