Will we ever be able to accurately predict solubility?

Accurate prediction of thermodynamic solubility by machine learning remains a challenge. Recent models often display good performances, but their reliability may be deceiving when used prospectively. This study investigates the origins of these discrepancies, following three directions: a historical perspective, an analysis of the aqueous solubility dataverse and data quality. We investigated over 20 years of published solubility datasets and models, highlighting overlooked datasets and the overlaps between popular sets. We benchmarked recently published models on a novel curated solubility dataset and report poor performances. We also propose a workflow to cure aqueous solubility data aiming at producing useful models for bench chemist. Our results demonstrate that some state-of-the-art models are not ready for public usage because they lack a well-defined applicability domain and overlook historical data sources. We report the impact of factors influencing the utility of the models: interlaboratory standard deviation, ionic state of the solute and data sources. The herein obtained models, and quality-assessed datasets are publicly available.

Time dynamics of the dose deposited by relativistic ultra-short electron beams.

Ultra-short electron beams are used as ultra-fast radiation source for radiobiology experiments aiming at very high energy electron beams (VHEE) radiotherapy with very high dose rates. Laser plasma accelerators are capable of producing electron beams as short as 1 fs and with tunable energy from few MeV up to multi-GeV with compact footprint. This makes them an attractive source for applications in different fields, where the ultra-short (fs) duration plays an important role. The time dynamics of the dose deposited by electron beams with energies in the range 50-250 MeV have been studied and the results are presented here. The results set a quantitative limit to the maximum dose rate at which the electron beams can impart dose.

Diffusion in liquid mixtures.

The understanding of transport and mixing in fluids in the presence and in the absence of external fields and reactions represents a challenging topic of strategic relevance for space exploration. Indeed, mixing and transport of components in a fluid are especially important during long-term space missions where fuels, food and other materials, needed for the sustainability of long space travels, must be processed under microgravity conditions. So far, the processes of transport and mixing have been investigated mainly at the macroscopic and microscopic scale. Their investigation at the mesoscopic scale is becoming increasingly important for the understanding of mass transfer in confined systems, such as porous media, biological systems and microfluidic systems. Microgravity conditions will provide the opportunity to analyze the effect of external fields and reactions on optimizing mixing and transport in the absence of the convective flows induced by buoyancy on Earth. This would be of great practical applicative relevance to handle complex fluids under microgravity conditions for the processing of materials in space.

Characterization of OSL dosimeters used at the ELI-beamlines laser-driven accelerator facility.

The Extreme Light Infrastructure (ELI) Beamlines laser-driven accelerator facility is set to operate the most intense non-military laser system in the world, with ultra-high power up to 10 PW, concentrated plasma intensities of up to 1024W cm-2, and ultra-short laser pulses of the order of few femtoseconds. A robust and redundant radiation monitoring system is in place to minimise risks to personnel and general public. Beryllium oxide optically stimulated luminescence (BeO-OSL) detectors are used to monitor radiation levels in the experimental building and surrounding grounds. In fact, in recent years, BeO-OSL have become an increasingly more popular choice for personal and environmental dosimetry. At ELI Beamlines, an exhaustive and thorough characterization process of the BeO-OSLs is in place. Dosimeter responses are studied as a function of delivered air kerma and photon energies. Calibration curves are calculated. Results from the latest calibration campaign are presented.

Teosinte (Zea mays ssp parviglumis) growth and transcriptomic response to weed stress identifies similarities and differences between varieties and with modern maize varieties.

Transcriptomic responses of plants to weed presence gives insight on the physiological and molecular mechanisms involved in the stress response. This study evaluated transcriptomic and morphological responses of two teosinte (Zea mays ssp parviglumis) (an ancestor of domesticated maize) lines (Ames 21812 and Ames 21789) to weed presence and absence during two growing seasons. Responses were compared after 6 weeks of growth in Aurora, South Dakota, USA. Plant heights between treatments were similar in Ames 21812, whereas branch number decreased when weeds were present. Ames 21789 was 45% shorter in weedy vs weed-free plots, but branch numbers were similar between treatments. Season-long biomass was reduced in response to weed stress in both lines. Common down-regulated subnetworks in weed-stressed plants were related to light, photosynthesis, and carbon cycles. Several unique response networks (e.g. aging, response to chitin) and gene sets were present in each line. Comparing transcriptomic responses of maize (determined in an adjacent study) and teosinte lines indicated three common gene ontologies up-regulated when weed-stressed: jasmonic acid response/signaling, UDP-glucosyl and glucuronyltransferases, and quercetin glucosyltransferase (3-O and 7-O). Overall, morphologic and transcriptomic differences suggest a greater varietal (rather than a conserved) response to weed stress, and implies multiple responses are possible. These findings offer insights into opportunities to define and manipulate gene expression of several different pathways of modern maize varieties to improve performance under weedy conditions.

Autoignition temperature: comprehensive data analysis and predictive models.

Here we report a new predictive model for autoignition temperature (AIT), an important physical parameter widely used to assess potential safety hazards of combustible materials. Available structure-AIT data extracted from different sources were critically analysed. Support vector regression (SVR) models on different data subsets were built in order to identify a reliable compound set on which a realistic model could be built. This led to a selection of the dataset containing 875 compounds annotated with AIT values. The thereupon-based SVR model performs reasonably well in cross-validation with the determination coefficient r 2 = 0.77 and mean absolute error MAE = 37.8°C. External validation on 20 industrial compounds missing in the training set confirmed its good predictive power (MAE = 28.7°C).

In Silico implementation of evolutionary paradigm in therapy design: Towards anti-cancer therapy as Darwinian process.

In here presented in silico study we suggest a way how to implement the evolutionary principles into anti-cancer therapy design. We hypothesize that instead of its ongoing supervised adaptation, the therapy may be constructed as a self-sustaining evolutionary process in a dynamic fitness landscape established implicitly by evolving cancer cells, microenvironment and the therapy itself. For these purposes, we replace a unified therapy with the 'therapy species', which is a population of heterogeneous elementary therapies, and propose a way how to turn the toxicity of the elementary therapy into its fitness in a way conforming to evolutionary causation. As a result, not only the therapies govern the evolution of different cell phenotypes, but the cells' resistances govern the evolution of the therapies as well. We illustrate the approach by the minimalistic ad hoc evolutionary model. Its results indicate that the resistant cells could bias the evolution towards more toxic elementary therapies by inhibiting the less toxic ones. As the evolutionary causation of cancer drug resistance has been intensively studied for a few decades, we refer to cancer as a special case to illustrate purely theoretical analysis.

Modelling of ready biodegradability based on combined public and industrial data sources.

The European Registration, Evaluation, Authorization and Restriction of Chemical Substances Regulation, requires marketed chemicals to be evaluated for Ready Biodegradability (RB), considering in silico prediction as valid alternative to experimental testing. However, currently available models may not be relevant to predict compounds of industrial interest, due to accuracy and applicability domain restriction issues. In this work, we present a new and extended RB dataset (2830 compounds), issued by the merging of several public data sources. It was used to train classification models, which were externally validated and benchmarked against already-existing tools on a set of 316 compounds coming from the industrial context. New models showed good performances in terms of predictive power (Balance Accuracy (BA) = 0.74-0.79) and data coverage (83-91%). The Generative Topographic Mapping approach identified several chemotypes and structural motifs unique to the industrial dataset, highlighting for which chemical classes currently available models may have less reliable predictions. Finally, public and industrial data were merged into global dataset containing 3146 compounds. This is the biggest dataset reported in the literature so far, covering some chemotypes absent in the public data. Thus, predictive model developed on the Global dataset has larger applicability domain than the existing ones.

Consensus models to predict oral rat acute toxicity and validation on a dataset coming from the industrial context.

We report predictive models of acute oral systemic toxicity representing a follow-up of our previous work in the framework of the NICEATM project. It includes the update of original models through the addition of new data and an external validation of the models using a dataset relevant for the chemical industry context. A regression model for LD50 and multi-class classification model for toxicity classes according to the Global Harmonized System categories were prepared. ISIDA descriptors were used to encode molecular structures. Machine learning algorithms included support vector machine (SVM), random forest (RF) and naïve Bayesian. Selected individual models were combined in consensus. The different datasets were compared using the generative topographic mapping approach. It appeared that the NICEATM datasets were lacking some relevant chemotypes for chemical industry. The new models trained on enlarged data sets have applicability domains (AD) sufficiently large to accommodate industrial compounds. The fraction of compounds inside the models' AD increased from 58% (NICEATM model) to 94% (new model). The increase of training sets improved models' prediction performance: RMSE values decreased from 0.56 to 0.47 and balanced accuracies increased from 0.69 to 0.71 for NICEATM and new models, respectively.

QSPR models for bioconcentration factor (BCF): are they able to predict data of industrial interest?

The bioconcentration factor (BCF), a key parameter required by the REACH regulation, estimates the tendency for a xenobiotic to concentrate inside living organisms. In silico methods can be valid alternatives to costly data measurements. However, in the industrial context, these theoretical approaches may fail to predict BCF with reasonable accuracy. We analyzed whether models built on public data only have adequate performances when challenged to predict industrial compounds. A new set of 1129 compounds has been collected by merging publicly available datasets. Generative Topographic Mapping was employed to compare this chemical space with a set of new compounds issued from the industry. Some new chemotypes absent in the training set (such as siloxanes) have been detected. A new BCF model has been built using ISIDA (In SIlico design and Data Analysis) fragment descriptors, support vector regression and random forest machine-learning methods. It has been externally validated on: (i) collected data from the literature and (ii) industrial data. The latter also served as benchmark for the freely available tools VEGA, EPISuite, TEST, OPERA. New model performs (RMSE of 0.58 log BCF units) comparably to existing ones but benefits of an extended applicability, covering the industrial set chemical space (78% data coverage).

In silico Design, Virtual Screening and Synthesis of Novel Electrolytic Solvents.

We report the building, validation and release of QSPR (Quantitative Structure Property Relationship) models aiming to guide the design of new solvents for the next generation of Li-ion batteries. The dataset compiled from the literature included oxidation potentials (Eox ), specific ionic conductivities (κ), melting points (Tm ) and boiling points (Tb ) for 103 electrolytes. Each of the resulting consensus models assembled 9-19 individual Support Vector Machine models built on different sets of ISIDA fragment descriptors.(1) They were implemented in the ISIDA/Predictor software. Developed models were used to screen a virtual library of 9965 esters and sulfones. The most promising compounds prioritized according to theoretically estimated properties were synthesized and experimentally tested.

Crystal structure of a xylulose 5-phosphate phosphoketolase. Insights into the substrate specificity for xylulose 5-phosphate.

Phosphoketolases (PK) are TPP-dependent enzymes which play essential roles in carbohydrate metabolism of numerous bacteria. Depending on the substrate specificity PKs can be subdivided into xylulose 5-phosphate (X5P) specific PKs (XPKs) and PKs which accept both X5P and fructose 6-phosphate (F6P) (XFPKs). Despite their key metabolic importance, so far only the crystal structures of two XFPKs have been reported. There are no reported structures for any XPKs and for any complexes between PK and substrate. One of the major unknowns concerning PKs mechanism of action is related to the structural determinants of PKs substrate specificity for X5P or F6P. We report here the crystal structure of XPK from Lactococcus lactis (XPK-Ll) at 2.1 Šresolution. Using small angle X-ray scattering (SAXS) we proved that XPK-Ll is a dimer in solution. Towards better understanding of PKs substrate specificity, we performed flexible docking of TPP-X5P and TPP-F6P on crystal structures of XPK-Ll, two XFPKs and transketolase (TK). Calculated structure-based binding energies consistently support XPK-Ll preference for X5P. Analysis of structural models thus obtained show that substrates adopt moderately different conformation in PKs active sites following distinct networks of polar interactions. Based on the here reported structure of XPK-Ll we propose the most probable amino acid residues involved in the catalytic steps of reaction mechanism. Altogether our results suggest that PKs substrate preference for X5P or F6P is the outcome of a fine balance between specific binding network and dissimilar catalytic residues depending on the enzyme (XPK or XFPK) - substrate (X5P or F6P) couples.

Interaction of Pure Marangoni Convection with a Propagating Reactive Interface under Microgravity.

A reactive interface in the form of an autocatalytic reaction front propagating in a bulk phase can generate a dynamic contact line upon reaching the free surface when a surface tension gradient builds up due to the change in chemical composition. Experiments in microgravity evidence the existence of a self-organized autonomous and localized coupling of a pure Marangoni flow along the surface with the reaction in the bulk. This dynamics results from the advancement of the contact line at the surface that acts as a moving source of the reaction, leading to the reorientation of the front propagation. Microgravity conditions allow one to isolate the transition regime during which the surface propagation is enhanced, whereas diffusion remains the main mode of transport in the bulk with negligible convective mixing, a regime typically concealed on Earth because of buoyancy-driven convection.

A new conceptual framework for the therapy by optimized multidimensional pulses of therapeutic activity. The case of multiple myeloma model.

We developed simulation methodology to assess eventual therapeutic efficiency of exogenous multiparametric changes in a four-component cellular system described by the system of ordinary differential equations. The method is numerically implemented to simulate the temporal behavior of a cellular system of multiple myeloma cells. The problem is conceived as an inverse optimization task where the alternative temporal changes of selected parameters of the ordinary differential equations represent candidate solutions and the objective function quantifies the goals of the therapy. The system under study consists of two main cellular components, tumor cells and their cellular environment, respectively. The subset of model parameters closely related to the environment is substituted by exogenous time dependencies - therapeutic pulses combining continuous functions and discrete parameters subordinated thereafter to the optimization. Synergistic interaction of temporal parametric changes has been observed and quantified whereby two or more dynamic parameters show effects that absent if either parameter is stimulated alone. We expect that the theoretical insight into unstable tumor growth provided by the sensitivity and optimization studies could, eventually, help in designing combination therapies.

Peristalticity-driven banded chemical garden.

Complex structures in nature are often formed by self-assembly. In order to mimic the formation, to enhance the production, or to modify the structures, easy-to-use methods are sought to couple engineering and self-assembly. Chemical-garden-like precipitation reactions are frequently used to study such couplings because of the intrinsic chemical and hydrodynamic interplays. In this work, we present a simple method of applying periodic pressure fluctuations given by a peristaltic pump which can be used to achieve regularly banded precipitate membranes in the copper-phosphate system.

Neighboring Structure Visualization on a Grid-based Layout.

Here, we describe an algorithm to visualize chemical structures on a grid-based layout in such a way that similar structures are neighboring. It is based on structure reordering with the help of the Hilbert Schmidt Independence Criterion, representing an empirical estimate of the Hilbert-Schmidt norm of the cross-covariance operator. The method can be applied to any layout of bi- or three-dimensional shape. The approach is demonstrated on a set of dopamine D5 ligands visualized on squared, disk and spherical layouts.

Vitamin D deficiency and seasonal variation over the years in São Paulo, Brazil.

Brazil is a tropical/subtropical geographic area with elevated ultraviolet (UV) radiation. We report very high prevalence of vitamin D deficiency in a large database of Brazilian subjects and show seasonal and reciprocal relationship between vitamin D and parathyroid hormone (PTH) over the years in this tropical area. INTRODUCTION: We aim to examine the prevalence of vitamin D deficiency, characterize the temporal relationship between 25-hydroxyvitamin D levels (25(OH)D) and intact PTH (iPTH) according to seasons, and investigate potential associations between 25(OH)D levels and extra-skeletal outcomes in a Brazilian population. METHODS: We retrospectively determined population weekly mean concentrations of unpaired 25(OH)D and iPTH using 39,004 laboratory results of Brazilian individuals of both genders aged 2 to 95 years. The 25(OH)D and iPTH distributions were normalized, and the means fit with a sinusoidal function. Potential associations between 25(OH)D serum levels and inflammatory markers, fasting glucose, HbA1c and Homeostasis Model Assessment index (HOMA) were examined. RESULTS: Of the samples, 33.9 % had 25(OH)D serum concentrations lower than 20 ng/mL, while the vast majority (70.7 %) were found to be vitamin D deficient or insufficient (<30 ng/mL). Vitamin D deficiency was significantly higher during the winter as compared to the summer (38.4 % <20 ng/mL and 75.5 % <30 ng/mL versus 23.3 % <20 ng/mL and 62.5 % <30 ng/mL, respectively; p < 0.001). Seasonal variation was observed for both 25(OH)D and iPTH. 25(OH)D peaks occurred in March and troughs in September. iPTH levels showed an inverted pattern of peaks and troughs with a delay of 1 ± 5 week. 25(OH)D was significantly associated with inflammatory markers but not with glucose homeostasis. CONCLUSIONS: A sinusoidal interrelationship has been detected between vitamin D and PTH in this tropical population. A large percentage of the individuals showed vitamin D deficiency. Public health strategies are needed to better understand and manage this very high and apparently contradictory prevalence of vitamin D deficiency.

Preliminary study of the applicability of the thin gap method on alpha emitters.

The thin gap method as an in-situ radiotracer technique is widely used. This study investigated the applicability of alpha emitters. PIPS and CsI alpha spectrometers were applied in a thin gap cell. A suitable (210)Po source was prepared by spontaneous deposition, Mylar foil was used to simulate water. A maximum intensity decrement of 7% within 25 µm was observed. Even though this method is suitable for the study of surface phenomena, further investigation is necessary e.g. into water and heat sensitivity.

Expert system for predicting reaction conditions: the Michael reaction case.

A generic chemical transformation may often be achieved under various synthetic conditions. However, for any specific reagents, only one or a few among the reported synthetic protocols may be successful. For example, Michael ß-addition reactions may proceed under different choices of solvent (e.g., hydrophobic, aprotic polar, protic) and catalyst (e.g., Brønsted acid, Lewis acid, Lewis base, etc.). Chemoinformatics methods could be efficiently used to establish a relationship between the reagent structures and the required reaction conditions, which would allow synthetic chemists to waste less time and resources in trying out various protocols in search for the appropriate one. In order to address this problem, a number of 2-classes classification models have been built on a set of 198 Michael reactions retrieved from literature. Trained models discriminate between processes that are compatible and respectively processes not feasible under a specific reaction condition option (feasible or not with a Lewis acid catalyst, feasible or not in hydrophobic solvent, etc.). Eight distinct models were built to decide the compatibility of a Michael addition process with each considered reaction condition option, while a ninth model was aimed to predict whether the assumed Michael addition is feasible at all. Different machine-learning methods (Support Vector Machine, Naive Bayes, and Random Forest) in combination with different types of descriptors (ISIDA fragments issued from Condensed Graphs of Reactions, MOLMAP, Electronic Effect Descriptors, and Chemistry Development Kit computed descriptors) have been used. Models have good predictive performance in 3-fold cross-validation done three times: balanced accuracy varies from 0.7 to 1. Developed models are available for the users at http://infochim.u-strasbg.fr/webserv/VSEngine.html . Eventually, these were challenged to predict feasibility conditions for ∼50 novel Michael reactions from the eNovalys database (originally from patent literature).

GTM-Based QSAR Models and Their Applicability Domains.

In this paper we demonstrate that Generative Topographic Mapping (GTM), a machine learning method traditionally used for data visualisation, can be efficiently applied to QSAR modelling using probability distribution functions (PDF) computed in the latent 2-dimensional space. Several different scenarios of the activity assessment were considered: (i) the "activity landscape" approach based on direct use of PDF, (ii) QSAR models involving GTM-generated on descriptors derived from PDF, and, (iii) the k-Nearest Neighbours approach in 2D latent space. Benchmarking calculations were performed on five different datasets: stability constants of metal cations Ca(2+) , Gd(3+) and Lu(3+) complexes with organic ligands in water, aqueous solubility and activity of thrombin inhibitors. It has been shown that the performance of GTM-based regression models is similar to that obtained with some popular machine-learning methods (random forest, k-NN, M5P regression tree and PLS) and ISIDA fragment descriptors. By comparing GTM activity landscapes built both on predicted and experimental activities, we may visually assess the model's performance and identify the areas in the chemical space corresponding to reliable predictions. The applicability domain used in this work is based on data likelihood. Its application has significantly improved the model performances for 4 out of 5 datasets.