Quality assurance of polyolefin regrinds requires assessment and control of their odor properties. Aiming to develop efficient tools for this purpose, we applied analytical and sensory methods to investigate odor properties in 51 virgin and recycled plastics, primarily polypropylene. We report on the odor characteristics of polyolefin materials with a differentiation between virgin and post-consumer recycled material, the latter exhibiting higher odor intensities. Gas chromatography combined with software-based odorant detection facilitated a comparison of samples and indicated higher odorant loads in regrinds. Chemometric tools using sensory and analytical datasets classified samples from virgin and PCR material with prediction accuracies of 97.2-100 %. The prediction of overall odor intensity from analytical data using a support vector machine model delivered root mean square errors of 0.46 (training) and 0.22 (test). This represents a crucial step towards instrumental measurement for odor monitoring in plastics, especially in view of the increasing reuse of post-consumer materials.

In this paper we present a method that allows leveraging 3D electron density information to train a deep neural network pipeline to segment regions of high, medium and low electronegativity and classify substances as health hazardous or non-hazardous. We show that this can be used for use-cases such as cosmetics and food products. For this purpose, we first generate 3D electron density cubes using semiempirical molecular calculations for a custom European Chemicals Agency (ECHA) subset consisting of substances labelled as hazardous and non-hazardous for cosmetic usage. Together with their 3-class electronegativity maps we train a modified 3D-UNet with electron density cubes to segment reactive sites in molecules and classify substances with an accuracy of 78.1%. We perform the same process on a custom food dataset (CompFood) consisting of hazardous and non-hazardous substances compiled from European Food Safety Authority (EFSA) OpenFoodTox, Food and Drug Administration (FDA) Generally Recognized as Safe (GRAS) and FooDB datasets to achieve a classification accuracy of 64.1%. Our results show that 3D electron densities and particularly masked electron densities, calculated by taking a product of original electron densities and regions of high and low electronegativity can be used to classify molecules for different use-cases and thus serve not only to guide safe-by-design product development but also aid in regulatory decisions. SCIENTIFIC CONTRIBUTION: We aim to contribute to the diverse 3D molecular representations used for training machine learning algorithms by showing that a deep learning network can be trained on 3D electron density representation of molecules. This approach has previously not been used to train machine learning models and it allows utilization of the true spatial domain of the molecule for prediction of properties such as their suitability for usage in cosmetics and food products and in future, to other molecular properties. The data and code used for training is accessible at https://github.com/s-singh-ivv/eDen-Substances .

Aroma compositions are usually complex mixtures of odor-active compounds exhibiting diverse molecular structures. Due to chemical interactions of these compounds in the olfactory system, assessing or even predicting the olfactory quality of such mixtures is a difficult task, not only for statistical models, but even for trained assessors. Here, we combine fast automated analytical assessment tools with human sensory data of 11 experienced panelists and machine learning algorithms. Using 16 previously analyzed whisky samples (American or Scotch origin), we apply the linear classifier OWSum to distinguish the samples based on their detected molecules and to gain insights into the key molecular structure characteristics and odor descriptors for sample type. Moreover, we use OWSum and a Convolutional Neural Network (CNN) architecture to classify the five most relevant odor attributes of each sample and predict their sensory scores with promising accuracies (up to F1: 0.71, MCC: 0.68, ROCAUC: 0.78). The predictions outperform the inter-panelist agreement and thus demonstrate previously impossible data-driven sensory assessment in mixtures.

Exhaustive analysis of chemical measurements requires considerable expenditure of time and personnel. However, many aspects of this can be automated by translating manual work into objective algorithmic routines. To this end, we developed adaptable software for gas chromatography data and validated analysis steps using whisky samples. We employed an unspecific, larger, in-house volatile organic compound (VOC) database and another specifically curated database of 217 known whisky chemicals, to automate database-matching based on mass spectra and retention indices. We managed to reduce the amount of necessary interaction, facilitated complex analysis for the less experienced user, and showed that characteristic whisky components constituted the majority of detected molecules in all 16 analyzed samples. With this approach, we present a decisive contribution towards the automated assessment of aroma profiles in food.

We derived and implemented a linear classification algorithm for the prediction of a molecule's odor, called Olfactory Weighted Sum (OWSum). Our approach relies solely on structural patterns of the molecules as features for algorithmic treatment and uses conditional probabilities combined with tf-idf values. In addition to the prediction of molecular odor, OWSum provides insights into properties of the dataset and allows to understand how algorithmic classifications are reached by quantitatively assigning structural patterns to odors. This provides chemists with an intuitive understanding of underlying interactions. To deal with ambiguities of the natural language used to describe odor, we introduced descriptor overlap as a metric for the quantification of semantic overlap between descriptors. Thus, grouping of descriptors and derivation of higher-level descriptors becomes possible. Our approach poses a large leap forward in our capabilities to understand and predict molecular features.

A possible way to reduce the size and complexity of common gas chromatography (GC) systems is the economization of the column temperature regulation system. To this end, a temperature compensation method was developed and validated on a benchtop GC-PDD (pulsed discharge detector) with ethene. An in-house-developed algorithm correlates the retention index of a test gas to the retention index of a previously selected reference gas. To investigate further methods of cost reduction, commercial gas sensors were tested as cheap, sensitive, and versatile detectors. Therefore, CO2 was chosen as a naturally occurring reference gas, while ethene was chosen as a maturity marker for climacteric fruits and hence as a test gas. A demonstrator, consisting of a simple syringe injection system, a PLOT (porous layer open tubular) column boxed in a polystyrene-foam housing, a commercial MOS (metal-oxide semiconductor) sensor for the test gas, and a CO2-specific IR (infrared) sensor, was used to set up a simple GC system and to apply this method on test measurements. Sorption parameters for ethene and CO2 were determined via a van 't Hoff plot, where the entropy S was -11.982 J mol(-1) K-1 (Delta S-Ethene(0)) and 1.351 J mol(-1) K-1 (Delta S-Carbon dioxide(0)), and the enthalpy H was -20.622 kJ mol(-1) (Delta H-Ethene(0)) and -14.792 kJ mol(-1) (Delta H-Carbon dioxide(0)), respectively. Ethene (100 ppm) measurements revealed a system-specific correction term of 0.652 min. Further measurements of ethene and interfering gases revealed a mean retention time for ethene of 3.093 min; the mean predicted retention time is 3.099 min. The demonstrator was able to identify the test gas, ethene, as a function of the reference gas, CO2, in a first approach, without a column heating system and in a gas mixture by applying a temperature compensation algorithm and a system-specific holdup time correction term.

The evaluation of aroma properties of beverages, both analytically as well as with human sensory studies, is a challenging task and most often related to time- and cost-intensive analyses. Whisky is a spirit offering a wide variety of aroma impressions caused by a complex mixture of aroma active compounds. In the present study, methods for the efficient evaluation of aroma characteristics are evaluated for 16 whisky samples of different origins (Scotch and American). Rate all that apply (RATA) was applied as a rapid method for the sensory evaluation of whiskies. Sensory evaluation of the samples led to the determination of eight significant aroma attributes: caramel-/cream caramel-/toffee-like, vanilla-like, (canned) peach-like, phenolic, smoky, fruity, flowery and (fermented) apple-/cider-like. Chemical analysis was conducted by stir bar sorptive extraction (SBSE) coupled to gas chromatography-mass spectrometry in combination with an in-house data processing tool for semi-automated analyte detection. Through chemical analysis of the whisky samples and automated compound detection, we report over 200 mostly aroma-active volatiles. To test both approaches for their potential for sample classification, a simple classification problem (Scotch vs. American) was applied. Linear discriminant analysis (LDA) indicates both that sensory evaluation by RATA (97.86%) and the applied analytical procedure (96.94%) are suitable for the distinction between the two whisky types. In both approaches, potential markers were determined for the classification. These investigations build a solid foundation for the implementation of a versatile platform facilitating rapid and efficient aroma evaluation in various foodstuffs and beverages.

Mimicking the human sense of olfaction with technical systems is a challenging task that requires consideration and understanding of biological olfactory systems, high quality gas analysis, sensor technology and data science. We present an approach for such an artificial intelligent olfactory system, which is just as the olfactory system of humans divided in four units: a sampling unit, a gas chromatograph, the detector/sensor unit and an artificial intelligence as brain. Model-based approaches for characterizing tools and system compounds are used to optimize the sampling of odorants, to predict the chromatographical behavior of analytes and to find the ideal composition of system compounds for an application-specific sensor system. This is complemented with machine-learning software for discriminating and identifying single compounds in an odor mixture. Therefore, a comprehensive toolbox, consisting of different software tools as well as hardware tools is considered as the key to artificial intelligent olfactory systems. The integration of these four units brings together all information to make sense of what is measured. The four system compounds have been evaluated and optimized individually. The interplay between the components has been described and herein 1 we demonstrate our roadmap towards a small, inexpensive system.

Synaptic transmission relies on presynaptic neurotransmitter (NT) release from synaptic vesicles (SVs) and on NT detection by postsynaptic receptors. Transmission exists in two principal modes: action-potential (AP) evoked and AP-independent, “spontaneous” transmission. AP-evoked neurotransmission is considered the primary mode of inter-neuronal communication, whereas spontaneous transmission is required for neuronal development, homeostasis, and plasticity. While some synapses appear dedicated to spontaneous transmission only, all AP-responsive synapses also engage spontaneously, but whether this encodes functional information regarding their excitability is unknown. Here we report on functional interdependence of both transmission modes at individual synaptic contacts of Drosophila larval neuromuscular junctions (NMJs) which were identified by the presynaptic scaffolding protein Bruchpilot (BRP) and whose activities were quantified using the genetically encoded Ca 2+ indicator GCaMP. Consistent with the role of BRP in organizing the AP-dependent release machinery (voltage-dependent Ca 2+ channels and SV fusion machinery), most active BRP-positive synapses (>85%) responded to APs. At these synapses, the level of spontaneous activity was a predictor for their responsiveness to AP-stimulation. AP-stimulation resulted in cross-depletion of spontaneous activity and both transmission modes were affected by the non-specific Ca 2+ channel blocker cadmium and engaged overlapping postsynaptic receptors. Thus, by using overlapping machinery, spontaneous transmission is a continuous, stimulus independent predictor for the AP-responsiveness of individual synapses.

Chemical synaptic transmission relies on the Ca2+-induced fusion of transmitter-laden vesicles whose coupling distance to Ca2+ channels determines synaptic release probability and short-term plasticity, the facilitation or depression of repetitive responses. Here, using electron- and super-resolution microscopy at the Drosophila neuromuscular junction we quantitatively map vesicle:Ca2+ channel coupling distances. These are very heterogeneous, resulting in a broad spectrum of vesicular release probabilities within synapses. Stochastic simulations of transmitter release from vesicles placed according to this distribution revealed strong constraints on short-term plasticity; particularly facilitation was difficult to achieve. We show that postulated facilitation mechanisms operating via activity-dependent changes of vesicular release probability (e.g. by a facilitation fusion sensor) generate too little facilitation and too much variance. In contrast, Ca2+-dependent mechanisms rapidly increasing the number of releasable vesicles reliably reproduce short-term plasticity and variance of synaptic responses. We propose activity-dependent inhibition of vesicle un-priming or release site activation as novel facilitation mechanisms.

Neuronal synapses transduce information via the consecutive action of three transducers: voltage-gated Ca2+-channels, fusion-competent synaptic vesicles, and postsynaptic receptors. Their physical distance is thought to influence the speed and efficiency of neurotransmission. However, technical limitations have hampered resolving their nanoscale arrangement. Here, we developed a new method for live-labeling proteins for electron microscopy (EM), revealing that release-competent vesicles preferentially align with Ca2+-channels and postsynaptic AMPA receptors within 20-30 nm and thereby forming a transsynaptic tripartite nanocomplex. Using functional EM, we show that single action potentials cause vesicles within the nanocomplex to fuse with a 50% probability. The loss of the presynaptic scaffold disrupts the formation of the tripartite transducers. Strikingly, the forced transsynaptic alignment of the Ca2+-channel subunit α2δ1 and AMPA receptors suffice to restore neurotransmission in a scaffold lacking synapse. Our results demonstrate a synaptic transducer nanocomplex that actively contributes to the organization of central synapses.

Every cell produces thousands of distinct lipid species, but insight into how lipid chemical diversity contributes to biological signaling is lacking, particularly because of a scarcity of methods for quantitatively studying lipid function in living cells. Using the example of diacylglycerols, prominent second messengers, we here investigate whether lipid chemical diversity can provide a basis for cellular signal specification. We generated photo-caged lipid probes, which allow acute manipulation of distinct diacylglycerol species in the plasma membrane. Combining uncaging experiments with mathematical modeling, we were able to determine binding constants for diacylglycerol-protein interactions, and kinetic parameters for diacylglycerol transbilayer movement and turnover in quantitative live-cell experiments. Strikingly, we find that affinities and kinetics vary by orders of magnitude due to diacylglycerol side-chain composition. These differences are sufficient to explain differential recruitment of diacylglycerol binding proteins and, thus, differing downstream phosphorylation patterns. Our approach represents a generally applicable method for elucidating the biological function of single lipid species on subcellular scales in quantitative live-cell experiments.

Neuronal communication across synapses relies on neurotransmitter release from presynaptic active zones (AZs) followed by postsynaptic transmitter detection. Synaptic plasticity homeostatically maintains functionality during perturbations and enables memory formation. Postsynaptic plasticity targets neurotransmitter receptors, but presynaptic mechanisms regulating the neurotransmitter release apparatus remain largely enigmatic. By studying Drosophila neuromuscular junctions (NMJs) we show that AZs consist of nano-modular release sites and identify a molecular sequence that adds modules within minutes of inducing homeostatic plasticity. This requires cognate transport machinery and specific AZ-scaffolding proteins. Structural remodeling is not required for immediate potentiation of neurotransmitter release, but necessary to sustain potentiation over longer timescales. Finally, mutations in Unc13 disrupting homeostatic plasticity at the NMJ also impair short-term memory when central neurons are targeted, suggesting that both plasticity mechanisms utilize Unc13. Together, while immediate synaptic potentiation capitalizes on available material, it triggers the coincident incorporation of modular release sites to consolidate synaptic potentiation.

Every cell produces thousands of distinct lipid species, but methodology for studying the biological roles of individual lipids is insufficient. Using the example of diacylglycerols, prominent second messengers, we here investigate whether lipid chemical diversity can provide a basis for cellular signal specification. We developed novel photo-caged lipid probes, which allow acute manipulation of distinct diacylglycerol species in the plasma membrane. Combining uncaging experiments with mathematical modelling enabled the determination of binding constants for diacylglycerol-protein interactions and kinetic parameters for diacylglycerol transbilayer movement and turnover in quantitative live-cell experiments. Strikingly, we find that affinities and kinetics vary by orders of magnitude due to diacylglycerol structural diversity. These differences are sufficient to explain differential recruitment of diacylglycerol binding proteins and thus differing downstream phosphorylation patterns. Our approach represents a generally applicable method for elucidating the biological function of single lipid species on subcellular scales.

Synaptic transmission relies on the rapid fusion of neurotransmitter-containing synaptic vesicles (SVs), which happens in response to action potential (AP)-induced Ca2+ influx at active zones (AZs). A highly conserved molecular machinery cooperates at SV-release sites to mediate SV plasma membrane attachment and maturation, Ca2+ sensing, and membrane fusion. Despite this high degree of conservation, synapses - even within the same organism, organ or neuron - are highly diverse regarding the probability of APs to trigger SV fusion. Additionally, repetitive activation can lead to either strengthening or weakening of transmission. In this review, we discuss mechanisms controlling release probability and this short-term plasticity. We argue that an important layer of control is exerted by evolutionarily conserved AZ scaffolding proteins, which determine the coupling distance between SV fusion sites and voltage-gated Ca2+ channels (VGCC) and, thereby, shape synapse-specific input/output behaviors. We propose that AZ-scaffold modifications may occur to adapt the coupling distance during synapse maturation and plastic regulation of synapse strength.