Reducing overconfident errors in molecular property classification using Posterior Network

Deep-learning-based classification models are increasingly used for predicting molecular properties in drug development. However, traditional classification models using the Softmax function often give overconfident mispredictions for out-of-distribution samples, highlighting a critical lack of accurate uncertainty estimation. Such limitations can result in substantial costs and should be avoided during drug development. Inspired by advances in evidential deep learning and Posterior Network, we replaced the Softmax function with a normalizing flow to enhance the uncertainty estimation ability of the model in molecular property classification. The proposed strategy was evaluated across diverse scenarios, including simulated experiments based on a synthetic dataset, ADMET predictions, and ligand-based virtual screening. The results demonstrate that compared with the vanilla model, the proposed strategy effectively alleviates the problem of giving overconfident but incorrect predictions. Our findings support the promising application of evidential deep learning in drug development and offer a valuable framework for further research.

Identification of important gene signatures in schizophrenia through feature fusion and genetic algorithm

Schizophrenia is a debilitating psychiatric disorder that can significantly affect a patient's quality of life and lead to permanent brain damage. Although medical research has identified certain genetic risk factors, the specific pathogenesis of the disorder remains unclear. Despite the prevalence of research employing magnetic resonance imaging, few studies have focused on the gene level and gene expression profile involving a large number of screened genes. However, the high dimensionality of genetic data presents a great challenge to accurately modeling the data. To tackle the current challenges, this study presents a novel feature selection strategy that utilizes heuristic feature fusion and a multi-objective optimization genetic algorithm. The goal is to improve classification performance and identify the key gene subset for schizophrenia diagnostics. Traditional gene screening techniques are inadequate for accurately determining the precise number of key genes associated with schizophrenia. Our innovative approach integrates a filter-based feature selection method to reduce data dimensionality and a multi-objective optimization genetic algorithm for improved classification tasks. By combining the filtering and wrapper methods, our strategy leverages their respective strengths in a deliberate manner, leading to superior classification accuracy and a more efficient selection of relevant genes. This approach has demonstrated significant improvements in classification results across 11 out of 14 relevant datasets. The performance on the remaining three datasets is comparable to the existing methods. Furthermore, visual and enrichment analyses have confirmed the practicality of our proposed method as a promising tool for the early detection of schizophrenia.

The SMAC mimetic GDC-0152 is a direct ABCB1-ATPase activity modulator and BIRC5 expression suppressor in cancer cells

Upregulation of the multidrug efflux pump ABCB1/MDR1 (P-gp) and the anti-apoptotic protein BIRC5/Survivin promotes multidrug resistance in various human cancers. GDC-0152 is a DIABLO/SMAC mimetic currently being tested in patients with solid tumors. However, it is still unclear whether GDC-0152 is therapeutically applicable for patients with ABCB1-overexpressing multidrug-resistant tumors, and the molecular mechanism of action of GDC-0152 in cancer cells is still incompletely understood. In this study, we found that the potency of GDC-0152 is unaffected by the expression of ABCB1 in cancer cells. Interestingly, through in silico and in vitro analysis, we discovered that GDC-0152 directly modulates the ABCB1-ATPase activity and inhibits ABCB1 multidrug efflux activity at sub-cytotoxic concentrations (i.e., 0.25×IC50 or less). Further investigation revealed that GDC-0152 also decreases BIRC5 expression, induces mitophagy, and lowers intracellular ATP levels in cancer cells at low cytotoxic concentrations (i.e., 0.5×IC50). Co-treatment with GDC-0152 restored the sensitivity to the known ABCB1 substrates, including paclitaxel, vincristine, and YM155 in ABCB1-expressing multidrug-resistant cancer cells, and it also restored the sensitivity to tamoxifen in BIRC5-overexpressing tamoxifen-resistant breast cancer cells in vitro. Moreover, co-treatment with GDC-0152 restored and potentiated the anticancer effects of paclitaxel in ABCB1 and BIRC5 co-expressing xenograft tumors in vivo. In conclusion, GDC-0152 has the potential for use in the management of cancer patients with ABCB1 and BIRC5-related drug resistance. The findings of our study provide essential information to physicians for designing a more patient-specific GDC-0152 clinical trial program in the future.

Combined in silico and in vitro approaches to identify P-glycoprotein-inhibiting pesticides

The P-glycoprotein (P-gp) efflux pump plays a major role in xenobiotic detoxification. The inhibition of its activity by environmental contaminants remains however rather little characterised. The present study was designed to develop a combination of different approaches to identify P-gp inhibitors among a large number of pesticides using in silico and in vitro models. First, the prediction performance of four web tools was evaluated alone or in combination using a set of recently marketed drugs. The best combination of web tools-AdmetSAR2.0/PgpRules/pkCSM-was next used to predict P-gp activity inhibition by 762 pesticides. Among the 187 pesticides predicted to be P-gp inhibitors, 11 were tested in vitro for their ability to inhibit the efflux of reference substrates (rhodamine 123 and Hoechst 33342) in P-gp overexpressing MCF7R cells and to inhibit the efflux of the reference substrate rhodamine 123 in the Caco-2 cell monolayer. In MCF7R cell assays, ivermectin B1a, emamectin B1 benzoate, spinosad, dimethomorph and tralkoxydim inhibited P-gp activity; ivermectin B1a, emamectin B1 benzoate and spinosad were determined to be stronger inhibitors (half-maximal inhibitory concentration [IC50 ] of 3 ± 1, 5 ± 1 and 7 ± 1 µM, respectively) than dimethomorph and tralkoxydim (IC50 of 102 ± 7 and 88 ± 7 µM, respectively). Ivermectin B1a, emamectin B1 benzoate, spinosad and dimethomorph also inhibited P-gp activity in Caco-2 cell monolayer assays, with dimethomorph being a weaker P-gp inhibitor. These combined approaches could be used to identify P-gp inhibitors among food contaminants, but need to be optimised and adapted for high-throughput screening.

Principal Component Analysis (PCA) of Molecular Descriptors for Improving Permeation through the Blood-Brain Barrier of Quercetin Analogues

Despite its beneficial pharmacological effects in the brain, partly by modulating inositol phosphate multikinase (IPMK) activity, the therapeutic use of quercetin is limited due to its poor solubility, low oral bioavailability, and low permeability through the blood-brain barrier (BBB). We aimed to identify quercetin analogues with improved BBB permeability and preserved binding affinities towards IPMK and to identify the molecular characteristics required for them to permeate the BBB. Binding affinities of quercetin analogues towards IPMK were determined by molecular docking. Principal component analysis (PCA) was applied to identify the molecular descriptors contributing to efficient permeation through the BBB. Among 34 quercetin analogues, 19 compounds were found to form more stable complexes with IPMK, and the vast majority were found to be more lipophilic than quercetin. Using two distinct in silico techniques, insufficient BBB permeation was determined for all quercetin analogues. However, using the PCA method, the descriptors related to intrinsic solubility and lipophilicity (logP) were identified as mainly responsible for clustering four quercetin analogues (trihydroxyflavones) with the highest BBB permeability. The application of PCA revealed that quercetin analogues could be classified with respect to their structural characteristics, which may be utilized in further analogue syntheses and lead optimization of BBB-penetrating IPMK modulators as neuroprotective agents.

Virtual Screening and ADMET Prediction to Uncover the Potency of Flavonoids from Genus Erythrina as Antibacterial Agent through Inhibition of Bacterial ATPase DNA Gyrase B

The emergence of antimicrobial resistance due to the widespread and inappropriate use of antibiotics has now become the global health challenge. Flavonoids have long been reported to be a potent antimicrobial agent against a wide range of pathogenic microorganisms in vitro. Therefore, new antibiotics development based on flavonoid structures could be a potential strategy to fight against antibiotic-resistant infections. This research aims to screen the potency of flavonoids of the genus Erythrina as an inhibitor of bacterial ATPase DNA gyrase B. From the 378 flavonoids being screened, 49 flavonoids show potential as an inhibitor of ATPase DNA gyrase B due to their lower binding affinity compared to the inhibitor and ATP. Further screening for their toxicity, we identified 6 flavonoids from these 49 flavonoids, which are predicted to have low toxicity. Among these flavonoids, erystagallin B (334) is predicted to have the best pharmacokinetic properties, and therefore, could be further developed as new antibacterial agent.

Recent Studies of Artificial Intelligence on In Silico Drug Absorption

Absorption is an important area of research in pharmacochemistry and drug development, because the drug has to be absorbed before any drug effects can occur. Furthermore, the ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) profile of drugs can be directly and considerably altered by modulating factors affecting absorption. Many drugs in development fail because of poor absorption. The research and continuous efforts of researchers in recent years have brought many successes and promises in drug absorption property prediction, especially in silico, which helps to reduce the time and cost significantly for screening undesirable drug candidates. In this report, we explicitly provide an overview of recent in silico studies on predicting absorption properties, especially from 2019 to the present, using artificial intelligence. Additionally, we have collected and investigated public databases that support absorption prediction research. On those grounds, we also proposed the challenges and development directions of absorption prediction in the future. We hope this review can provide researchers with valuable guidelines on absorption prediction to facilitate the development of newer approaches in drug discovery.

Recent Advances on P-Glycoprotein (ABCB1) Transporter Modelling with In Silico Methods

ABC transporters play a critical role in both drug bioavailability and toxicity, and with the discovery of the P-glycoprotein (P-gp), this became even more evident, as it plays an important role in preventing intracellular accumulation of toxic compounds. Over the past 30 years, intensive studies have been conducted to find new therapeutic molecules to reverse the phenomenon of multidrug resistance (MDR) ), that research has found is often associated with overexpression of P-gp, the most extensively studied drug efflux transporter; in MDR, therapeutic drugs are prevented from reaching their targets due to active efflux from the cell. The development of P-gp inhibitors is recognized as a good way to reverse this type of MDR, which has been the subject of extensive studies over the past few decades. Despite the progress made, no effective P-gp inhibitors to reverse multidrug resistance are yet on the market, mainly because of their toxic effects. Computational studies can accelerate this process, and in silico models such as QSAR models that predict the activity of compounds associated with P-gp (or analogous transporters) are of great value in the early stages of drug development, along with molecular modelling methods, which provide a way to explain how these molecules interact with the ABC transporter. This review highlights recent advances in computational P-gp research, spanning the last five years to 2022. Particular attention is given to the use of machine-learning approaches, drug-transporter interactions, and recent discoveries of potential P-gp inhibitors that could act as modulators of multidrug resistance.

Cancer multidrug-resistance reversal by ABCB1 inhibition: A recent update

Chemotherapy is one of the most common treatments for cancer that uses one or more anti-cancer drugs as a part of the standardized chemotherapy regimen. Cytotoxic chemicals delay and prevent cancer cells from multiplying, invading, and metastasizing. However, the significant drawbacks of cancer chemotherapy are the lack of selectivity of the cytotoxic drugs to tumour cells and normal cells and the development of resistance by cells for the particular drug or the combination of drugs. Multidrug resistance (MDR) is the low sensitivity of specific cells against drugs associated with cancer chemotherapy. The most common mechanisms of anticancer drug resistance are: (a) drug-dependent MDR (b) target-dependent MDR, and (c) drug target-independent MDR. In all the factors, the overexpression of multidrug efflux systems contributes significantly to the increased resistance in the cancer cells. Multidrug resistance due to efflux of anticancer drugs by membrane ABC transporters includes ABCB1, ABCC1, and ABCG2. ABCB1 inhibition can restore the sensitivity of the cancerous cells toward chemotherapeutic drugs. In this review, we discussed ABCB1 inhibitors under clinical studies with their mode of action, potency and selectivity. Also, we have highlighted the contribution of repurposing drugs, biologics and nano formulation strategies to combat multidrug resistance by modulating the ABCB1 activity.

In Silico Approaches for Addressing Challenges in CNS Radiopharmaceutical Design

Positron emission tomography (PET) is a highly sensitive and versatile molecular imaging modality that leverages radiolabeled molecules, known as radiotracers, to interrogate biochemical processes such as metabolism, enzymatic activity, and receptor expression. The ability to probe specific molecular and cellular events longitudinally in a noninvasive manner makes PET imaging a particularly powerful technique for studying the central nervous system (CNS) in both health and disease. Unfortunately, developing and translating a single CNS PET tracer for clinical use is typically an extremely resource-intensive endeavor, often requiring synthesis and evaluation of numerous candidate molecules. While existing in vitro methods are beginning to address the challenge of derisking molecules prior to costly in vivo PET studies, most require a significant investment of resources and possess substantial limitations. In the context of CNS drug development, significant time and resources have been invested into the development and optimization of computational methods, particularly involving machine learning, to streamline the design of better CNS therapeutics. However, analogous efforts developed and validated for CNS radiotracer design are conspicuously limited. In this Perspective, we overview the requirements and challenges of CNS PET tracer design, survey the most promising computational methods for in silico CNS drug design, and bridge these two areas by discussing the potential applications and impact of computational design tools in CNS radiotracer design.

i2APP: A Two-Step Machine Learning Framework For Antiparasitic Peptides Identification

Parasites can cause enormous damage to their hosts. Studies have shown that antiparasitic peptides can inhibit the growth and development of parasites and even kill them. Because traditional biological methods to determine the activity of antiparasitic peptides are time-consuming and costly, a method for large-scale prediction of antiparasitic peptides is urgently needed. We propose a computational approach called i2APP that can efficiently identify APPs using a two-step machine learning (ML) framework. First, in order to solve the imbalance of positive and negative samples in the training set, a random under sampling method is used to generate a balanced training data set. Then, the physical and chemical features and terminus-based features are extracted, and the first classification is performed by Light Gradient Boosting Machine (LGBM) and Support Vector Machine (SVM) to obtain 264-dimensional higher level features. These features are selected by Maximal Information Coefficient (MIC) and the features with the big MIC values are retained. Finally, the SVM algorithm is used for the second classification in the optimized feature space. Thus the prediction model i2APP is fully constructed. On independent datasets, the accuracy and AUC of i2APP are 0.913 and 0.935, respectively, which are better than the state-of-arts methods. The key idea of the proposed method is that multi-level features are extracted from peptide sequences and the higher-level features can distinguish well the APPs and non-APPs.

Ensemble modeling with machine learning and deep learning to provide interpretable generalized rules for classifying CNS drugs with high prediction power

The trade-off between a machine learning (ML) and deep learning (DL) model's predictability and its interpretability has been a rising concern in central nervous system-related quantitative structure-activity relationship (CNS-QSAR) analysis. Many state-of-the-art predictive modeling failed to provide structural insights due to their black box-like nature. Lack of interpretability and further to provide easy simple rules would be challenging for CNS-QSAR models. To address these issues, we develop a protocol to combine the power of ML and DL to generate a set of simple rules that are easy to interpret with high prediction power. A data set of 940 market drugs (315 CNS-active, 625 CNS-inactive) with support vector machine and graph convolutional network algorithms were used. Individual ML/DL modeling methods were also constructed for comparison. The performance of these models was evaluated using an additional external dataset of 117 market drugs (42 CNS-active, 75 CNS-inactive). Fingerprint-split validation was adopted to ensure model stringency and generalizability. The resulting novel hybrid ensemble model outperformed other constituent traditional QSAR models with an accuracy of 0.96 and an F1 score of 0.95. With the power of the interpretability provided with this protocol, our model laid down a set of simple physicochemical rules to determine whether a compound can be a CNS drug using six sub-structural features. These rules displayed higher classification ability than classical guidelines, with higher specificity and more mechanistic insights than just for blood-brain barrier permeability. This hybrid protocol can potentially be used for other drug property predictions.

Triterpenoids with modified A-ring as modulators of P-gp-dependent drug-resistance in cancer cells

Semi-synthetic A-cycle modified triterpenic derivatives with A-cycle condensed with a heterocyclic fragment (compound 1) and fragmented A-ring (compound 2) were tested for cytotoxicity against several tumor cell cultures and doxorubicin (Dox)-resistant cell lines. The equal cytotoxicity of the tested compounds to the parental tumor cell lines (HBL-100, K562) and their resistant subclones (HBL-100/Dox, K562/i-S9) was revealed. The overexpression of ABCB1 (MDR1) gene and P-glycoprotein (P-gp) was confirmed for both resistant subclones of tumor cells. Compounds 1 and 2 were shown to inhibit the ABC-transporter gene expression (MDR1, MRP, MVP, and BCRP) and the transport of well-known P-gp substrate Rhodamine 123 from resistant cells. The docking of triterpenoids 1 and 2 into the drug binding site of P-gp revealed a similarity between the conformation of the tested triterpenoids and that of classical inhibitor verapamil, thus assuming these compounds to be more likely the inhibitors than the substrates of P-gp. Any tested triterpenic derivatives, when combined at non-toxic concentrations with doxorubicin, improved cytotoxic effect of the therapeutic drug against resistant subclones of tumor cells.

FP-ADMET: a compendium of fingerprint-based ADMET prediction models

MOTIVATION

The absorption, distribution, metabolism, excretion, and toxicity (ADMET) of drugs plays a key role in determining which among the potential candidates are to be prioritized. In silico approaches based on machine learning methods are becoming increasing popular, but are nonetheless limited by the availability of data. With a view to making both data and models available to the scientific community, we have developed FPADMET which is a repository of molecular fingerprint-based predictive models for ADMET properties. In this article, we have examined the efficacy of fingerprint-based machine learning models for a large number of ADMET-related properties. The predictive ability of a set of 20 different binary fingerprints (based on substructure keys, atom pairs, local path environments, as well as custom fingerprints such as all-shortest paths) for over 50 ADMET and ADMET-related endpoints have been evaluated as part of the study. We find that for a majority of the properties, fingerprint-based random forest models yield comparable or better performance compared with traditional 2D/3D molecular descriptors.

AVAILABILITY

The models are made available as part of open access software that can be downloaded from https://gitlab.com/vishsoft/fpadmet .

Machine learning models for classification tasks related to drug safety

In this review, we outline the current trends in the field of machine learning-driven classification studies related to ADME (absorption, distribution, metabolism and excretion) and toxicity endpoints from the past six years (2015-2021). The study focuses only on classification models with large datasets (i.e. more than a thousand compounds). A comprehensive literature search and meta-analysis was carried out for nine different targets: hERG-mediated cardiotoxicity, blood-brain barrier penetration, permeability glycoprotein (P-gp) substrate/inhibitor, cytochrome P450 enzyme family, acute oral toxicity, mutagenicity, carcinogenicity, respiratory toxicity and irritation/corrosion. The comparison of the best classification models was targeted to reveal the differences between machine learning algorithms and modeling types, endpoint-specific performances, dataset sizes and the different validation protocols. Based on the evaluation of the data, we can say that tree-based algorithms are (still) dominating the field, with consensus modeling being an increasing trend in drug safety predictions. Although one can already find classification models with great performances to hERG-mediated cardiotoxicity and the isoenzymes of the cytochrome P450 enzyme family, these targets are still central to ADMET-related research efforts.

Comparative in silico prediction of P-glycoprotein-mediated transport for 2010-2020 US FDA-approved drugs using six Web-tools

P-glycoprotein (P-gp) is an efflux pump implicated in pharmacokinetics and drug-drug interactions. The identification of its substrates is consequently an important issue, notably for drugs under development. For such a purpose, various in silico methods have been developed, but their relevance remains to be fully established. The present study was designed to get insight about this point, through determining the performance values of six freely accessible Web-tools (ADMETlab, AdmetSAR2.0, PgpRules, pkCSM, SwissADME and vNN-ADMET), computationally predicting P-gp-mediated transport. Using an external test set of 231 marketed drugs, approved over the 2010-2020 period by the US Food and Drug Administration and fully in vitro characterized for their P-gp substrate status, various performance parameters (including sensitivity, specificity, accuracy, Matthews correlation coefficient and area under the receiver operating characteristics curve) were determined. They were found to rather poorly meet criteria commonly required for acceptable prediction, whatever the Web-tools were used alone or in combination. Predictions of being P-gp substrate or non-substrate by these online in silico methods may therefore be considered with caution.

Effect of CFTR correctors on the traffic and the function of intracellularly retained ABCB4 variants

BACKGROUND & AIM

ABCB4 is expressed at the canalicular membrane of hepatocytes. This ATP-binding cassette (ABC) transporter is responsible for the secretion of phosphatidylcholine into bile canaliculi. Missense genetic variations of ABCB4 are correlated with several rare cholestatic liver diseases, the most severe being progressive familial intrahepatic cholestasis type 3 (PFIC3). In a repurposing strategy to correct intracellularly retained ABCB4 variants, we tested 16 compounds previously validated as cystic fibrosis transmembrane conductance regulator (CFTR) correctors.

METHODS

The maturation, intracellular localization and activity of intracellularly retained ABCB4 variants were analyzed in cell models after treatment with CFTR correctors. In addition, in silico molecular docking calculations were performed to test the potential interaction of CFTR correctors with ABCB4.

RESULTS

We observed that the correctors C10, C13, and C17, as well as the combinations of C3 + C18 and C4 + C18, allowed the rescue of maturation and canalicular localization of four distinct traffic-defective ABCB4 variants. However, such treatments did not permit a rescue of the phosphatidylcholine secretion activity of these defective variants and were also inhibitory of the activity of wild type ABCB4. In silico molecular docking analyses suggest that these CFTR correctors might directly interact with transmembrane domains and/or ATP-binding sites of the transporter.

CONCLUSION

Our results illustrate the uncoupling between the traffic and the activity of ABCB4 because the same molecules can rescue the traffic of defective variants while they inhibit the secretion activity of the transporter. We expect that this study will help to design new pharmacological tools with potential clinical interest.

Current development of integrated web servers for preclinical safety and pharmacokinetics assessments in drug development

In drug development, preclinical safety and pharmacokinetics assessments of candidate drugs to ensure the safety profile are a must. While in vivo and in vitro tests are traditionally used, experimental determinations have disadvantages, as they are usually time-consuming and costly. In silico predictions of these preclinical endpoints have each been developed in the past decades. However, only a few web-based tools have integrated different models to provide a simple one-step platform to help researchers thoroughly evaluate potential drug candidates. To efficiently achieve this approach, a platform for preclinical evaluation must not only predict key ADMET (absorption, distribution, metabolism, excretion and toxicity) properties but also provide some guidance on structural modifications to improve the undesired properties. In this review, we organized and compared several existing integrated web servers that can be adopted in preclinical drug development projects to evaluate the subject of interest. We also introduced our new web server, Virtual Rat, as an alternative choice to profile the properties of drug candidates. In Virtual Rat, we provide not only predictions of important ADMET properties but also possible reasons as to why the model made those structural predictions. Multiple models were implemented into Virtual Rat, including models for predicting human ether-a-go-go-related gene (hERG) inhibition, cytochrome P450 (CYP) inhibition, mutagenicity (Ames test), blood-brain barrier penetration, cytotoxicity and Caco-2 permeability. Virtual Rat is free and has been made publicly available at https://virtualrat.cmdm.tw/.

Synthesis and evaluation of NLRP3-inhibitory sulfonylurea [11C]MCC950 in healthy animals

The diaryl sulfonylurea MCC950/CRID3 is a potent NLRP3 inhibitor (IC₅₀ = 8 nM) and, in animal models, MCC950 protects against numerous NLRP3-related neurodegenerative disorders. To evaluate the brain uptake and investigate target engagement of MCC950, we synthesised [¹¹C-urea]MCC950 via carrier added [¹¹C]CO₂ fixation chemistry (activity yield = 237 MBq; radiochemical purity >99%; molar activity = 7 GBq/µmol; radiochemical yield (decay-corrected from [¹¹C]CO₂) = 1.1%; synthesis time from end-of-bombardment = 31 min; radiochemically stable for >1 h). Despite preclinical efficacy in neurodegeneration studies, preclinical positron emission tomography (PET) imaging studies in mouse, rat and rhesus monkey revealed poor brain uptake of low molar activity [¹¹C]MCC950 and rapid washout. In silico prediction tools suggest efflux transporter liabilities for MCC950 at microdoses, and this information should be taken into account when developing next generation NLRP3 inhibitors and/or PET radiotracers.