A multimodal Transformer Network for protein-small molecule interactions enhances predictions of kinase inhibition and enzyme-substrate relationships.

The activities of most enzymes and drugs depend on interactions between proteins and small molecules. Accurate prediction of these interactions could greatly accelerate pharmaceutical and biotechnological research. Current machine learning models designed for this task have a limited ability to generalize beyond the proteins used for training. This limitation is likely due to a lack of information exchange between the protein and the small molecule during the generation of the required numerical representations. Here, we introduce ProSmith, a machine learning framework that employs a multimodal Transformer Network to simultaneously process protein amino acid sequences and small molecule strings in the same input. This approach facilitates the exchange of all relevant information between the two molecule types during the computation of their numerical representations, allowing the model to account for their structural and functional interactions. Our final model combines gradient boosting predictions based on the resulting multimodal Transformer Network with independent predictions based on separate deep learning representations of the proteins and small molecules. The resulting predictions outperform recently published state-of-the-art models for predicting protein-small molecule interactions across three diverse tasks: predicting kinase inhibitions; inferring potential substrates for enzymes; and predicting Michaelis constants KM. The Python code provided can be used to easily implement and improve machine learning predictions involving arbitrary protein-small molecule interactions.

Turnover number predictions for kinetically uncharacterized enzymes using machine and deep learning.

The turnover number kcat, a measure of enzyme efficiency, is central to understanding cellular physiology and resource allocation. As experimental kcat estimates are unavailable for the vast majority of enzymatic reactions, the development of accurate computational prediction methods is highly desirable. However, existing machine learning models are limited to a single, well-studied organism, or they provide inaccurate predictions except for enzymes that are highly similar to proteins in the training set. Here, we present TurNuP, a general and organism-independent model that successfully predicts turnover numbers for natural reactions of wild-type enzymes. We constructed model inputs by representing complete chemical reactions through differential reaction fingerprints and by representing enzymes through a modified and re-trained Transformer Network model for protein sequences. TurNuP outperforms previous models and generalizes well even to enzymes that are not similar to proteins in the training set. Parameterizing metabolic models with TurNuP-predicted kcat values leads to improved proteome allocation predictions. To provide a powerful and convenient tool for the study of molecular biochemistry and physiology, we implemented a TurNuP web server.

A general model to predict small molecule substrates of enzymes based on machine and deep learning.

For most proteins annotated as enzymes, it is unknown which primary and/or secondary reactions they catalyze. Experimental characterizations of potential substrates are time-consuming and costly. Machine learning predictions could provide an efficient alternative, but are hampered by a lack of information regarding enzyme non-substrates, as available training data comprises mainly positive examples. Here, we present ESP, a general machine-learning model for the prediction of enzyme-substrate pairs with an accuracy of over 91% on independent and diverse test data. ESP can be applied successfully across widely different enzymes and a broad range of metabolites included in the training data, outperforming models designed for individual, well-studied enzyme families. ESP represents enzymes through a modified transformer model, and is trained on data augmented with randomly sampled small molecules assigned as non-substrates. By facilitating easy in silico testing of potential substrates, the ESP web server may support both basic and applied science.

Thiophosphinoyl-Tethered Ylide-Substituted Heavier Carbenes: Synthesis, Structures and Stabilities.

The synthesis and structure analysis of a series of mono and diylide-substituted tetrylenes of type YEX and Y2 E (E=Ge, Sn, Pb; X=Cl or Br) using a thiophosphinoyl-tethered metallated ylide (Y=Ph2 P(S)-C-P(pip)Ph2 with pip=piperidyl) is reported, amongst the first ylide-substituted plumbylenes. The tetrylenes feature distinct trends in the spectroscopic and structural properties of the ylide ligand with increasing atomic number of the tetrel element. For instance, an increasingly high-field shifted signal for the thiophosphinoyl group is observed in the 31 P{1 H} NMR spectrum as a consequence of the increasing polarity of the element-carbon bond, which likewise results in a shortening of the ylidic C-P bond in the solid-state structure. The diylidyltetrylenes are unstable towards transylidation forming the mono(ylide)tetrylenes when treated with the tetrel dihalides according to the stability trend: Y2 Pb

Synthesis, Crystal and Electronic Structures of a Thiophosphinoyl- and Amino-Substituted Metallated Ylide.

Invited for this month's cover is the group of Viktoria H. Gessner at the Ruhr-University in Bochum (Germany). The cover shows the structure of the newly reported, isolated metallated ylide. Due to the high negative charge at the ylidic carbon center this compound is "on fire", but can be stabilized by smart molecular design. Structure analyses of the different alkali metal complexes combined with computational studies provide insights into the electronic structure of the compounds Read the full text of their Communication at 10.1002/open.202100178.

Deep learning allows genome-scale prediction of Michaelis constants from structural features.

The Michaelis constant KM describes the affinity of an enzyme for a specific substrate and is a central parameter in studies of enzyme kinetics and cellular physiology. As measurements of KM are often difficult and time-consuming, experimental estimates exist for only a minority of enzyme-substrate combinations even in model organisms. Here, we build and train an organism-independent model that successfully predicts KM values for natural enzyme-substrate combinations using machine and deep learning methods. Predictions are based on a task-specific molecular fingerprint of the substrate, generated using a graph neural network, and on a deep numerical representation of the enzyme's amino acid sequence. We provide genome-scale KM predictions for 47 model organisms, which can be used to approximately relate metabolite concentrations to cellular physiology and to aid in the parameterization of kinetic models of cellular metabolism.

Synthesis, Crystal and Electronic Structures of a Thiophosphinoyl- and Amino-Substituted Metallated Ylide.

α-Metallated ylides have revealed themselves to be versatile reagents for the introduction of ylide groups. Herein, we report the synthesis of the thiophosphinoyl and piperidyl (Pip) substituted α-metallated ylide [Ph2 (Pip)P=C-P(S)Ph2 ]M (M=Li, Na, K) through a four-step synthetic procedure starting from diphenylmethylphosphine sulfide. Metallation of the ylide intermediate was successfully accomplished with different alkali metal bases delivering the lithium, sodium and potassium salts, the latter isolable in high yields. Structure analyses of the lithium and potassium compounds in the solid state with and without crown ether revealed different aggregates (monomer, dimer and hexamer) with the metals coordinated by the thiophosphoryl moiety and ylidic carbon atom. Although the piperidyl group does not coordinate to the metal, it significantly contributes to the stability of the yldiide by charge delocalization through negative hyperconjugation.

Towards the rational design of ylide-substituted phosphines for gold(i)-catalysis: from inactive to ppm-level catalysis.

The implementation of gold catalysis into large-scale processes suffers from the fact that most reactions still require high catalyst loadings to achieve efficient catalysis thus making upscaling impractical. Here, we report systematic studies on the impact of the substituent in the backbone of ylide-substituted phosphines (YPhos) on the catalytic activity in the hydroamination of alkynes, which allowed us to increase the catalyst performance by orders of magnitude. While electronic changes of the ligand properties by introduction of aryl groups with electron-withdrawing or electron-donating groups had surprisingly little impact on the activity of the gold complexes, the use of bulky aryl groups with ortho-substituents led to a remarkable boost in the catalyst activity. However, this catalyst improvement is not a result of an increased steric demand of the ligand towards the metal center, but due to steric protection of the reactive ylidic carbon centre in the ligand backbone. The gold complex of the thus designed mesityl-substituted YPhos ligand YMesPCy2, which is readily accessible in one step from a simple phosphonium salt, exhibited a high catalyst stability and allowed for turnover numbers up to 20 000 in the hydroamination of a series of different alkynes and amines. Furthermore, the catalyst was also active in more challenging reactions including enyne cyclisation and the formation of 1,2-dihydroquinolines.

Antineoplastic Agents Targeting Sphingolipid Pathways.

Emerging studies in the enigmatic area of bioactive lipids have made many exciting new discoveries in recent years. Once thought to play a strictly structural role in cellular function, it has since been determined that sphingolipids and their metabolites perform a vast variety of cellular functions beyond what was previously believed. Of utmost importance is their role in cellular signaling, for it is now well understood that select sphingolipids serve as bioactive molecules that play critical roles in both cancer cell death and survival, as well as other cellular responses such as chronic inflammation, protection from intestinal pathogens, and intrinsic protection from intestinal contents, each of which are associated with oncogenesis. Importantly, it has been demonstrated time and time again that many different tumors display dysregulation of sphingolipid metabolism, and the exact profile of said dysregulation has been proven to be useful in determining not only the presence of a tumor, but also the susceptibility to various chemotherapeutic drugs, as well as the metastasizing characteristics of the malignancies. Since these discoveries surfaced it has become apparent that the understanding of sphingolipid metabolism and profile will likely become of great importance in the clinic for both chemotherapy and diagnostics of cancer. The goal of this paper is to provide a comprehensive review of the current state of chemotherapeutic agents that target sphingolipid metabolism that are undergoing clinical trials. Additionally, we will formulate questions involving the use of sphingolipid metabolism as chemotherapeutic targets in need of further research.

A diamino-substituted carbodiphosphorane as strong C-donor and weak N-donor: isolation of monomeric trigonal-planar L·ZnCl2.

The isolation, structural characterization and coordination chemistry of a di(amino)-substituted carbodiphosphorane (CDP) are reported. Compared to the analogue, dianionic bis(iminophosphoryl)methandiides, the CDP is a stronger C-, but much weaker N-donor which led to the isolation of solely C-coordinated metal complexes amongst an unusual monomeric trigonal-planar L·ZnCl2 complex.

Towards the Preparation of Stable Cyclic Amino(ylide)Carbenes.

Cyclic amino(ylide)carbenes (CAYCs) are the ylide-substituted analogues of N-heterocyclic Carbenes (NHCs). Due to the stronger π donation of the ylide compared to an amino moiety they are stronger donors and thus are desirable ligands for catalysis. However, no stable CAYC has been reported until today. Here, we describe experimental and computational studies on the synthesis and stability of CAYCs based on pyrroles with trialkyl onium groups. Attempts to isolate two CAYCs with trialkyl phosphonium and sulfonium ylides resulted in the deprotonation of the alkyl groups instead of the formation of the desired CAYCs. In case of the PCy3-substituted system, the corresponding ylide was isolated, while deprotonation of the SMe2-functionalized compound led to the formation of ethene and the thioether. Detailed computational studies on various trialkyl onium groups showed that both the α- and ß-deprotonated compounds were energetically favored over the free carbene. The most stable candidates were revealed to be α-hydrogen-free adamantyl-substituted onium groups, for which ß-deprotonation is less favorable at the bridgehead position. Overall, the calculations showed that the isolation of CAYCs should be possible, but careful design is required to exclude decomposition pathways such as deprotonations at the onium group.

Cationic Phosphorus Compounds Based on a Bis(1-piperidinyl)-Substituted Carbodiphosphorane: Syntheses, Structures, and Csp3 -H Activation.

The use of the bis(1-piperidinyl)-substituted carbodiphosphorane (Ph2(Pip)P)2C (1) as an NCN ligand for the stabilization of phosphorus cations was studied. A simple ligand for halide exchange allowed the synthesis and isolation of a series of phosphorus monocations of the type [1-PR2]+ (with R = Cl, Br, I, CyCl, Ph). These cations exhibit characteristic NMR and structural properties which nicely correlate with the charge at the central phosphorus atom and the interaction between the ligand and the PR2 moiety. Halide abstraction from the monocations does not result in isolable dicationic compounds but in an unexpected intramolecular Csp3 -H activation in the piperidinyl group. DFT studies show that the selective activation of the CH2 group next to the nitrogen atom instead of a CH group at the phenyl substituents proceeds via an iminium intermediate formed by hydride transfer from the carbon atom to the cationic phosphorus center. This observation clearly demonstrates the pronounced π acidity of the dicationic phosphorus species in comparison to compounds with a further π-donor substituent.

Increased BOLD sensitivity in the orbitofrontal cortex using slice-dependent echo times at 3 T.

Functional magnetic resonance imaging (fMRI) exploits the blood oxygenation level dependent (BOLD) effect to detect neuronal activation related to various experimental paradigms. Some of these, such as reversal learning, involve the orbitofrontal cortex and its interaction with other brain regions like the amygdala, striatum or dorsolateral prefrontal cortex. These paradigms are commonly investigated with event-related methods and gradient echo-planar imaging (EPI) with short echo time of 27 ms. However, susceptibility-induced signal losses and image distortions in the orbitofrontal cortex are still a problem for this optimized sequence as this brain region consists of several slices with different optimal echo times. An EPI sequence with slice-dependent echo times is suitable to maximize BOLD sensitivity in all slices and might thus improve signal detection in the orbitofrontal cortex. To test this hypothesis, we first optimized echo times via BOLD sensitivity simulation. Second, we measured 12 healthy volunteers using a standard EPI sequence with an echo time of 27 ms and a modified EPI sequence with echo times ranging from 22 ms to 47 ms. In the orbitofrontal cortex, the number of activated voxels increased from 87 ± 44 to 549 ± 83 and the maximal t-value increased from 4.4 ± 0.3 to 5.4 ± 0.3 when the modified EPI was used. We conclude that an EPI with slice-dependent echo times may be a valuable tool to mitigate susceptibility artifacts in event-related whole-brain fMRI studies with a focus on the orbitofrontal cortex.

Biopsy targeting gliomas: do functional imaging techniques identify similar target areas?

OBJECTIVE: Because of the heterogeneous nature of glioma, biopsies performed should be targeted at the most anaplastic region. Several functional magnetic resonance imaging (MRI) or positron emission tomography (PET) techniques have been proposed for identifying the most anaplastic tumor area. However, it is unclear whether the recommended biopsy targets based on these various functional imaging modalities correspond with each other. Thus, the purpose was to evaluate whether they identify similar target areas. MATERIALS AND METHODS: A total of 61 patients with suspected glioma were assessed within 2.3 +/- 3.5 days by MRI, 18F-fluorothymidine-, and 18F-fluorodeoxyglucose-PET. Thirty-five patients underwent gross total resection and 26 were stereotactically biopsied. MRI was performed on a 1.5 Tesla broadband transmit/receive system, using a double-resonant birdcage coil. The MRI protocol comprised of sodium (23Na)-MRI (3D-radial projection imaging), proton spectroscopic imaging (1H-MRSI, point-resolved spectroscopy), arterial spin-labeling (ASL) perfusion MRI, dynamic contrast-enhanced (DCE) MRI, and dynamic-susceptibility-weighted (DSC) perfusion MRI after a single dose each of gadobenate dimeglumine. Also, apparent diffusion coefficient (ADC) maps were processed from diffusion tensor images. Image analysis comprised a detailed semiquantitative region of interest analysis of the different parameter values as well as visual identification of the most conspicuous tumor areas on parameter maps, for example, areas with maximum tumor perfusion, highest metabolite ratios of choline-containing compounds/N-acetyl-aspartate, or lowest ADC values within tumor tissue. Colocalization of these areas was then assessed. RESULTS: Regarding tumor vascularity-related parameters and tumor proliferation-related parameters, the higher the glioma grade the higher were the respective parameters in semiquantitative analysis. ADC values decreased with glioma grade. In the whole study population comprising low- (N = 15) and high-grade gliomas (N = 42), except for 23Na-MRI, there was good (>50%) or perfect (100%) agreement of the tumor areas with highest values on parameter images in the majority of cases (>80%), that is, tumor areas with increased thymidine-uptake and highest choline, both suggestive of increased tumor proliferation, and elevated microcirculation as demonstrated by DSC-, arterial spin-labeling-, and DCE-MRI. 23Na-MRI depicted the highest signal within necrotic tumor areas, but non-necrotic gliomas also showed a perfect agreement in more than 61%. 18F-fluorothymidine-PET, DSC-, and DCE-MRI, diffusion-weighted imaging as well as MR spectroscopic imaging correctly detected no glioma heterogeneity in all 15 histologically proven grade II gliomas but identified suspicious areas in all 3 nonenhancing grade III gliomas. CONCLUSION: Both imaging techniques that depict microcirculation and techniques that visualize proliferation identify similar target areas.

Cerebral processing of pain in school-aged children with neonatal nociceptive input: an exploratory fMRI study.

Due to maturation-related plasticity of the developing nociceptive system, neonatal nociceptive input, as induced by medical procedures in the neonatal intensive care unit (NICU), may cause long-term alterations in pain processing. Using functional magnetic resonance imaging, this study investigated the cerebral pain response in school-aged children and adolescents (11-16 yr) with experience in a NICU after preterm (or=37 weeks gestational age, N=9) as compared to fullterm control children without early hospitalization (N=9). NICU children had been recruited retrospectively among former patients of the Children's University Hospital Mannheim. All children had participated in our previous studies [46,49] entailing psychophysical measurements. In response to tonic (30s) heat stimuli of individually adjusted moderate pain intensity, which were of comparable temperature across groups, the preterm but not the fullterm NICU children exhibited significant activations in a number of brain regions (thalamus, anterior cingulate cortex, cerebellum, basal ganglia, and periaquaeductal gray) that were not significantly activated in controls. The preterms showed significantly higher activations than controls in primary somatosensory cortex, anterior cingulate cortex, and insula. This exaggerated brain response was pain-specific and was not observed during non-painful warmth stimulation. Preterms' continuous pain ratings revealed a tendency for increased sensitization within and a lack of habituation across trials. In highly vulnerable children such as preterms, neonatal nociceptive input may, aside from other neurodevelopmental consequences, persistently increase the gain within pain pathways.

Context conditioning and extinction in humans: differential contribution of the hippocampus, amygdala and prefrontal cortex.

Functional magnetic resonance imaging was used to investigate the role of the hippocampus, amygdala and medial prefrontal cortex (mPFC) in a contextual conditioning and extinction paradigm provoking anxiety. Twenty-one healthy persons participated in a differential context conditioning procedure with two different background colours as contexts. During acquisition increased activity to the conditioned stimulus (CS+) relative to the CS- was found in the left hippocampus and anterior cingulate cortex (ACC). The amygdala, insula and inferior frontal cortex were differentially active during late acquisition. Extinction was accompanied by enhanced activation to CS+ vs. CS- in the dorsal anterior cingulate cortex (dACC). The results are in accordance with animal studies and provide evidence for the important role of the hippocampus in contextual learning in humans. Connectivity analyses revealed correlated activity between the left posterior hippocampus and dACC (BA32) during early acquisition and the dACC, left posterior hippocampus and right amygdala during extinction. These data are consistent with theoretical models that propose an inhibitory effect of the mPFC on the amygdala. The interaction of the mPFC with the hippocampus may reflect the context-specificity of extinction learning.

Nanosizing of fluorescent objects by spatially modulated illumination microscopy.

A new approach to measuring the sizes of small fluorescent objects by use of spatially modulated illumination (SMI) far-field light microscopy is presented. This method is based on SME measurements combined with a new SMI virtual microscopy (VIM) data analysis calibration algorithm. Here, experimental SMI measurements of fluorescent objects with known diameter (size) were made. From the SMI data obtained, the size was determined in an independent way by use of the SMI VIM algorithm. The results showed that with SMI microscopy in combination with SMI VIM calibration, subwavelength object size measurements as small as 40 nm are experimentally feasible with high accuracy.