A Factor Analysis Perspective on Linear Regression in the 'More Predictors than Samples' Case.

Linear regression (LR) is a core model in supervised machine learning performing a regression task. One can fit this model using either an analytic/closed-form formula or an iterative algorithm. Fitting it via the analytic formula becomes a problem when the number of predictors is greater than the number of samples because the closed-form solution contains a matrix inverse that is not defined when having more predictors than samples. The standard approach to solve this issue is using the Moore-Penrose inverse or the L2 regularization. We propose another solution starting from a machine learning model that, this time, is used in unsupervised learning performing a dimensionality reduction task or just a density estimation one-factor analysis (FA)-with one-dimensional latent space. The density estimation task represents our focus since, in this case, it can fit a Gaussian distribution even if the dimensionality of the data is greater than the number of samples; hence, we obtain this advantage when creating the supervised counterpart of factor analysis, which is linked to linear regression. We also create its semisupervised counterpart and then extend it to be usable with missing data. We prove an equivalence to linear regression and create experiments for each extension of the factor analysis model. The resulting algorithms are either a closed-form solution or an expectation-maximization (EM) algorithm. The latter is linked to information theory by optimizing a function containing a Kullback-Leibler (KL) divergence or the entropy of a random variable.

IL-9 is a key component of memory TH cell peanut-specific responses from children with peanut allergy.

BACKGROUND: Differentiation between patients with peanut allergy (PA) and those with peanut sensitization (PS) who tolerate peanut but have peanut-specific IgE, positive skin prick test responses, or both represents a significant diagnostic difficulty. Previously, gene expression microarrays were successfully used to identify biomarkers and explore immune responses during PA immunotherapy. OBJECTIVE: We aimed to characterize peanut-specific responses from patients with PA, subjects with PS, and atopic children without peanut allergy (NA children). METHODS: A preliminary exploratory microarray investigation of gene expression in peanut-activated memory TH subsets from 3 children with PA and 3 NA children identified potential PA diagnostic biomarkers. Microarray findings were confirmed by using real-time quantitative PCR in 30 subjects (12 children with PA, 12 children with PS, and 6 NA children). Flow cytometry was used to identify the TH subsets involved. RESULTS: Among 12,257 differentially expressed genes, IL9 showed the greatest difference between children with PA and NA children (45.59-fold change, P < .001), followed by IL5 and then IL13. Notably, IL9 allowed the most accurate classification of children with PA and NA children by using a machine-learning approach with recursive feature elimination and the random forest algorithm. Skin- and gut-homing TH cells from donors with PA expressed similar TH2- and TH9-associated genes. Real-time quantitative PCR confirmed that IL9 was the highest differentially expressed gene between children with PA and NA children (23.3-fold change, P < .01) and children with PS (18.5-fold change, P < .05). Intracellular cytokine staining showed that IL-9 and the TH2-specific cytokine IL-5 are produced by distinct TH populations. CONCLUSION: In this study IL9 best differentiated between children with PA and children with PS (and atopic NA children). Mutually exclusive production of IL-9 and the TH2-specific cytokine IL-5 suggests that the IL-9-producing cells belong to the recently described TH9 subset.

MiRNA recognition with the yasMiR system: the quest for further improvements.

The paper "Using Base Pairing Probabilities for MiRNA Recognition" by Daniel Pasaila, Irina Mohorianu, and Liviu Ciortuz, that has been published in Proceedings of the International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC) 2008, IEEE Computer Society, pp. 519-525, has introduced a new SVM for microRNA identification, whose novelty is twofolded: first, many of its features incorporate the base-pairing probabilities provided by McCaskill's algorithm, and second the classification performance is improved using a certain similarity ("profile"-based) measure between the training and test microRNAs and a set of carefully chosen ("pivot") RNA sequences. Comparisons with some of the best existing SVMs for microRNA identification proved that our SVM obtains truly competitive results. Here we add several significant extensions to the work reported in Daniel Pasaila et al. Proceedings of the International (SYNASC) 2008, pp. 519-525: testing this classifier on a more recent version of miRBase (12.0), evaluating the effect of using probabilistic patterns instead of non-probabilistic ones, analysing the discriminative power of different categories of features we used, and automatically searching for good pivot RNA sequences, which are critical for classification in our approach.