An OMICs-based meta-analysis to support infection state stratification.

MOTIVATION: A fundamental problem for disease treatment is that while antibiotics are a powerful counter to bacteria, they are ineffective against viruses. Often, bacterial and viral infections are confused due to their similar symptoms and lack of rapid diagnostics. With many clinicians relying primarily on symptoms for diagnosis, overuse and misuse of modern antibiotics are rife, contributing to the growing pool of antibiotic resistance. To ensure an individual receives optimal treatment given their disease state and to reduce over-prescription of antibiotics, the host response can in theory be measured quickly to distinguish between the two states. To establish a predictive biomarker panel of disease state (viral/bacterial/no-infection), we conducted a meta-analysis of human blood infection studies using machine learning. RESULTS: We focused on publicly available gene expression data from two widely used platforms, Affymetrix and Illumina microarrays as they represented a significant proportion of the available data. We were able to develop multi-class models with high accuracies with our best model predicting 93% of bacterial and 89% viral samples correctly. To compare the selected features in each of the different technologies, we reverse-engineered the underlying molecular regulatory network and explored the neighbourhood of the selected features. The networks highlighted that although on the gene-level the models differed, they contained genes from the same areas of the network. Specifically, this convergence was to pathways including the Type I interferon Signalling Pathway, Chemotaxis, Apoptotic Processes and Inflammatory/Innate Response. AVAILABILITY: Data and code are available on the Gene Expression Omnibus and github. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Hyperparameter Learning Under Data Poisoning: Analysis of the Influence of Regularization via Multiobjective Bilevel Optimization.

Machine learning (ML) algorithms are vulnerable to poisoning attacks, where a fraction of the training data is manipulated to deliberately degrade the algorithms' performance. Optimal attacks can be formulated as bilevel optimization problems and help to assess their robustness in worst case scenarios. We show that current approaches, which typically assume that hyperparameters remain constant, lead to an overly pessimistic view of the algorithms' robustness and of the impact of regularization. We propose a novel optimal attack formulation that considers the effect of the attack on the hyperparameters and models the attack as a multiobjective bilevel optimization problem. This allows us to formulate optimal attacks, learn hyperparameters, and evaluate robustness under worst case conditions. We apply this attack formulation to several ML classifiers using L2 and L1 regularization. Our evaluation on multiple datasets shows that choosing an "a priori" constant value for the regularization hyperparameter can be detrimental to the performance of the algorithms. This confirms the limitations of previous strategies and evidences the benefits of using L2 and L1 regularization to dampen the effect of poisoning attacks, when hyperparameters are learned using a small trusted dataset. Additionally, our results show that the use of regularization plays an important robustness and stability role in complex models, such as deep neural networks (DNNs), where the attacker can have more flexibility to manipulate the decision boundary.

Spectroscopic analysis of bacterial biological warfare simulants and the effects of environmental conditioning on a bacterial spectrum.

The ability to distinguish bacteria from mixed samples is of great interest, especially in the medical and defence arenas. This paper reports a step towards the aim of differentiating pathogenic endospores in situ, to aid any required response for hazard management using infrared spectroscopy combined with multivariate analysis. We describe a proof-of-principle study aimed at discriminating biological warfare simulants from common environmental bacteria. We also report an evaluation of multiple pre-processing techniques and subsequent differences in cross-validation of two pattern recognition models (Support Vector Machines and Principal Component-Linear Discriminant Analysis) for a six-class classification (bacterial classification). These classifications were possible with an average sensitivity of 88.0 and 86.9 %, and an average specificity of 97.6 and 97.5 % for the SVM and the PC-LDA models, respectively. Most spectroscopic models are built upon spectra from bacteria that have been specifically prepared for analysis by a particular method; this paper will comment upon the differences in the bacterial spectrum that occur between specific preparations when the bacteria have spent 30 days in the simulated weather conditions of a hot dry climate.

Comparison of eight methods for the extraction of Bacillus atrophaeus spore DNA from eleven common interferents and a common swab.

Eight DNA extraction products or methods (Applied Biosystems PrepFiler Forensic DNA Extraction Kit; Bio-Rad Instagene Only, Bio-Rad Instagene & Spin Column Purification; EpiCentre MasterPure DNA & RNA Kit; FujiFilm QuickGene Mini80; Idaho Technologies 1-2-3 Q-Flow Kit; MoBio UltraClean Microbial DNA Isolation Kit; Sigma Extract-N-Amp Plant and Seed Kit) were adapted to facilitate extraction of DNA under BSL3 containment conditions. DNA was extracted from 12 common interferents or sample types, spiked with spores of Bacillus atropheaus. Resulting extracts were tested by real-time PCR. No one method was the best, in terms of DNA extraction, across all sample types. Statistical analysis indicated that the PrepFiler method was the best method from six dry powders (baking, biological washing, milk, plain flour, filler and talcum) and one solid (Underarm deodorant), the UltraClean method was the best from four liquids (aftershave, cola, nutrient broth, vinegar), and the MasterPure method was the best from the swab sample type. The best overall method, in terms of DNA extraction, across all sample types evaluated was the UltraClean method.