Julien Prados

Stability of feature selection algorithms: a study on high-dimensional spaces

With the proliferation of extremely high-dimensional data, feature selection algorithms have become indispensable components of the learning process. Strangely, despite extensive work on the stability of learning algorithms, the stability of feature selection algorithms has been relatively neglected. This study is an attempt to fill that gap by quantifying the sensitivity of feature selection algorithms to variations in the training set. We assess the stability of feature selection algorithms based on the stability of the feature preferences that they express in the form of weights-scores, ranks, or a selected feature subset. We examine a number of measures to quantify the stability of feature preferences and propose an empirical way to estimate them. We perform a series of experiments with several feature selection algorithms on a set of proteomics datasets. The experiments allow us to explore the merits of each stability measure and create stability profiles of the feature selection algorithms. Finally, we show how stability profiles can support the choice of a feature selection algorithm.

Sequential transcriptional waves direct the differentiation of newborn neurons in the mouse neocortex

Tracking neuronal transcriptional programs Early in brain development, cortical neurons are born near the ventricles, then migrate to their functional destinations. Telley et al. used a fluorescent labeling technique to see what transcripts characterize these earliest stages of neural development. Waves of transcriptional programs are initiated, then passed by as the neuron progresses from proliferative to migratory and finally to connectivity phases. Science, this issue p. 1443 In vivo fluorescence labeling reveals neuron-specific primordial transcriptional programs as they dynamically unfold. During corticogenesis, excitatory neurons are born from progenitors located in the ventricular zone (VZ), from where they migrate to assemble into circuits. How neuronal identity is dynamically specified upon progenitor division is unknown. Here, we study this process using a high-temporal-resolution technology allowing fluorescent tagging of isochronic cohorts of newborn VZ cells. By combining this in vivo approach with single-cell transcriptomics in mice, we identify and functionally characterize neuron-specific primordial transcriptional programs as they dynamically unfold. Our results reveal early transcriptional waves that instruct the sequence and pace of neuronal differentiation events, guiding newborn neurons toward their final fate, and contribute to a road map for the reverse engineering of specific classes of cortical neurons from undifferentiated cells.

Stability of feature selection algorithms

With the proliferation of extremely high-dimensional data, feature selection algorithms have become indispensable components of the learning process. Strangely, despite extensive work on the stability of learning algorithms, the stability of feature selection algorithms has been relatively neglected. This study is an attempt to fill that gap by quantifying the sensitivity of feature selection algorithms to variations in the training set. We assess the stability of feature selection algorithms based on the stability of the feature preferences that they express in the form of weights-scores, ranks, or a selected feature subset. We examine a number of measures to quantify the stability of feature preferences and propose an empirical way to estimate them. We perform a series of experiments with several feature selection algorithms on a set of proteomics datasets. The experiments allow us to explore the merits of each stability measure and create stability profiles of the feature selection algorithms. Finally we show how stability profiles can support the choice of a feature selection algorithm.

BDNF promoter I methylation correlates between post-mortem human peripheral and brain tissues

Several psychiatric disorders have been associated with CpG methylation changes in CG rich promoters of the brain-derived neurotrophic factor (BDNF) mainly by extracting DNA from peripheral blood cells. Whether changes in peripheral DNA methylation can be used as a proxy for brain-specific alterations remains an open question. In this study we aimed to compare DNA methylation levels in BDNF promoter regions in human blood cells, muscle and brain regions using bisulfite-pyrosequencing. We found a significant correlation between the levels of BDNF promoter I methylation measured in quadriceps and vPFC tissues extracted from the same individuals (n = 98, Pearson, r = 0.48, p = 4.5 × 10(-7)). In the hippocampus, BDNF promoter I and IV methylation levels were strongly correlated (Pearson, n = 37, r = 0.74, p = 1.4 × 10(-7)). We found evidence for sex-dependent effect on BDNF promoter methylation levels in the various tissues and blood samples. Taken together, these data indicate a strong intra-individual correlation between peripheral and brain tissue. They also suggest that sex determines methylation patterns in BDNF promoter region across different types of tissue, including muscle, brain, and blood.

Decay-Initiating Endoribonucleolytic Cleavage by RNase Y Is Kept under Tight Control via Sequence Preference and Sub-cellular Localisation.

Bacteria depend on efficient RNA turnover, both during homeostasis and when rapidly altering gene expression in response to changes. Nevertheless, remarkably few details are known about the rate-limiting steps in targeting and decay of RNA. The membrane-anchored endoribonuclease RNase Y is a virulence factor in Gram-positive pathogens. We have obtained a global picture of Staphylococcus aureus RNase Y sequence specificity using RNA-seq and the novel transcriptome-wide EMOTE method. Ninety-nine endoribonucleolytic sites produced in vivo were precisely mapped, notably inside six out of seven genes whose half-lives increase the most in an RNase Y deletion mutant, and additionally in three separate transcripts encoding degradation ribonucleases, including RNase Y itself, suggesting a regulatory network. We show that RNase Y is required to initiate the major degradation pathway of about a hundred transcripts that are inaccessible to other ribonucleases, but is prevented from promiscuous activity by membrane confinement and sequence preference for guanosines.

Borderline personality disorder and childhood maltreatment: a genome‐wide methylation analysis

Early life adversity plays a critical role in the emergence of borderline personality disorder (BPD) and this could occur through epigenetic programming. In this perspective, we aimed to determine whether childhood maltreatment could durably modify epigenetic processes by the means of a whole‐genome methylation scan of BPD subjects. Using the Illumina Infinium® HumanMethylation450 BeadChip, global methylation status of DNA extracted from peripheral blood leucocytes was correlated to the severity of childhood maltreatment in 96 BPD subjects suffering from a high level of child adversity and 93 subjects suffering from major depressive disorder (MDD) and reporting a low rate of child maltreatment. Several CpGs within or near the following genes (IL17RA, miR124‐3, KCNQ2, EFNB1, OCA2, MFAP2, RPH3AL, WDR60, CST9L, EP400, A2ML1, NT5DC2, FAM163A and SPSB2) were found to be differently methylated, either in BPD compared with MDD or in relation to the severity of childhood maltreatment. A highly relevant biological result was observed for cg04927004 close to miR124‐3 that was significantly associated with BPD and severity of childhood maltreatment. miR124‐3 codes for a microRNA (miRNA) targeting several genes previously found to be associated with BPD such as NR3C1. Our results highlight the potentially important role played by miRNAs in the etiology of neuropsychiatric disorders such as BPD and the usefulness of using methylome‐wide association studies to uncover such candidate genes. Moreover, they offer new understanding of the impact of maltreatments on biological processes leading to diseases and may ultimately result in the identification of relevant biomarkers.

The C-terminal region of the RNA helicase CshA is required for the interaction with the degradosome and turnover of bulk RNA in the opportunistic pathogen Staphylococcus aureus

Staphylococcus aureus is a versatile opportunistic pathogen that adapts readily to a variety of different growth conditions. This adaptation requires a rapid regulation of gene expression including the control of mRNA abundance. The CshA DEAD-box RNA helicase was previously shown to be required for efficient turnover of the agr quorum sensing mRNA. Here we show by transcriptome-wide RNA sequencing and microarray analyses that CshA is required for the degradation of bulk mRNA. Moreover a subset of mRNAs is significantly stabilised in absence of CshA. Deletion of the C-terminal extension affects RNA turnover similar to the full deletion of the cshA gene. In accordance with RNA decay data, the C-terminal region of CshA is required for an RNA-independent interaction with components of the RNA degradation machinery. The C-terminal truncation of CshA reduces its ATPase activity and this reduction cannot be compensated at high RNA concentrations. Finally, the deletion of the C-terminal extension does affect growth at low temperatures, but to a significantly lesser degree than the full deletion, indicating that the core of the helicase can assume a partial function and opening the possibility that CshA is involved in different cellular processes.

Prenatal Exposure to DEHP Affects Spermatogenesis and Sperm DNA Methylation in a Strain-Dependent Manner

Di-(2-ethylhexyl)phtalate (DEHP) is a plasticizer with endocrine disrupting properties found ubiquitously in the environment and altering reproduction in rodents. Here we investigated the impact of prenatal exposure to DEHP on spermatogenesis and DNA sperm methylation in two distinct, selected, and sequenced mice strains. FVB/N and C57BL/6J mice were orally exposed to 300 mg/kg/day of DEHP from gestation day 9 to 19. Prenatal DEHP exposure significantly decreased spermatogenesis in C57BL/6J (fold-change = 0.6, p-value = 8.7*10-4), but not in FVB/N (fold-change = 1, p-value = 0.9). The number of differentially methylated regions (DMRs) by DEHP-exposure across the entire genome showed increased hyper- and decreased hypo-methylation in C57BL/6J compared to FVB/N. At the promoter level, three important subsets of genes were massively affected. Promoters of vomeronasal and olfactory receptors coding genes globally followed the same trend, more pronounced in the C57BL/6J strain, of being hyper-methylated in DEHP related conditions. In contrast, a large set of micro-RNAs were hypo-methylated, with a trend more pronounced in the FVB/N strain. We additionally analyze both the presence of functional genetic variations within genes that were associated with the detected DMRs and that could be involved in spermatogenesis, and DMRs related with the DEHP exposure that affected both strains in an opposite manner. The major finding in this study indicates that prenatal exposure to DEHP can decrease spermatogenesis in a strain-dependent manner and affects sperm DNA methylation in promoters of large sets of genes putatively involved in both sperm chemotaxis and post-transcriptional regulatory mechanisms.

Transcriptomic and anatomic parcellation of 5-HT3AR expressing cortical interneuron subtypes revealed by single-cell RNA sequencing

Cortical GABAergic interneurons constitute a highly diverse population of inhibitory neurons that are key regulators of cortical microcircuit function. An important and heterogeneous group of cortical interneurons specifically expresses the serotonin receptor 3A (5-HT3AR) but how this diversity emerges during development is poorly understood. Here we use single-cell transcriptomics to identify gene expression patterns operating in Htr3a-GFP+ interneurons during early steps of cortical circuit assembly. We identify three main molecular types of Htr3a-GFP+ interneurons, each displaying distinct developmental dynamics of gene expression. The transcription factor Meis2 is specifically enriched in a type of Htr3a-GFP+ interneurons largely confined to the cortical white matter. These MEIS2-expressing interneurons appear to originate from a restricted region located at the embryonic pallial–subpallial boundary. Overall, this study identifies MEIS2 as a subclass-specific marker for 5-HT3AR-containing interstitial interneurons and demonstrates that the transcriptional and anatomical parcellation of cortical interneurons is developmentally coupled.

On Preprocessing of SELDI-MS Data and its Evaluation

Mass spectrometry is becoming an important tool in proteomics. Mass spectral data are characterized by very high dimensionality and a high level of redundancy. Both issues are quite challenging when one wants to perform knowledge discovery and push existing tools to their limits. We tackle both via a preprocessing pipeline that drastically reduces dimensionality and redundancy of the initial representation in order to focus on biologically relevant information. Essentially preprocessing performs feature extraction in a manner that reflects domain knowledge. We propose a framework for the evaluation of the given pipeline and in fact of any mass spectrometry preprocessing pipeline which is based on the level of conservation of discriminatory information. The discriminatory information content of a given representation is objectively measured by the classification performance of a number of classification algorithms evaluated on the given representation. This approach also allows us to compare a number of different preprocessing possibilities, namely using peak intensities vs peak areas to represent peaks and how non observed peaks should be treated, and demonstrate which is the most informative one