AI is a viable alternative to high throughput screening: a 318-target study

High throughput screening (HTS) is routinely used to identify bioactive small molecules. This requires physical compounds, which limits coverage of accessible chemical space. Computational approaches combined with vast on-demand chemical libraries can access far greater chemical space, provided that the predictive accuracy is sufficient to identify useful molecules. Through the largest and most diverse virtual HTS campaign reported to date, comprising 318 individual projects, we demonstrate that our AtomNet® convolutional neural network successfully finds novel hits across every major therapeutic area and protein class. We address historical limitations of computational screening by demonstrating success for target proteins without known binders, high-quality X-ray crystal structures, or manual cherry-picking of compounds. We show that the molecules selected by the AtomNet® model are novel drug-like scaffolds rather than minor modifications to known bioactive compounds. Our empirical results suggest that computational methods can substantially replace HTS as the first step of small-molecule drug discovery.

Psychometric evaluation of the Decision Support System (DSS) for municipal nurses encountering health deterioration among older adults

BACKGROUND

A valid and reliable tool is crucial for municipal registered nurses (RNs) to make quick decisions in older adults who show rapid signs of health deterioration. The aim of this study was to investigate the psychometric properties of the Decision Support System (DSS) among older adults in the municipal healthcare system.

METHODS

Firstly, we utilized the Rasch dichotomous model to analyze the DSS assessments (n=281) that were collected from municipal RNs working with older adults in the municipal healthcare system. We examined the properties of the DSS in terms of its unidimensionality, item fit, and separation indices. Secondly, to investigate inter-rater agreement in using the DSS, four experienced municipal RNs used the DSS to assess 60 health deterioration scenarios presented by one human patient simulators. The 60 DSS assessments were then analyzed using the ICC (2,1), percentage agreement, and Cohen κ statistics.

RESULTS

The sample of older adults had a mean age of 82.8 (SD 11.7). The DSS met the criteria for unidimensionality, although two items did not meet the item fit statistics when all the DSS items were analyzed together. The person separation index was 0.47, indicating a limited level of separation among the sample. The item separation index was 11.43, suggesting that the DSS has good ability to discriminate between and separate the items. At the overall DSS level, inter-rater agreements were good according to the ICC. At the individual DSS item level, the percentage agreements were 75% or above, while the Cohen κ statistics ranged from 0.46 to 1.00.

CONCLUSIONS

The Rasch analysis revealed that the psychometric properties of the instrument were acceptable, although further research with a larger sample size and more items is needed. The DSS has the potential to assist municipal RNs in making clinical decisions regarding health deterioration in older adults, thereby avoiding unnecessary emergency admistion and helping.

Designed switch from covalent to non-covalent inhibitors of carboxylesterase Notum activity

N-Acyl indolines 4 are potent, non-covalent Notum inhibitors developed from a covalent virtual screening hit 2a. The lead compounds were simple to synthesise, achieved excellent potency in a biochemical Notum-OPTS assay and restored Wnt signalling in a cell-based TCF/LEF reporter assay. Multiple high resolution X-ray structures established a common binding mode of these inhibitors with the indoline bound centred in the palmiteolate pocket with key interactions being aromatic stacking and a water mediated hydrogen bond to the oxyanion hole. These N-acyl indolines 4 will be useful tools for use in vitro studies to investigate the role of Notum in disease models, especially when paired with a structurally related covalent inhibitor (e.g. 4w and 2a). Overall, this study highlights the designed switch from covalent to non-covalent Notum inhibitors and so illustrates a complementary approach for hit generation and target inhibition.

Swefoot - The Swedish national quality register for foot and ankle surgery

BACKGROUND

Population-based register data could be used to improve our knowledge of patients surgically treated for foot and ankle disorders. The quality register Swefoot was recently created to collect surgical and patient-reported data of foot and ankle surgery. This manuscript aims to describe the development and current use of the register.

METHODS

The development of Swefoot started in 2014 and currently, data on 16 different diagnoses are collected in 49 units performing foot and ankle surgery. Registrations are performed by the surgeon and the patient.

RESULTS

Between 2014 and 2020 approximately 20,000 surgical procedures have been registered. 75.1% of the registered patients were women, 9.3% were smokers, 9.3% had a concomitant rheumatoid disease, and 18.4% a BMI larger than 30 kg/m²_. CONCLUSIONS: The Swefoot is a unique national register for foot and ankle surgery. It is by now possible to present demographic, surgical, and outcome parameters based on Swefoot.

Design of a Potent, Selective, and Brain-Penetrant Inhibitor of Wnt-Deactivating Enzyme Notum by Optimization of a Crystallographic Fragment Hit

Notum is a carboxylesterase that suppresses Wnt signaling through deacylation of an essential palmitoleate group on Wnt proteins. There is a growing understanding of the role Notum plays in human diseases such as colorectal cancer and Alzheimer's disease, supporting the need to discover improved inhibitors, especially for use in models of neurodegeneration. Here, we have described the discovery and profile of 8l (ARUK3001185) as a potent, selective, and brain-penetrant inhibitor of Notum activity suitable for oral dosing in rodent models of disease. Crystallographic fragment screening of the Diamond-SGC Poised Library for binding to Notum, supported by a biochemical enzyme assay to rank inhibition activity, identified 6a and 6b as a pair of outstanding hits. Fragment development of 6 delivered 8l that restored Wnt signaling in the presence of Notum in a cell-based reporter assay. Assessment in pharmacology screens showed 8l to be selective against serine hydrolases, kinases, and drug targets.

Studying and mitigating the effects of data drifts on ML model performance at the example of chemical toxicity data

Machine learning models are widely applied to predict molecular properties or the biological activity of small molecules on a specific protein. Models can be integrated in a conformal prediction (CP) framework which adds a calibration step to estimate the confidence of the predictions. CP models present the advantage of ensuring a predefined error rate under the assumption that test and calibration set are exchangeable. In cases where the test data have drifted away from the descriptor space of the training data, or where assay setups have changed, this assumption might not be fulfilled and the models are not guaranteed to be valid. In this study, the performance of internally valid CP models when applied to either newer time-split data or to external data was evaluated. In detail, temporal data drifts were analysed based on twelve datasets from the ChEMBL database. In addition, discrepancies between models trained on publicly-available data and applied to proprietary data for the liver toxicity and MNT in vivo endpoints were investigated. In most cases, a drastic decrease in the validity of the models was observed when applied to the time-split or external (holdout) test sets. To overcome the decrease in model validity, a strategy for updating the calibration set with data more similar to the holdout set was investigated. Updating the calibration set generally improved the validity, restoring it completely to its expected value in many cases. The restored validity is the first requisite for applying the CP models with confidence. However, the increased validity comes at the cost of a decrease in model efficiency, as more predictions are identified as inconclusive. This study presents a strategy to recalibrate CP models to mitigate the effects of data drifts. Updating the calibration sets without having to retrain the model has proven to be a useful approach to restore the validity of most models.

Virtual Screening Directly Identifies New Fragment-Sized Inhibitors of Carboxylesterase Notum with Nanomolar Activity

Notum is a negative regulator of Wnt signaling acting through the hydrolysis of a palmitoleoylate ester, which is required for Wnt activity. Inhibitors of Notum could be of use in diseases where dysfunctional Notum activity is an underlying cause. A docking-based virtual screen (VS) of a large commercial library was used to shortlist 952 compounds for experimental validation as inhibitors of Notum. The VS was successful with 31 compounds having an IC50 < 500 nM. A critical selection process was then applied with two clusters and two singletons (1-4d) selected for hit validation. Optimization of 4d guided by structural biology identified potent inhibitors of Notum activity that restored Wnt/β-catenin signaling in cell-based models. The [1,2,4]triazolo[4,3-b]pyradizin-3(2H)-one series 4 represent a new chemical class of Notum inhibitors and the first to be discovered by a VS campaign. These results demonstrate the value of VS with well-designed docking models based on X-ray structures.

Synergy conformal prediction applied to large-scale bioactivity datasets and in federated learning

Confidence predictors can deliver predictions with the associated confidence required for decision making and can play an important role in drug discovery and toxicity predictions. In this work we investigate a recently introduced version of conformal prediction, synergy conformal prediction, focusing on the predictive performance when applied to bioactivity data. We compare the performance to other variants of conformal predictors for multiple partitioned datasets and demonstrate the utility of synergy conformal predictors for federated learning where data cannot be pooled in one location. Our results show that synergy conformal predictors based on training data randomly sampled with replacement can compete with other conformal setups, while using completely separate training sets often results in worse performance. However, in a federated setup where no method has access to all the data, synergy conformal prediction is shown to give promising results. Based on our study, we conclude that synergy conformal predictors are a valuable addition to the conformal prediction toolbox.

Structural Insights into Notum Covalent Inhibition

The carboxylesterase Notum hydrolyzes a palmitoleate moiety from Wingless/Integrated(Wnt) ligands and deactivates Wnt signaling. Notum inhibitors can restore Wnt signaling which may be of therapeutic benefit for pathologies such as osteoporosis and Alzheimer's disease. We report the identification of a novel class of covalent Notum inhibitors, 4-(indolin-1-yl)-4-oxobutanoate esters. High-resolution crystal structures of the Notum inhibitor complexes reveal a common covalent adduct formed between the nucleophile serine-232 and hydrolyzed butyric esters. The covalent interaction in solution was confirmed by mass spectrometry analysis. Inhibitory potencies vary depending on the warheads used. Mechanistically, the resulting acyl-enzyme intermediate carbonyl atom is positioned at an unfavorable angle for the approach of the active site water, which, combined with strong hydrophobic interactions with the enzyme pocket residues, hinders the intermediate from being further processed and results in covalent inhibition. These insights into Notum catalytic inhibition may guide development of more potent Notum inhibitors.

Deep Learning-Based Conformal Prediction of Toxicity

Predictive modeling for toxicity can help reduce risks in a range of applications and potentially serve as the basis for regulatory decisions. However, the utility of these predictions can be limited if the associated uncertainty is not adequately quantified. With recent studies showing great promise for deep learning-based models also for toxicity predictions, we investigate the combination of deep learning-based predictors with the conformal prediction framework to generate highly predictive models with well-defined uncertainties. We use a range of deep feedforward neural networks and graph neural networks in a conformal prediction setting and evaluate their performance on data from the Tox21 challenge. We also compare the results from the conformal predictors to those of the underlying machine learning models. The results indicate that highly predictive models can be obtained that result in very efficient conformal predictors even at high confidence levels. Taken together, our results highlight the utility of conformal predictors as a convenient way to deliver toxicity predictions with confidence, adding both statistical guarantees on the model performance as well as better predictions of the minority class compared to the underlying models.

Assessing the calibration in toxicological in vitro models with conformal prediction

Machine learning methods are widely used in drug discovery and toxicity prediction. While showing overall good performance in cross-validation studies, their predictive power (often) drops in cases where the query samples have drifted from the training data’s descriptor space. Thus, the assumption for applying machine learning algorithms, that training and test data stem from the same distribution, might not always be fulfilled. In this work, conformal prediction is used to assess the calibration of the models. Deviations from the expected error may indicate that training and test data originate from different distributions. Exemplified on the Tox21 datasets, composed of chronologically released Tox21Train, Tox21Test and Tox21Score subsets, we observed that while internally valid models could be trained using cross-validation on Tox21Train, predictions on the external Tox21Score data resulted in higher error rates than expected. To improve the prediction on the external sets, a strategy exchanging the calibration set with more recent data, such as Tox21Test, has successfully been introduced. We conclude that conformal prediction can be used to diagnose data drifts and other issues related to model calibration. The proposed improvement strategy—exchanging the calibration data only—is convenient as it does not require retraining of the underlying model.

Carboxylesterase Notum Is a Druggable Target to Modulate Wnt Signaling

Regulation of the Wnt signaling pathway is critically important for a number of cellular processes in both development and adult mammalian biology. This Perspective will provide a summary of current and emerging therapeutic opportunities in modulating Wnt signaling, especially through inhibition of Notum carboxylesterase activity. Notum was recently shown to act as a negative regulator of Wnt signaling through the removal of an essential palmitoleate group. Inhibition of Notum activity may represent a new approach to treat disease where aberrant Notum activity has been identified as the underlying cause. Reliable screening technologies are available to identify inhibitors of Notum, and structural studies are accelerating the discovery of new inhibitors. A selection of these hits have been optimized to give fit-for-purpose small molecule inhibitors of Notum. Three noteworthy examples are LP-922056 (26), ABC99 (27), and ARUK3001185 (28), which are complementary chemical tools for exploring the role of Notum in Wnt signaling.

Functional Diversity Facilitates Stability Under Environmental Changes in an Outdoor Microalgal Cultivation System

Functionally uniform monocultures have remained the paradigm in microalgal cultivation despite the apparent challenges to avoid invasions by other microorganisms. A mixed microbial consortium approach has the potential to optimize and maintain biomass production despite of seasonal changes and to be more resilient toward contaminations. Here we present a 3-year outdoor production of mixed consortia of locally adapted microalgae and bacteria in cold temperate latitude. Microalgal consortia were cultivated in flat panel photobioreactors using brackish Baltic Sea water and CO₂ from a cement factory (Degerhamn, Cementa AB, Heidelberg Cement Group) as a sustainable CO₂ source. To evaluate the ability of the microbial consortia to maintain stable biomass production while exposed to seasonal changes in both light and temperature, we tracked changes in the microbial community using molecular methods (16S and 18S rDNA amplicon sequencing) and monitored the biomass production and quality (lipid, protein, and carbohydrate content) over 3 years. Despite changes in environmental conditions, the mixed consortia maintained stable biomass production by alternating between two different predominant green microalgae (Monoraphidium and Mychonastes) with complementary tolerance to temperature. The bacterial population was few taxa co-occured over time and the composition did not have any connection to the shifts in microalgal taxa. We propose that a locally adapted and mixed microalgal consortia, with complementary traits, can be useful for optimizing yield of commercial scale microalgal cultivation.

TREM2/PLCγ2 signalling in immune cells: function, structural insight, and potential therapeutic modulation

The central role of the resident innate immune cells of the brain (microglia) in neurodegeneration has become clear over the past few years largely through genome-wide association studies (GWAS), and has rapidly become an active area of research. However, a mechanistic understanding (gene to function) has lagged behind. That is now beginning to change, as exemplified by a number of recent exciting and important reports that provide insight into the function of two key gene products – TREM2 (Triggering Receptor Expressed On Myeloid Cells 2) and PLCγ2 (Phospholipase C gamma2) – in microglia, and their role in neurodegenerative disorders. In this review we explore and discuss these recent advances and the opportunities that they may provide for the development of new therapies.

Systematic Analysis of Protein Targets Associated with Adverse Events of Drugs from Clinical Trials and Postmarketing Reports

Adverse drug reactions (ADRs) are undesired effects of medicines that can harm patients and are a significant source of attrition in drug development. ADRs are anticipated by routinely screening drugs against secondary pharmacology protein panels. However, there is still a lack of quantitative information on the links between these off-target proteins and the reporting of ADRs in humans. Here, we present a systematic analysis of associations between measured and predicted in vitro bioactivities of drugs and adverse events (AEs) in humans from two sources of data: the Side Effect Resource, derived from clinical trials, and the Food and Drug Administration Adverse Event Reporting System, derived from postmarketing surveillance. The ratio of a drug's therapeutic unbound plasma concentration over the drug's in vitro potency against a given protein was used to select proteins most likely to be relevant to in vivo effects. In examining individual target bioactivities as predictors of AEs, we found a trade-off between the positive predictive value and the fraction of drugs with AEs that can be detected. However, considering sets of multiple targets for the same AE can help identify a greater fraction of AE-associated drugs. Of the 45 targets with statistically significant associations to AEs, 30 are included on existing safety target panels. The remaining 15 targets include 9 carbonic anhydrases, of which CA5B is significantly associated with cholestatic jaundice. We include the full quantitative data on associations between measured and predicted in vitro bioactivities and AEs in humans in this work, which can be used to make a more informed selection of safety profiling targets.

5-Phenyl-1,3,4-oxadiazol-2(3H)-ones Are Potent Inhibitors of Notum Carboxylesterase Activity Identified by the Optimization of a Crystallographic Fragment Screening Hit

Carboxylesterase Notum is a negative regulator of the Wnt signaling pathway. There is an emerging understanding of the role Notum plays in disease, supporting the need to discover new small-molecule inhibitors. A crystallographic X-ray fragment screen was performed, which identified fragment hit 1,2,3-triazole 7 as an attractive starting point for a structure-based drug design hit-to-lead program. Optimization of 7 identified oxadiazol-2-one 23dd as a preferred example with properties consistent with drug-like chemical space. Screening 23dd in a cell-based TCF/LEF reporter gene assay restored the activation of Wnt signaling in the presence of Notum. Mouse pharmacokinetic studies with oral administration of 23dd demonstrated good plasma exposure and partial blood-brain barrier penetration. Significant progress was made in developing fragment hit 7 into lead 23dd (>600-fold increase in activity), making it suitable as a new chemical tool for exploring the role of Notum-mediated regulation of Wnt signaling.

Screening of a Custom-Designed Acid Fragment Library Identifies 1-Phenylpyrroles and 1-Phenylpyrrolidines as Inhibitors of Notum Carboxylesterase Activity

The Wnt family of proteins are secreted signaling proteins that play key roles in regulating cellular functions. Recently, carboxylesterase Notum was shown to act as a negative regulator of Wnt signaling by mediating the removal of an essential palmitoleate. Here we disclose two new chemical scaffolds that inhibit Notum enzymatic activity. Our approach was to create a fragment library of 250 acids for screening against Notum in a biochemical assay followed by structure determination by X-ray crystallography. Twenty fragments were identified as hits for Notum inhibition, and 14 of these fragments were shown to bind in the palmitoleate pocket of Notum. Optimization of 1-phenylpyrrole 20, guided by structure-based drug design, identified 20z as the most potent compound from this series. Similarly, the optimization of 1-phenylpyrrolidine 8 gave acid 26. This work demonstrates that inhibition of Notum activity can be achieved by small, drug-like molecules possessing favorable in vitro ADME profiles.

Changing the HTS Paradigm: AI-Driven Iterative Screening for Hit Finding

Iterative screening is a process in which screening is done in batches, with each batch filled by using machine learning to select the most promising compounds from the library based on the previous results. We believe iterative screening is poised to enhance the screening process by improving hit finding while at the same time reducing the number of compounds screened. In addition, we see this process as a key enabler of next-generation high-throughput screening (HTS), which uses more complex assays that better describe the biology but demand more resource per screened compound. To demonstrate the utility of these methods, we retrospectively analyze HTS data from PubChem with a focus on machine learning–based screening strategies that can be readily implemented in practice. Our results show that over a variety of HTS experimental paradigms, an iterative screening setup that screens a total of 35% of the screening collection over as few as three iterations has a median return rate of approximately 70% of the active compounds. Increasing the portion of the library screened to 50% yields median returns of approximately 80% of actives. Using six iterations increases these return rates to 78% and 90%, respectively. The best results were achieved with machine learning models that can be run on a standard desktop. By demonstrating that the utility of iterative screening holds true even with a small number of iterations, and without requiring significant computational resources, we provide a roadmap for the practical implementation of these techniques in hit finding.

Positively selected modifications in the pore of TbAQP2 allow pentamidine to enter Trypanosoma brucei

Mutations in the Trypanosoma brucei aquaporin AQP2 are associated with resistance to pentamidine and melarsoprol. We show that TbAQP2 but not TbAQP3 was positively selected for increased pore size from a common ancestor aquaporin. We demonstrate that TbAQP2’s unique architecture permits pentamidine permeation through its central pore and show how specific mutations in highly conserved motifs affect drug permeation. Introduction of key TbAQP2 amino acids into TbAQP3 renders the latter permeable to pentamidine. Molecular dynamics demonstrates that permeation by dicationic pentamidine is energetically favourable in TbAQP2, driven by the membrane potential, although aquaporins are normally strictly impermeable for ionic species. We also identify the structural determinants that make pentamidine a permeant although most other diamidine drugs are excluded. Our results have wide-ranging implications for optimising antitrypanosomal drugs and averting cross-resistance. Moreover, these new insights in aquaporin permeation may allow the pharmacological exploitation of other members of this ubiquitous gene family.

African sleeping sickness is a potentially deadly illness caused by the parasite Trypanosoma brucei. The disease is treatable, but many of the current treatments are old and are becoming increasingly ineffective. For instance, resistance is growing against pentamidine, a drug used in the early stages in the disease, as well as against melarsoprol, which is deployed when the infection has progressed to the brain. Usually, cases resistant to pentamidine are also resistant to melarsoprol, but it is still unclear why, as the drugs are chemically unrelated.

Studies have shown that changes in a water channel called aquaglyceroporin 2 (TbAQP2) contribute to drug resistance in African sleeping sickness; this suggests that it plays a role in allowing drugs to kill the parasite. This molecular ‘drain pipe’ extends through the surface of T. brucei, and should allow only water and a molecule called glycerol in and out of the cell. In particular, the channel should be too narrow to allow pentamidine or melarsoprol to pass through. One possibility is that, in T. brucei, the TbAQP2 channel is abnormally wide compared to other members of its family. Alternatively, pentamidine and melarsoprol may only bind to TbAQP2, and then ‘hitch a ride’ when the protein is taken into the parasite as part of the natural cycle of surface protein replacement.

Alghamdi et al. aimed to tease out these hypotheses. Computer models of the structure of the protein were paired with engineered changes in the key areas of the channel to show that, in T. brucei, TbAQP2 provides a much broader gateway into the cell than observed for similar proteins. In addition, genetic analysis showed that this version of TbAQP2 has been actively selected for during the evolution process of T. brucei. This suggests that the parasite somehow benefits from this wider aquaglyceroporin variant.

This is a new resistance mechanism, and it is possible that aquaglyceroporins are also larger than expected in other infectious microbes. The work by Alghamdi et al. therefore provides insight into how other germs may become resistant to drugs.

Using Predicted Bioactivity Profiles to Improve Predictive Modeling

Predictive modeling is a cornerstone in early drug development. Using information for multiple domains or across prediction tasks has the potential to improve the performance of predictive modeling. However, aggregating data often leads to incomplete data matrices that might be limiting for modeling. In line with previous studies, we show that by generating predicted bioactivity profiles, and using these as additional features, prediction accuracy of biological endpoints can be improved. Using conformal prediction, a type of confidence predictor, we present a robust framework for the calculation of these profiles and the evaluation of their impact. We report on the outcomes from several approaches to generate the predicted profiles on 16 datasets in cytotoxicity and bioactivity and show that efficiency is improved the most when including the p-values from conformal prediction as bioactivity profiles.

Scaffold-hopping identifies furano[2,3-d]pyrimidine amides as potent Notum inhibitors

The carboxylesterase Notum is a key negative regulator of the Wnt signaling pathway by mediating the depalmitoleoylation of Wnt proteins. Our objective was to discover potent small molecule inhibitors of Notum suitable for exploring the regulation of Wnt signaling in the central nervous system. Scaffold-hopping from thienopyrimidine acids 1 and 2, supported by X-ray structure determination, identified 3-methylimidazolin-4-one amides 20-24 as potent inhibitors of Notum with activity across three orthogonal assay formats (biochemical, extra-cellular, occupancy). A preferred example 24 demonstrated good stability in mouse microsomes and plasma, and cell permeability in the MDCK-MDR1 assay albeit with modest P-gp mediated efflux. Pharmacokinetic studies with 24 were performed in vivo in mouse with single oral administration of 24 showing good plasma exposure and reasonable CNS penetration. We propose that 24 is a new chemical tool suitable for cellular studies to explore the fundamental biology of Notum.

Elevated intracellular cAMP concentration mediates growth suppression in glioma cells

Supressed levels of intracellular cAMP have been associated with malignancy. Thus, elevating cAMP through activation of adenylyl cyclase (AC) or by inhibition of phosphodiesterase (PDE) may be therapeutically beneficial. Here, we demonstrate that elevated cAMP levels suppress growth in C6 cells (a model of glioma) through treatment with forskolin, an AC activator, or a range of small molecule PDE inhibitors with differing selectivity profiles. Forskolin suppressed cell growth in a PKA-dependent manner by inducing a G₂/M phase cell cycle arrest. In contrast, trequinsin (a non-selective PDE2/3/7 inhibitor), not only inhibited cell growth via PKA, but also stimulated (independent of PKA) caspase-3/-7 and induced an aneuploidy phenotype. Interestingly, a cocktail of individual PDE 2,3,7 inhibitors suppressed cell growth in a manner analogous to forskolin but not trequinsin. Finally, we demonstrate that concomitant targeting of both AC and PDEs synergistically elevated intracellular cAMP levels thereby potentiating their antiproliferative actions.

LightGBM: An Effective and Scalable Algorithm for Prediction of Chemical Toxicity-Application to the Tox21 and Mutagenicity Data Sets

Machine learning algorithms have attained widespread use in assessing the potential toxicities of pharmaceuticals and industrial chemicals because of their faster speed and lower cost compared to experimental bioassays. Gradient boosting is an effective algorithm that often achieves high predictivity, but historically the relative long computational time limited its applications in predicting large compound libraries or developing in silico predictive models that require frequent retraining. LightGBM, a recent improvement of the gradient boosting algorithm, inherited its high predictivity but resolved its scalability and long computational time by adopting a leaf-wise tree growth strategy and introducing novel techniques. In this study, we compared the predictive performance and the computational time of LightGBM to deep neural networks, random forests, support vector machines, and XGBoost. All algorithms were rigorously evaluated on publicly available Tox21 and mutagenicity data sets using a Bayesian optimization integrated nested 10-fold cross-validation scheme that performs hyperparameter optimization while examining model generalizability and transferability to new data. The evaluation results demonstrated that LightGBM is an effective and highly scalable algorithm offering the best predictive performance while consuming significantly shorter computational time than the other investigated algorithms across all Tox21 and mutagenicity data sets. We recommend LightGBM for applications of in silico safety assessment and also other areas of cheminformatics to fulfill the ever-growing demand for accurate and rapid prediction of various toxicity or activity related end points of large compound libraries present in the pharmaceutical and chemical industry.

Scrutinizing the cut-off for "pathological" meniscal body extrusion on knee MRI

OBJECTIVES

Medial meniscal body extrusion ≥ 3 mm on MRI is often considered "pathologic." The aims of this study were to (1) assess the adequacy of 3 mm as cut-off for "pathological" extrusion and (2) find an optimal cut-off for meniscal extrusion cross-sectionally associated with radiographic knee osteoarthritis, bone marrow lesions (BMLs), and cartilage damage.

METHODS

Nine hundred fifty-eight persons, aged 50-90 years from Framingham, MA, USA, had readable 1.5 T MRI scans of the right knee for meniscal body extrusion (measured in mm). BMLs and cartilage damage were read using the whole organ magnetic resonance imaging score (WORMS). Knee X-rays were read according to the Kellgren and Lawrence (KL) scale. We evaluated the performance of the 3-mm cut-off with respect to the three outcomes and estimated a new cut-off maximizing the sum of sensitivity and specificity.

RESULTS

The study persons had mean age of 62.2 years, 57.0% were women and the mean body mass index was 28.5 kg/m2. Knees with radiographic osteoarthritis, BMLs, and cartilage damage had overall more meniscal extrusion than knees without. The 3-mm cut-off had moderate sensitivity and low specificity for all three outcomes (sensitivity between 0.68 [95% CI 0.63-0.73] and 0.81 [0.73-0.87], specificity between 0.49 [0.45-0.52] and 0.54 [0.49-0.58]). Using 4 mm maximized the sum of sensitivity and specificity and improved the percentage of correctly classified subjects (from between 54 and 61% to between 64 and 79%).

CONCLUSIONS

The 4-mm cut-off may be used as an alternative cut-off for denoting pathological meniscal extrusion.

KEY POINTS

• Medial meniscal body extrusion is strongly associated with osteoarthritis. • The 3-mm cut-off for medial meniscal body extrusion has high sensitivity but low specificity with respect to bone marrow lesions, cartilage damage, and radiographic osteoarthritis. • The 4-mm cut-off maximizes the sensitivity and specificity with respect to all three osteoarthritis features.

Multitask Modeling with Confidence Using Matrix Factorization and Conformal Prediction

Multitask prediction of bioactivities is often faced with challenges relating to the sparsity of data and imbalance between different labels. We propose class conditional (Mondrian) conformal predictors using underlying Macau models as a novel approach for large scale bioactivity prediction. This approach handles both high degrees of missing data and label imbalances while still producing high quality predictive models. When applied to ten assay end points from PubChem, the models generated valid models with an efficiency of 74.0-80.1% at the 80% confidence level with similar performance both for the minority and majority class. Also when deleting progressively larger portions of the available data (0-80%) the performance of the models remained robust with only minor deterioration (reduction in efficiency between 5 and 10%). Compared to using Macau without conformal prediction the method presented here significantly improves the performance on imbalanced data sets.

Meniscal body extrusion and cartilage coverage in middle-aged and elderly without radiographic knee osteoarthritis

OBJECTIVES

To determine meniscal extrusion and cartilage coverage on magnetic resonance (MR) images and factors associated with these parameters in knees of middle-aged and elderly persons free from radiographic tibiofemoral osteoarthritis (OA).

METHODS

Seven hundred eighteen persons, free of radiographic tibiofemoral OA, aged 50-90 years from Framingham, MA, USA, were included. We measured meniscal extrusion on 1.5 T MRI of both knees to evaluate both medial and lateral meniscal body extrusion and cartilage coverage. We also determined meniscal morphology and structural integrity. The multivariable association with age, body mass index (BMI), and ipsilateral meniscal damage was also evaluated.

RESULTS

The mean meniscal body extrusion medially was 2.7 mm and laterally 1.8 mm. The tibial cartilage coverage was about 30% of ipsilateral cartilage surface (both compartments). The presence of ipsilateral meniscal damage was associated with more extrusion in only the medial compartment, 1.0 mm in men and 0.6 mm in women, and less cartilage coverage proportion, -5.5% in men and -4.6% in women.

CONCLUSIONS

Mean medial meniscal body extrusion in middle-aged or older persons without radiographic tibiofemoral OA approximates the commonly used cutoff (3 mm) to denote pathological extrusion. Medial meniscal damage is a factor associated with medial meniscal body extrusion and less cartilage coverage.

KEY POINTS

• Medial meniscal extrusion in middle-aged/older persons without OA is around 3 mm. • Lateral meniscal extrusion in middle-aged/older persons without OA is around 2 mm. • Meniscal damage is associated with medial meniscal extrusion and less cartilage coverage.

Therapy for glioblastoma: is it working?

Glioblastoma (GBM) remains one of the most intransigent of cancers, with a median overall survival of only 15 months after diagnosis. Drug treatments have largely proven ineffective; it is thought that this is related to the heterogeneous nature and plasticity of GBM-initiating stem cell lineages. Although many combination drug therapies are being positioned to address tumour heterogeneity, the most promising therapeutic approaches for GBM to date appear to be those targeting GBM by vaccination or antibody- and cell-based immunotherapy. We review the most recent clinical trials for GBM and discuss the role of adaptive clinical trials in developing personalised treatment strategies to address intra- and inter-tumoral heterogeneity.

Spring and Late Summer Phytoplankton Biomass Impact on the Coastal Sediment Microbial Community Structure

Two annual Baltic Sea phytoplankton blooms occur in spring and summer. The bloom intensity is determined by nutrient concentrations in the water, while the period depends on weather conditions. During the course of the bloom, dead cells sink to the sediment where their degradation consumes oxygen to create hypoxic zones (< 2 mg/L dissolved oxygen). These zones prevent the establishment of benthic communities and may result in fish mortality. The aim of the study was to determine how the spring and autumn sediment chemistry and microbial community composition changed due to degradation of diatom or cyanobacterial biomass, respectively. Results from incubation of sediment cores showed some typical anaerobic microbial processes after biomass addition such as a decrease in NO₂^- + NO₃^- in the sediment surface (0-1 cm) and iron in the underlying layer (1-2 cm). In addition, an increase in NO₂^- + NO₃^- was observed in the overlying benthic water in all amended and control incubations. The combination of NO₂^- + NO₃^- diffusion plus nitrification could not account for this increase. Based on 16S rRNA gene sequences, the addition of cyanobacterial biomass during autumn caused a large increase in ferrous iron-oxidizing archaea while diatom biomass amendment during spring caused minor changes in the microbial community. Considering that OTUs sharing lineages with acidophilic microorganisms had a high relative abundance during autumn, it was suggested that specific niches developed in sediment microenvironments. These findings highlight the importance of nitrogen cycling and early microbial community changes in the sediment due to sinking phytoplankton before potential hypoxia occurs.

Information-Derived Mechanistic Hypotheses for Structural Cardiotoxicity

Adverse events resulting from drug therapy can be a cause of drug withdrawal, reduced and or restricted clinical use, as well as a major economic burden for society. To increase the safety of new drugs, there is a need to better understand the mechanisms causing the adverse events. One way to derive new mechanistic hypotheses is by linking data on drug adverse events with the drugs' biological targets. In this study, we have used data mining techniques and mutual information statistical approaches to find associations between reported adverse events collected from the FDA Adverse Event Reporting System and assay outcomes from ToxCast, with the aim to generate mechanistic hypotheses related to structural cardiotoxicity (morphological damage to cardiomyocytes and/or loss of viability). Our workflow identified 22 adverse event-assay outcome associations. From these associations, 10 implicated targets could be substantiated with evidence from previous studies reported in the literature. For two of the identified targets, we also describe a more detailed mechanism, forming putative adverse outcome pathways associated with structural cardiotoxicity. Our study also highlights the difficulties deriving these type of associations from the very limited amount of data available.

WINDOW consortium: A path towards increased therapy efficacy against glioblastoma

Glioblastoma is the most common and malignant form of brain cancer, for which the standard treatment is maximal surgical resection, radiotherapy and chemotherapy. Despite these interventions, mean overall survival remains less than 15 months, during which extensive tumor infiltration throughout the brain occurs. The resulting metastasized cells in the brain are characterized by chemotherapy resistance and extensive intratumoral heterogeneity. An orthogonal approach attacking both intracellular resistance mechanisms as well as intercellular heterogeneity is necessary to halt tumor progression. For this reason, we established the WINDOW Consortium (Window for Improvement for Newly Diagnosed patients by Overcoming disease Worsening), in which we are establishing a strategy for rational selection and development of effective therapies against glioblastoma. Here, we overview the many challenges posed in treating glioblastoma, including selection of drug combinations that prevent therapy resistance, the need for drugs that have improved blood brain barrier penetration and strategies to counter heterogeneous cell populations within patients. Together, this forms the backbone of our strategy to attack glioblastoma.

Orthologue chemical space and its influence on target prediction

Motivation

In silico approaches often fail to utilize bioactivity data available for orthologous targets due to insufficient evidence highlighting the benefit for such an approach. Deeper investigation into orthologue chemical space and its influence toward expanding compound and target coverage is necessary to improve the confidence in this practice.

Results

Here we present analysis of the orthologue chemical space in ChEMBL and PubChem and its impact on target prediction. We highlight the number of conflicting bioactivities between human and orthologues is low and annotations are overall compatible. Chemical space analysis shows orthologues are chemically dissimilar to human with high intra-group similarity, suggesting they could effectively extend the chemical space modelled. Based on these observations, we show the benefit of orthologue inclusion in terms of novel target coverage. We also benchmarked predictive models using a time-series split and also using bioactivities from Chemistry Connect and HTS data available at AstraZeneca, showing that orthologue bioactivity inclusion statistically improved performance.

Availability and implementation

Orthologue-based bioactivity prediction and the compound training set are available at www.github.com/lhm30/PIDGINv2.

Contact

ab454@cam.ac.uk.

Supplementary information

Supplementary data are available at Bioinformatics online.

Maximizing gain in high-throughput screening using conformal prediction

Iterative screening has emerged as a promising approach to increase the efficiency of screening campaigns compared to traditional high throughput approaches. By learning from a subset of the compound library, inferences on what compounds to screen next can be made by predictive models, resulting in more efficient screening. One way to evaluate screening is to consider the cost of screening compared to the gain associated with finding an active compound. In this work, we introduce a conformal predictor coupled with a gain-cost function with the aim to maximise gain in iterative screening. Using this setup we were able to show that by evaluating the predictions on the training data, very accurate predictions on what settings will produce the highest gain on the test data can be made. We evaluate the approach on 12 bioactivity datasets from PubChem training the models using 20% of the data. Depending on the settings of the gain-cost function, the settings generating the maximum gain were accurately identified in 8-10 out of the 12 datasets. Broadly, our approach can predict what strategy generates the highest gain based on the results of the cost-gain evaluation: to screen the compounds predicted to be active, to screen all the remaining data, or not to screen any additional compounds. When the algorithm indicates that the predicted active compounds should be screened, our approach also indicates what confidence level to apply in order to maximize gain. Hence, our approach facilitates decision-making and allocation of the resources where they deliver the most value by indicating in advance the likely outcome of a screening campaign.

Computer-aided design of multi-target ligands at A1R, A2AR and PDE10A, key proteins in neurodegenerative diseases

Compounds designed to display polypharmacology may have utility in treating complex diseases, where activity at multiple targets is required to produce a clinical effect. In particular, suitable compounds may be useful in treating neurodegenerative diseases by promoting neuronal survival in a synergistic manner via their multi-target activity at the adenosine A1 and A2A receptors (A1R and A2AR) and phosphodiesterase 10A (PDE10A), which modulate intracellular cAMP levels. Hence, in this work we describe a computational method for the design of synthetically feasible ligands that bind to A1 and A2A receptors and inhibit phosphodiesterase 10A (PDE10A), involving a retrosynthetic approach employing in silico target prediction and docking, which may be generally applicable to multi-target compound design at several target classes. This approach has identified 2-aminopyridine-3-carbonitriles as the first multi-target ligands at A1R, A2AR and PDE10A, by showing agreement between the ligand and structure based predictions at these targets. The series were synthesized via an efficient one-pot scheme and validated pharmacologically as A1R/A2AR-PDE10A ligands, with IC50 values of 2.4-10.0 μM at PDE10A and Ki values of 34-294 nM at A1R and/or A2AR. Furthermore, selectivity profiling of the synthesized 2-amino-pyridin-3-carbonitriles against other subtypes of both protein families showed that the multi-target ligand 8 exhibited a minimum of twofold selectivity over all tested off-targets. In addition, both compounds 8 and 16 exhibited the desired multi-target profile, which could be considered for further functional efficacy assessment, analog modification for the improvement of selectivity towards A1R, A2AR and PDE10A collectively, and evaluation of their potential synergy in modulating cAMP levels.

Modelling compound cytotoxicity using conformal prediction and PubChem HTS data

The assessment of compound cytotoxicity is an important part of the drug discovery process. Accurate predictions of cytotoxicity have the potential to expedite decision making and save considerable time and effort. In this work we apply class conditional conformal prediction to model the cytotoxicity of compounds based on 16 high throughput cytotoxicity assays from PubChem. The data span 16 cell lines and comprise more than 440 000 unique compounds. The data sets are heavily imbalanced with only 0.8% of the tested compounds being cytotoxic. We trained one classification model for each cell line and validated the performance with respect to validity and accuracy. The generated models deliver high quality predictions for both toxic and non-toxic compounds despite the imbalance between the two classes. On external data collected from the same assay provider as one of the investigated cell lines the model had a sensitivity of 74% and a specificity of 65% at the 80% confidence level among the compounds assigned to a single class. Compared to previous approaches for large scale cytotoxicity modelling, this represents a balanced performance in the prediction of the toxic and non-toxic classes. The conformal prediction framework also allows the modeller to control the error frequency of the predictions, allowing predictions of cytotoxicity outcomes with confidence.

A Decision support system (DSS) for municipal nurses encountering health deterioration among older people

Background

This study is part of a larger project called ViSam and includes testing of a decision support system developed and adapted for older people on the basis of M (R) ETTS (Rapid Emergency Triage and Treatment System). The system is designed to allow municipal nurses to determine the optimal level of care for older people whose health has deteriorated. This new system will allow more structured assessment, the patient should receive optimal care and improved data transmission to the next caregiver.

Methods

This study has an explanatory approach, commencing with quantitative data collection phase followed by qualitative data arising from focus group discussions over the RNs professional experience using the Decision Support system. Focus group discussions were performed to complement the quantitative data to get a more holistic view of the decision support system.

Results

Using elements of the decision support system (vital parameters for saturation, pain and affected general health) together with the nurses' decision showed that 94 % of the older persons referred to hospital were ultimately hospitalized. Nurses felt that they worked more systematically, communicated more effectively with others and felt more professional when using the decision support system.

Conclusions

The results of this study showed that, with the help of a decision support system, the correct patients are sent to the Emergency Department from municipal home care. Unnecessary referrals of older patients that might lead to poorer health, decreased well-being and confusion can thus be avoided. Using the decision support system means that healthcare co-workers (nurses, ambulance/emergency department/district doctor/SOS alarm) begin to communicate more optimally. There is increased understanding leading to the risk of misinterpretation being reduced and the relationship between healthcare co-workers is improved. However, the decision support system requires more extensive testing in order to enhance the evidence base relating to the vital parameters among older people and the use of the decision support system.

Inhibition of Insulin-Regulated Aminopeptidase (IRAP) by Arylsulfonamides

The inhibition of insulin-regulated aminopeptidase (IRAP, EC 3.4.11.3) by angiotenesin IV is known to improve memory and learning in rats. Screening 10 500 low-molecular-weight compounds in an enzyme inhibition assay with IRAP from Chinese Hamster Ovary (CHO) cells provided an arylsulfonamide (N-(3-(1H-tetrazol-5-yl)phenyl)-4-bromo-5-chlorothiophene-2-sulfonamide), comprising a tetrazole in the meta position of the aromatic ring, as a hit. Analogues of this hit were synthesized, and their inhibitory capacities were determined. A small structure–activity relationship study revealed that the sulfonamide function and the tetrazole ring are crucial for IRAP inhibition. The inhibitors exhibited a moderate inhibitory potency with an IC₅₀=1.1±0.5 μm for the best inhibitor in the series. Further optimization of this new class of IRAP inhibitors is required to make them attractive as research tools and as potential cognitive enhancers.