Unleashing high content screening in hit detection - Benchmarking AI workflows including novelty detection.

Complex mixtures containing natural products are still an interesting source of novel drug candidates. High content screening (HCS) is a popular tool to screen for such. In particular, multiplexed HCS assays promise comprehensive bioactivity profiles, but generate also high amounts of data. Yet, only some machine learning (ML) applications for data analysis are available and these usually require a profound knowledge of the underlying cell biology. Unfortunately, there are no applications that simply predict if samples are biologically active or not (any kind of bioactivity). Within this work, we benchmark ML algorithms for binary classification, starting with classical ML models, which are the standard classifiers of the scikit-learn library or ensemble models of these classifiers (a total of 92 models tested). Followed by a partial least square regression (PLSR)-based classification (44 tested models in total) and simple artificial neural networks (ANNs) with dense layers (72 tested models in total). In addition, a novelty detection (ND) was examined, which is supposed to handle unknown patterns. For the final analysis the models, with and without upstream ND, were tested with two independent data sets. In our analysis, a stacking model, an ensamble model of class ML algorithms, performed best to predict new and unknown data. ND improved the predictions of the models and was useful to handle unknown patterns. Importantly, the classifier presented here can be easily rebuilt and be adapted to the data and demands of other groups. The hit detector (ND + stacking model) is universal and suitable for a broader application to support the search for new drug candidates.

Nicotine reverses hypofrontality in animal models of addiction and schizophrenia.

The prefrontal cortex (PFC) underlies higher cognitive processes that are modulated by nicotinic acetylcholine receptor (nAChR) activation by cholinergic inputs. PFC spontaneous default activity is altered in neuropsychiatric disorders, including schizophrenia-a disorder that can be accompanied by heavy smoking. Recently, genome-wide association studies (GWAS) identified single-nucleotide polymorphisms (SNPs) in the human CHRNA5 gene, encoding the α5 nAChR subunit, that increase the risks for both smoking and schizophrenia. Mice with altered nAChR gene function exhibit PFC-dependent behavioral deficits, but it is unknown how the corresponding human polymorphisms alter the cellular and circuit mechanisms underlying behavior. Here we show that mice expressing a human α5 SNP exhibit neurocognitive behavioral deficits in social interaction and sensorimotor gating tasks. Two-photon calcium imaging in awake mouse models showed that nicotine can differentially influence PFC pyramidal cell activity by nAChR modulation of layer II/III hierarchical inhibitory circuits. In α5-SNP-expressing and α5-knockout mice, lower activity of vasoactive intestinal polypeptide (VIP) interneurons resulted in an increased somatostatin (SOM) interneuron inhibitory drive over layer II/III pyramidal neurons. The decreased activity observed in α5-SNP-expressing mice resembles the hypofrontality observed in patients with psychiatric disorders, including schizophrenia and addiction. Chronic nicotine administration reversed this hypofrontality, suggesting that administration of nicotine may represent a therapeutic strategy for the treatment of schizophrenia, and a physiological basis for the tendency of patients with schizophrenia to self-medicate by smoking.

DI-ICR-FT-MS-based high-throughput deep metabotyping: a case study of the Caenorhabditis elegans-Pseudomonas aeruginosa infection model.

In metabolomics there is an ever-growing need for faster and more comprehensive analysis methods to cope with the increasing size of biological studies. Direct-infusion ion-cyclotron-resonance Fourier-transform spectrometry (DI-ICR-FT-MS) is used in non-targeted metabolomics to obtain high-resolution snapshots of the metabolic state of a system. We applied this technology to a Caenorhabditis elegans-Pseudomonas aeruginosa infection model and optimized times needed for cultivation and mass-spectrometric analysis. Our results reveal that DI-ICR-FT-MS is a promising tool for high-throughput in-depth non-targeted metabolomics. We performed whole-worm metabolomics and recovered markers of the induced metabolic changes in C. elegans brought about by interaction with pathogens. In this investigation, we reveal complex metabolic phenotypes enabling clustering based upon challenge. Specifically, we observed a marked decrease in amino-acid metabolism with infection by P. aeruginosa and a marked increase in sugar metabolism with infection by Salmonella enterica. We were also able to discriminate between infection with a virulent wild-type Pseudomonas and with an attenuated mutant, making it possible to use this method in larger genetic screens to identify host and pathogen effectors affecting the metabolic phenotype of infection.

Distinct signatures of host-microbial meta-metabolome and gut microbiome in two C57BL/6 strains under high-fat diet.

A combinatory approach using metabolomics and gut microbiome analysis techniques was performed to unravel the nature and specificity of metabolic profiles related to gut ecology in obesity. This study focused on gut and liver metabolomics of two different mouse strains, the C57BL/6J (C57J) and the C57BL/6N (C57N) fed with high-fat diet (HFD) for 3 weeks, causing diet-induced obesity in C57N, but not in C57J mice. Furthermore, a 16S-ribosomal RNA comparative sequence analysis using 454 pyrosequencing detected significant differences between the microbiome of the two strains on phylum level for Firmicutes, Deferribacteres and Proteobacteria that propose an essential role of the microbiome in obesity susceptibility. Gut microbial and liver metabolomics were followed by a combinatory approach using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) and ultra performance liquid chromatography time of tlight MS/MS with subsequent multivariate statistical analysis, revealing distinctive host and microbial metabolome patterns between the C57J and the C57N strain. Many taurine-conjugated bile acids (TBAs) were significantly elevated in the cecum and decreased in liver samples from the C57J phenotype likely displaying different energy utilization behavior by the bacterial community and the host. Furthermore, several metabolite groups could specifically be associated with the C57N phenotype involving fatty acids, eicosanoids and urobilinoids. The mass differences based metabolite network approach enabled to extend the range of known metabolites to important bile acids (BAs) and novel taurine conjugates specific for both strains. In summary, our study showed clear alterations of the metabolome in the gastrointestinal tract and liver within a HFD-induced obesity mouse model in relation to the host-microbial nutritional adaptation.

Metabolic features of Protochlamydia amoebophila elementary bodies--a link between activity and infectivity in Chlamydiae.

The Chlamydiae are a highly successful group of obligate intracellular bacteria, whose members are remarkably diverse, ranging from major pathogens of humans and animals to symbionts of ubiquitous protozoa. While their infective developmental stage, the elementary body (EB), has long been accepted to be completely metabolically inert, it has recently been shown to sustain some activities, including uptake of amino acids and protein biosynthesis. In the current study, we performed an in-depth characterization of the metabolic capabilities of EBs of the amoeba symbiont Protochlamydia amoebophila. A combined metabolomics approach, including fluorescence microscopy-based assays, isotope-ratio mass spectrometry (IRMS), ion cyclotron resonance Fourier transform mass spectrometry (ICR/FT-MS), and ultra-performance liquid chromatography mass spectrometry (UPLC-MS) was conducted, with a particular focus on the central carbon metabolism. In addition, the effect of nutrient deprivation on chlamydial infectivity was analyzed. Our investigations revealed that host-free P. amoebophila EBs maintain respiratory activity and metabolize D-glucose, including substrate uptake as well as host-free synthesis of labeled metabolites and release of labeled CO2 from (13)C-labeled D-glucose. The pentose phosphate pathway was identified as major route of D-glucose catabolism and host-independent activity of the tricarboxylic acid (TCA) cycle was observed. Our data strongly suggest anabolic reactions in P. amoebophila EBs and demonstrate that under the applied conditions D-glucose availability is essential to sustain metabolic activity. Replacement of this substrate by L-glucose, a non-metabolizable sugar, led to a rapid decline in the number of infectious particles. Likewise, infectivity of Chlamydia trachomatis, a major human pathogen, also declined more rapidly in the absence of nutrients. Collectively, these findings demonstrate that D-glucose is utilized by P. amoebophila EBs and provide evidence that metabolic activity in the extracellular stage of chlamydiae is of major biological relevance as it is a critical factor affecting maintenance of infectivity.

High metabolomic microdiversity within co-occurring isolates of the extremely halophilic bacterium Salinibacter ruber.

Salinibacter ruber is an extremely halophilic member of the Bacteroidetes that thrives in crystallizer ponds worldwide. Here, we have analyzed two sets of 22 and 35 co-occurring S. ruber strains, newly isolated respectively, from 100 microliters water samples from crystalizer ponds in Santa Pola and Mallorca, located in coastal and inland Mediterranean Spain and 350 km apart from each other. A set of old strains isolated from the same setting were included in the analysis. Genomic and taxonomy relatedness of the strains were analyzed by means of PFGE and MALDI-TOF, respectively, while their metabolomic potential was explored with high resolution ion cyclotron resonance Fourier transform mass spectrometry (ICR-FT/MS). Overall our results show a phylogenetically very homogeneous species expressing a very diverse metabolomic pool. The combination of MALDI-TOF and PFGE provides, for the newly isolated strains, the same scenario presented by the previous studies of intra-specific diversity of S. ruber using a more restricted number of strains: the species seems to be very homogeneous at the ribosomal level while the genomic diversity encountered was rather high since no identical genome patterns could be retrieved from each of the samples. The high analytical mass resolution of ICR-FT/MS enabled the description of thousands of putative metabolites from which to date only few can be annotated in databases. Some metabolomic differences, mainly related to lipid metabolism and antibiotic-related compounds, provided enough specificity to delineate different clusters within the co-occurring strains. In addition, metabolomic differences were found between old and new strains isolated from the same ponds that could be related to extended exposure to laboratory conditions.

Liquid chromatography-mass spectrometry in metabolomics research: mass analyzers in ultra high pressure liquid chromatography coupling.

The present review gives an introduction into the concept of metabolomics and provides an overview of the analytical tools applied in non-targeted metabolomics with a focus on liquid chromatography (LC). LC is a powerful analytical tool in the study of complex sample matrices. A further development and configuration employing Ultra-High Pressure Liquid Chromatography (UHPLC) is optimized to provide the largest known liquid chromatographic resolution and peak capacity. Reasonably UHPLC plays an important role in separation and consequent metabolite identification of complex molecular mixtures such as bio-fluids. The most sensitive detectors for these purposes are mass spectrometers. Almost any mass analyzer can be optimized to identify and quantify small pre-defined sets of targets; however, the number of analytes in metabolomics is far greater. Optimized protocols for quantification of large sets of targets may be rendered inapplicable. Results on small target set analyses on different sample matrices are easily comparable with each other. In non-targeted metabolomics there is almost no analytical method which is applicable to all different matrices due to limitations pertaining to mass analyzers and chromatographic tools. The specifications of the most important interfaces and mass analyzers are discussed. We additionally provide an exemplary application in order to demonstrate the level of complexity which remains intractable up to date. The potential of coupling a high field Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (ICR-FT/MS), the mass analyzer with the largest known mass resolving power, to UHPLC is given with an example of one human pre-treated plasma sample. This experimental example illustrates one way of overcoming the necessity of faster scanning rates in the coupling with UHPLC. The experiment enabled the extraction of thousands of features (analytical signals). A small subset of this compositional space could be mapped into a mass difference network whose topology shows specificity toward putative metabolite classes and retention time.

Molecular cartography in acute Chlamydia pneumoniae infections--a non-targeted metabolomics approach.

Infections with Chlamydia pneumoniae cause several respiratory diseases, such as community-acquired pneumonia, bronchitis or sinusitis. Here, we present an integrated non-targeted metabolomics analysis applying ultra-high-resolution mass spectrometry and ultra-performance liquid chromatography mass spectrometry to determine metabolite alterations in C. pneumoniae-infected HEp-2 cells. Most important permutations are elaborated using uni- and multivariate statistical analysis, logD retention time regression and mass defect-based network analysis. Classes of metabolites showing high variations upon infection are lipids, carbohydrates and amino acids. Moreover, we observed several non-annotated compounds as predominantly abundant after infection, which are promising biomarker candidates for drug-target and diagnostic research.