Optimized model architectures for deep learning on genomic data.

The success of deep learning in various applications depends on task-specific architecture design choices, including the types, hyperparameters, and number of layers. In computational biology, there is no consensus on the optimal architecture design, and decisions are often made using insights from more well-established fields such as computer vision. These may not consider the domain-specific characteristics of genome sequences, potentially limiting performance. Here, we present GenomeNet-Architect, a neural architecture design framework that automatically optimizes deep learning models for genome sequence data. It optimizes the overall layout of the architecture, with a search space specifically designed for genomics. Additionally, it optimizes hyperparameters of individual layers and the model training procedure. On a viral classification task, GenomeNet-Architect reduced the read-level misclassification rate by 19%, with 67% faster inference and 83% fewer parameters, and achieved similar contig-level accuracy with ~100 times fewer parameters compared to the best-performing deep learning baselines.

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning.

Late-stage functionalization is an economical approach to optimize the properties of drug candidates. However, the chemical complexity of drug molecules often makes late-stage diversification challenging. To address this problem, a late-stage functionalization platform based on geometric deep learning and high-throughput reaction screening was developed. Considering borylation as a critical step in late-stage functionalization, the computational model predicted reaction yields for diverse reaction conditions with a mean absolute error margin of 4-5%, while the reactivity of novel reactions with known and unknown substrates was classified with a balanced accuracy of 92% and 67%, respectively. The regioselectivity of the major products was accurately captured with a classifier F-score of 67%. When applied to 23 diverse commercial drug molecules, the platform successfully identified numerous opportunities for structural diversification. The influence of steric and electronic information on model performance was quantified, and a comprehensive simple user-friendly reaction format was introduced that proved to be a key enabler for seamlessly integrating deep learning and high-throughput experimentation for late-stage functionalization.

Identifying opportunities for late-stage C-H alkylation with high-throughput experimentation and in silico reaction screening.

Enhancing the properties of advanced drug candidates is aided by the direct incorporation of specific chemical groups, avoiding the need to construct the entire compound from the ground up. Nevertheless, their chemical intricacy often poses challenges in predicting reactivity for C-H activation reactions and planning their synthesis. We adopted a reaction screening approach that combines high-throughput experimentation (HTE) at a nanomolar scale with computational graph neural networks (GNNs). This approach aims to identify suitable substrates for late-stage C-H alkylation using Minisci-type chemistry. GNNs were trained using experimentally generated reactions derived from in-house HTE and literature data. These trained models were then used to predict, in a forward-looking manner, the coupling of 3180 advanced heterocyclic building blocks with a diverse set of sp3-rich carboxylic acids. This predictive approach aimed to explore the substrate landscape for Minisci-type alkylations. Promising candidates were chosen, their production was scaled up, and they were subsequently isolated and characterized. This process led to the creation of 30 novel, functionally modified molecules that hold potential for further refinement. These results positively advocate the application of HTE-based machine learning to virtual reaction screening.

A parsimonious approach to predict regions affected by sewer-borne contaminants in urban aquifers.

Leaky urban drainage networks (UDNs) exfiltrating wastewater can contaminate aquifers. Detailed knowledge on spatiotemporal distributions of water-dissolved, sewer-borne contaminants in groundwater is essential to protect urban aquifers and to optimize monitoring systems. We evaluated the effect of UDN layouts on the spreading of sewer-borne contaminants in groundwater using a parsimonious approach. Due to the UDN's long-term leakage behavior and the existence of non-degradable sewer-borne contaminants (equivalent to a conservative and constant contaminant source), we employed a concept of horizontal line sources to mimic the UDN layout. This does not require the consideration of bio-degradation processes or temporal delay and effectively bypasses the vadose zone, thus reducing computational requirements associated with a full simulation of leakages. We used a set of synthetic leakage scenarios which were generated using fractals and are based on a real-world UDN layout. We investigated the effects of typical leakage rates, varying groundwater flow directions, and UDN's layouts on the shape of the contaminant plume, disregarding the resulted concentration. Leakage rates showed minimal effects on the total covered plume area, whereas 89% of the variance of the plume's geometry is explained by both the UDN's layout (e.g., length and level of complexity) and groundwater flow direction. We demonstrated the potential of applying this approach to identify possible locations of groundwater observation wells using a real UDN layout. This straightforward and parsimonious method can serve as an initial step to strategically identify optimal monitoring systems locations within urban aquifers, and to improve sewer asset management at city scale.

A self-supervised deep learning method for data-efficient training in genomics.

Deep learning in bioinformatics is often limited to problems where extensive amounts of labeled data are available for supervised classification. By exploiting unlabeled data, self-supervised learning techniques can improve the performance of machine learning models in the presence of limited labeled data. Although many self-supervised learning methods have been suggested before, they have failed to exploit the unique characteristics of genomic data. Therefore, we introduce Self-GenomeNet, a self-supervised learning technique that is custom-tailored for genomic data. Self-GenomeNet leverages reverse-complement sequences and effectively learns short- and long-term dependencies by predicting targets of different lengths. Self-GenomeNet performs better than other self-supervised methods in data-scarce genomic tasks and outperforms standard supervised training with ~10 times fewer labeled training data. Furthermore, the learned representations generalize well to new datasets and tasks. These findings suggest that Self-GenomeNet is well suited for large-scale, unlabeled genomic datasets and could substantially improve the performance of genomic models.

Mass Balance of the Indoleamine 2,3-Dioxygenase Inhibitor Navoximod (GDC-0919) in Rats and Dogs: Unexpected Cyanide Release from Imidazo[5,1-a]isoindole and Species Differences in Glucuronidation.

Navoximod (GDC-0919) is a small molecule inhibitor of indoleamine 2,3-dioxygenase 1 (IDO1) developed to reduce T cell immunosuppression associated with cancer. This study describes the absorption, metabolism, and excretion (AME) of navoximod in rats and dogs after a single oral dose of [14C]-navoximod. An unexpected thiocyanate metabolite M1 and a chiral inversion metabolite M51 were captured as the major circulating metabolites in rats, accounting for 30% and 18% of 0-24 hours exposure, respectively. These two metabolites combined had much lower systemic exposure in dogs and humans (<6% and <1%). The novel cyanide release is proposed to occur via 4,5-epoxidation on the fused imidazole ring, leading to ring opening and rearrangement along with the release of cyanide. The decyanated metabolites were identified and confirmed by synthetic standards, which supported the proposed mechanism. In dogs, glucuronidation to M19 was the major clearance mechanism, representing 59% of the dose in the bile of bile duct-cannulated (BDC) dogs and 19% of the dose in the urine of intact dogs. Additionally, M19 also represented 52% of drug related exposure in circulation in dogs. In comparison, in humans, navoximod was mainly cleared through glucuronidation to M28 and excreted in urine (60% of the dose). The differences in the metabolism and elimination observed in vivo were qualitatively recapitulated in vitro with liver microsomes, suspended hepatocytes, and cocultured primary hepatocytes. The striking species differences in regioselective glucuronidation is likely explained by the species differences in UGT1A9, which was mainly responsible for M28 formation in humans. SIGNIFICANCE STATEMENT: The results from this study demonstrated significant species differences in metabolism (especially glucuronidation) and elimination of navoximod among rats, dogs, and humans. The study also illustrated the mechanism of a novel cyanide release metabolism from the fused imidazo[5,1-a]isoindole ring. Such biotransformation should be kept in mind when working with imidazole-containing new chemical entities in drug discovery and development.

Determining the impact of flow velocities on reactive processes associated with Enterococcus faecalis JH2-2.

This study focuses on the impact of infiltration rates on colloidal transport and reactive processes associated with Enterococcus faecalis JH2-2 using water-saturated sediment columns. The infiltration rates influence the physical transport of bacteria by controlling the mean flow velocity. This, in turn, impacts biological processes in pore water owing to the higher or lower residence time of the bacteria in the column. In the present study, continuous injection of E. faecalis (suspended in saline water with varying conditions of dissolved oxygen and nutrient concentrations) into a lab-scale sediment column was performed at flow velocities of 0.02 cm min-1 and 0.078 cm min-1, i.e., at residence times of 1-5 hours. The impact of residence times on reactive processes is significant for field scale setups. A process-based model with a first-order rate coefficient for each biological process was fitted for each obtained condition-specific dataset from the experimental observations (breakthrough curves). The coefficients were converted to a dimensionless form to facilitate the comparison of biological processes. These results indicate that the processes of attachment and growth were flow-dependent. The growth process in the absence of dissolved oxygen was the most dominant process, with a Damkoehler number of approximately 48.

Biological and Physical Clogging in Infiltration Wells: Effects of Well Diameter and Gravel Pack.

Gravity-driven infiltration into the shallow subsurface via small-diameter wells (SDWs), i.e., wells with an inner diameter smaller than 7.5 cm (3 inches) and no gravel pack) has proven to be a cost-efficient and flexible tool for managed aquifer recharge (MAR), as it provides relatively high recharge rates with minimal construction effort. SDWs have a significantly smaller open filter area than larger diameter wells with gravel pack, making the infiltration of low-quality waters through these wells more at risk clogging. To investigate their susceptibility for biological and physical clogging, 24 physical models with different well setups were evaluated by infiltrating either nutrient-poor but turbid water or nutrient-rich but clear water. The experiments showed that smaller diameters and the lack of a gravel pack increase the well's susceptibility to both kinds of clogging. However, this effect was observed to be much more pronounced for physical than for biological clogging. Our conclusion is that SDWs show severe disadvantages with respect to the infiltration of highly turbid waters in comparison to large diameter wells with a gravel pack. Nevertheless, this disadvantage is much less severe when it comes to the infiltration of clear but nutrient-rich waters (e.g., treated wastewater). Depending on the economic and geological circumstances of a MAR-project, this disadvantage could be outweighed by the significantly lower construction costs of SDWs.

Reactive-transport modelling of Enterococcus faecalis JH2-2 passage through water saturated sediment columns.

The reuse of treated wastewater (e.g. for irrigation) is a common practice to combat water scarcity problems world-wide. However, the potential spread of opportunistic pathogens and fecal contaminants like Enterococci within the subsoil could pose serious health hazards. Additional sources (e.g., leaky sewer systems, livestock farming) aggravate this situation. This study contributes to an understanding of pathogen spread in the environment, using a combined modelling and experimental approach. The impact of quartz sediment and certain wastewater characteristics on the dissemination of Enterococcus faecalis JH2-2 is investigated. The transport processes of advection-dispersion and straining were studied by injecting conservative saline tracer and fluorescent microspheres through sediment packed columns, and evaluating resulting breakthrough curves using models. Similarly, simultaneously occurring reactive processes of microbial attachment, decay, respiration and growth were studied by injecting Enterococcus faecalis JH2-2 suspended in water with or without dissolved oxygen (DO) and nutrients through sediment, and evaluating resulting inlet and outlet concentration curves. The processes of straining, microbial decay and growth, were important when DO was absent. Irreversible attachment was important when DO was present. Sensitivity analysis of each parameter was conducted, and field scale behavior of the processes was predicted, to facilitate future work.

Identification and Preclinical Development of a 2,5,6-Trisubstituted Fluorinated Pyridine Derivative as a Radioligand for the Positron Emission Tomography Imaging of Cannabinoid Type 2 Receptors.

Despite the broad implications of the cannabinoid type 2 receptor (CB2) in neuroinflammatory processes, a suitable CB2-targeted probe is currently lacking in clinical routine. In this work, we synthesized 15 fluorinated pyridine derivatives and tested their binding affinities toward CB2 and CB1. With a sub-nanomolar affinity (Ki for CB2) of 0.8 nM and a remarkable selectivity factor of >12,000 over CB1, RoSMA-18-d6 exhibited outstanding in vitro performance characteristics and was radiofluorinated with an average radiochemical yield of 10.6 ± 3.8% (n = 16) and molar activities ranging from 52 to 65 GBq/µmol (radiochemical purity > 99%). [18F]RoSMA-18-d6 showed exceptional CB2 attributes as demonstrated by in vitro autoradiography, ex vivo biodistribution, and positron emission tomography (PET). Further, [18F]RoSMA-18-d6 was used to detect CB2 upregulation on postmortem human ALS spinal cord tissues. Overall, these results suggest that [18F]RoSMA-18-d6 is a promising CB2 PET radioligand for clinical translation.

Novel strategies for expansion of tooth epithelial stem cells and ameloblast generation.

Enamel is secreted by ameloblasts derived from tooth epithelial stem cells (SCs). Humans cannot repair or regenerate enamel, due to early loss of tooth epithelial SCs. Contrarily in the mouse incisors, epithelial SCs are maintained throughout life and endlessly generate ameloblasts, and thus enamel. Here we isolated Sox2-GFP+ tooth epithelial SCs which generated highly cellular spheres following a novel in vitro strategy. This system enabled analysis of SC regulation by various signaling molecules, and supported the stimulatory and inhibitory roles of Shh and Bmp, respectively; providing better insight into the heterogeneity of the SCs. Further, we generated a novel mouse reporter, Enamelin-tdTomato for identification of ameloblasts in live tissues and cells, and used it to demonstrate presence of ameloblasts in the new 3D co-culture system of dental SCs. Collectively, our results provide means of generating 3D tooth epithelium from adult SCs which can be utilized toward future generation of enamel.

Collected Rain Water as Cost-Efficient Source for Aquifer Tracer Testing.

Locally collected precipitation water can be actively used as a groundwater tracer solution based on four inherent tracer signals: electrical conductivity, stable isotopic signatures of deuterium [δ2 H], oxygen-18 [δ18 O], and heat, which all may strongly differ from the corresponding background values in the tested groundwater. In hydrogeological practice, a tracer test is one of the most important methods for determining subsurface connections or field parameters, such as porosity, dispersivity, diffusion coefficient, groundwater flow velocity, or flow direction. A common problem is the choice of tracer and the corresponding permission by the appropriate authorities. This problem intensifies where tracer tests are conducted in vulnerable conservation or water protection areas (e.g., around drinking water wells). The use of (if required treated) precipitation as an elemental groundwater tracer is a practical solution for this problem, as it does not introduce foreign matters into the aquifer system, which may contribute positively to the permission delivery. Before tracer application, the natural variations of the participating end members' tracer signals have to be evaluated locally. To obtain a sufficient volume of tracer solution, precipitation can be collected as rain using a detached, large-scale rain collector, which will be independent from possibly existing surfaces like roofs or drained areas. The collected precipitation is then stored prior to a tracer experiment.

Functionally Distinctive Ptch Receptors Establish Multimodal Hedgehog Signaling in the Tooth Epithelial Stem Cell Niche.

Continuous growth of the mouse incisor teeth is due to the life-long maintenance of epithelial stem cells (SCs) in their niche called cervical loop (CL). Several signaling factors regulate SC maintenance and/or their differentiation to achieve organ homeostasis. Previous studies indicated that Hedgehog signaling is crucial for both the maintenance of the SCs in the niche, as well as for their differentiation. How Hedgehog signaling regulates these two opposing cellular behaviors within the confinement of the CL remains elusive. In this study, we used in vitro organ and cell cultures to pharmacologically attenuate Hedgehog signaling. We analyzed expression of various genes expressed in the SC niche to determine the effect of altered Hedgehog signaling on the cellular hierarchy within the niche. These genes include markers of SCs (Sox2 and Lgr5) and transit-amplifying cells (P-cadherin, Sonic Hedgehog, and Yap). Our results show that Hedgehog signaling is a critical survival factor for SCs in the niche, and that the architecture and the diversity of the SC niche are regulated by multiple Hedgehog ligands. We demonstrated the presence of an additional Hedgehog ligand, nerve-derived Desert Hedgehog, secreted in the proximity of the CL. In addition, we provide evidence that Hedgehog receptors Ptch1 and Ptch2 elicit independent responses, which enable multimodal Hedgehog signaling to simultaneously regulate SC maintenance and differentiation. Our study indicates that the cellular hierarchy in the continuously growing incisor is a result of complex interplay of two Hedgehog ligands with functionally distinct Ptch receptors. Stem Cells 2019;37:1238-1248.

The fate of DNAPL contaminants in non-consolidated subsurface systems - Discussion on the relevance of effective source zone geometries for plume propagation.

Dense non-aqueous phase liquids, i.e., DNAPLs and the evolving contaminant plumes in aquifers provide significant potential to pose hazards affecting both environment and human health. Therefore, a proper assessment of contaminant spreading within the subsurface is critical. This includes a sufficient characterization of governing parameters describing both the subsurface and the contaminant itself. Thereby, knowledge on the contaminant source zone and especially the source zone geometry, i.e., SZG is critically required, yet very uncertain. This study identifies current limitations and open research questions in the formation and shape determination of source zone geometry, as well as its relevance for contaminant plumes. Our literature review reveals that existing characterization methods are subject to data interpretation uncertainties, while the application of these methods on field scale is limited by technical demands and accompanied efforts. In a next step, methods to implement increased source zone information into calculation methods are discussed. By means of an exemplary application of selected assessment tools, i.e., plume response models, results clearly proof the relevance of SZGs for site assessment. However, existing plume response models consider over-simplified geometries that may compromise their suitability. Our findings identify the demand for improved characterization of complex SZGs and the need to better evaluate the dependency of DNAPL migration on system properties and external influences. With emphasized knowledge on the most relevant SZG features, the delineation of "effective" SZGs allowing for straightforward implementation into plume response models and an adaption of the latter to incorporate more information on SZGs should be possible.

Smile or die: Can subjective well-being increase survival in the face of substantive health impairments?

A robust relationship between subjective well-being and mortality has been established in the literature, but few studies address how subjective well-being interacts with the impact of concrete diseases on survival. In addition, issues of endogeneity between bad health and subjective well-being are ignored when it comes to survival. We assess both for the British Household Panel Survey (BHPS; 1991-2008) and specifically analyze whether subjective well-being predicts better chances of surviving diseases such as cancer or heart conditions. We find that several of the studied diseases consistently decrease survival chances in our sample (e.g. hazard ratio 3.47 for cancer), also when controlling for the severity of health problems. But our results do not suggest that well-being mitigates the effect these diseases have on mortality. Life satisfaction also does not predict longer survival in the data set if we control for the endogeneity of subjective well-being.

Metabolites of alectinib in human: their identification and pharmacological activity.

Two metabolites (M4 and M1b) in plasma and four metabolites (M4, M6, M1a and M1b) in faeces were detected through the human ADME study following a single oral administration of [14C]alectinib, a small-molecule anaplastic lymphoma kinase inhibitor, to healthy subjects. In the present study, M1a and M1b, which chemical structures had not been identified prior to the human ADME study, were identified as isomers of a carboxylate metabolite oxidatively cleaved at the morpholine ring. In faeces, M4 and M1b were the main metabolites, which shows that the biotransformation to M4 and M1b represents two main metabolic pathways for alectinib. In plasma, M4 was a major metabolite and M1b was a minor metabolite. The contribution to in vivo pharmacological activity of these circulating metabolites was assessed from their in vitro pharmacological activity and plasma protein binding. M4 had a similar cancer cell growth inhibitory activity and plasma protein binding to that of alectinib, suggesting its contribution to the antitumor activity of alectinib, whereas the pharmacological activity of M1b was insignificant.

A Straightforward Random Walk Model for Fast Push-Pull Tracer Test Evaluation.

In this article, we present a straightforward random walk model for fast evaluation of push-pull tracer tests. By developing an adaptive algorithm, we overcome the problem of manually defining how many particles have to be used to simulate the transport problem. Beside this, we validate the random walk model by evaluating a push-pull tracer test with drift phase and confirm the results with MT3DMS. The random walk model took less than 1% of computational time of MT3DMS, thus allowing a remarkable faster evaluation of push-pull tracer tests.

"I'm afraid I have bad news for you " Estimating the impact of different health impairments on subjective well-being.

Bad health decreases individuals' happiness, but few studies measure the impact of specific illnesses. We apply matching estimators to examine how changes in different (objective) conditions of bad health affect subjective well-being for a sample of 100,265 observations from the British Household Panel Survey (BHPS) database (1996-2006). The strongest effect is for alcohol and drug abuse, followed by anxiety, depression and other mental illnesses, stroke and cancer. Adaptation to health impairments varies across health impairments. There is also a puzzling asymmetry: strong adverse reactions to deteriorations in health appear alongside weak increases in well-being after health improvements. In conclusion, our analysis offers a more detailed account of how bad health influences happiness than accounts focusing on how bad self-assessed health affects individual well-being.

Ethene stabilization on Cu(111) by surface roughness.

The molecular vibrations of ethene adsorbed on roughened Cu(111) surfaces have been investigated with high resolution electron energy loss spectroscopy and density-functional-theory calculations. The roughness was introduced by sputtering or evaporation of copper, respectively, on the cooled surface. We found stabilization of the ethene layer compared to ethene adsorbed on pristine Cu(111). Furthermore, two new vibrational features observed on the rough surface can be assigned to frustrated translations and rotations of the ethene molecule on surface defects and are indicative of a different binding on the rough surface.