Exploiting machine learning for end-to-end drug discovery and development

The Commoditization of AI for Molecule Design

Anyone involved in designing or finding molecules in the life sciences over the past few years has witnessed a dramatic change in how we now work due to the COVID-19 pandemic. Computational technologies like artificial intelligence (AI) seemed to become ubiquitous in 2020 and have been increasingly applied as scientists worked from home and were separated from the laboratory and their colleagues. This shift may be more permanent as the future of molecule design across different industries will increasingly require machine learning models for design and optimization of molecules as they become "designed by AI". AI and machine learning has essentially become a commodity within the pharmaceutical industry. This perspective will briefly describe our personal opinions of how machine learning has evolved and is being applied to model different molecule properties that crosses industries in their utility and ultimately suggests the potential for tight integration of AI into equipment and automated experimental pipelines. It will also describe how many groups have implemented generative models covering different architectures, for de novo design of molecules. We also highlight some of the companies at the forefront of using AI to demonstrate how machine learning has impacted and influenced our work. Finally, we will peer into the future and suggest some of the areas that represent the most interesting technologies that may shape the future of molecule design, highlighting how we can help increase the efficiency of the design-make-test cycle which is currently a major focus across industries.

The autocrine loop of ALK receptor and ALKAL2 ligand is an actionable target in consensus molecular subtype 1 colon cancer

Background

In the last years, several efforts have been made to classify colorectal cancer (CRC) into well-defined molecular subgroups, representing the intrinsic inter-patient heterogeneity, known as Consensus Molecular Subtypes (CMSs).

Methods

In this work, we performed a meta-analysis of CRC patients stratified into four CMSs. We identified a negative correlation between a high level of anaplastic lymphoma kinase (ALK) expression and relapse-free survival, exclusively in CMS1 subtype. Stemming from this observation, we tested cell lines, patient-derived organoids and mice with potent ALK inhibitors, already approved for clinical use.

Results

ALK interception strongly inhibits cell proliferation already at nanomolar doses, specifically in CMS1 cell lines, while no effect was found in CMS2/3/4 groups. Furthermore, in vivo imaging identified a role for ALK in the dynamic formation of 3D tumor spheroids. Consistently, ALK appeares constitutively phosphorylated in CMS1, and it signals mainly through the AKT axis. Mechanistically, we found that CMS1 cells display several copies of ALKAL2 ligand and ALK-mRNAs, suggesting an autocrine loop mediated by ALKAL2 in the activation of ALK pathway, responsible for the invasive phenotype. Consequently, disruption of ALK axis mediates the pro-apoptotic action of CMS1 cell lines, both in 2D and 3D and enhanced cell-cell adhesion and e-cadherin organization. In agreement with all these findings, the ALK signature encompassing 65 genes statistically associated with worse relapse-free survival in CMS1 subtype. Finally, as a proof of concept, the efficacy of ALK inhibition was demonstrated in both patient-derived organoids and in tumor xenografts in vivo.

Conclusions

Collectively, these findings suggest that ALK targeting may represent an attractive therapy for CRC, and CMS classification may provide a useful tool to identify patients who could benefit from this treatment. These findings offer rationale and pharmacological strategies for the treatment of CMS1 CRC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13046-022-02309-1.

Graph neural networks for materials science and chemistry

Machine learning plays an increasingly important role in many areas of chemistry and materials science, being used to predict materials properties, accelerate simulations, design new structures, and predict synthesis routes of new materials. Graph neural networks (GNNs) are one of the fastest growing classes of machine learning models. They are of particular relevance for chemistry and materials science, as they directly work on a graph or structural representation of molecules and materials and therefore have full access to all relevant information required to characterize materials. In this Review, we provide an overview of the basic principles of GNNs, widely used datasets, and state-of-the-art architectures, followed by a discussion of a wide range of recent applications of GNNs in chemistry and materials science, and concluding with a road-map for the further development and application of GNNs.

Ponatinib modulates the metabolic profile of obese mice by inhibiting adipose tissue macrophage inflammation

Obesity-induced metabolic syndrome is a rapidly growing conundrum, reaching epidemic proportions globally. Chronic inflammation in obese adipose tissue plays a key role in metabolic syndrome with a series of local and systemic effects such as inflammatory cell infiltration and inflammatory cytokine secretion. Adipose tissue macrophages (ATM), as one of the main regulators in this process, are particularly crucial for pharmacological studies on obesity-related metabolic syndrome. Ponatinib, a multi-targeted tyrosine kinase inhibitor originally used to treat leukemia, has recently been found to improve dyslipidemia and atherosclerosis, suggesting that it may have profound effect on metabolic syndrome, although the mechanisms underlying have not yet been revealed. Here we discovered that ponatinib significantly improved insulin sensitivity in leptin deficient obese mice. In addition to that, ponatinib treatment remarkably ameliorated high fat diet-induced hyperlipidemia and inhibited ectopic lipid deposition in the liver. Interestingly, although ponatinib did not reduce but increase the weight of white adipose tissue (WAT), it remarkably suppressed the inflammatory response in WAT and preserved its function. Mechanistically, we showed that ponatinib had no direct effect on hepatocyte or adipocyte but attenuated free fatty acid (FFA) induced macrophage transformation from pro-inflammatory to anti-inflammatory phenotype. Moreover, adipocytes co-cultured with FFA-treated macrophages exhibited insulin resistance, while pre-treat these macrophages with ponatinib can ameliorate this process. These results suggested that the beneficial effects of ponatinib on metabolic disorders are achieved by inhibiting the inflammatory phenotypic transformation of ATMs, thereby maintaining the physiological function of adipose tissue under excessive obesity. The data here not only revealed the novel therapeutic function of ponatinib, but also provided a theoretical basis for the application of multi-target tyrosine kinase inhibitors in metabolic diseases.

Development of machine learning classifiers to predict compound activity on prostate cancer cell lines

Prostate cancer is the most common type of cancer in men. The disease presents good survival rates if treated at the early stages. However, the evolution of the disease in its most aggressive variant remains without effective therapeutic answers. Therefore, the identification of novel effective therapeutics is urgently needed. On these premises, we developed a series of machine learning models, based on compounds with reported highly homogeneous cell-based antiproliferative assay data, able to predict the activity of ligands towards the PC-3 and DU-145 prostate cancer cell lines. The data employed in the development of the computational models was finely-tuned according to a series of thresholds for the classification of active/inactive compounds, to the number of features to be implemented, and by using 10 different machine learning algorithms. Models’ evaluation allowed us to identify the best combination of activity thresholds and ML algorithms for the classification of active compounds, achieving prediction performances with MCC values above 0.60 for PC-3 and DU-145 cells. Moreover, in silico models based on the combination of PC-3 and DU-145 data were also developed, demonstrating excellent precision performances. Finally, an analysis of the activity annotations reported for the ligands in the curated datasets were conducted, suggesting associations between cellular activity and biological targets that might be explored in the future for the design of more effective prostate cancer antiproliferative agents.

A Mutual Attention Model for Drug Target Binding Affinity Prediction

Vrious machine learning approaches have been developed for drug-target interaction (DTI) prediction. One class of these approaches, DTBA, is interested in Drug-Target Binding Affinity strength, rather than focusing merely on the presence or absence of interaction. Several machine learning methods have been developed for this purpose. However, almost all depend heavily on the use of increasingly sophisticated inputs to improve their performance. In addition, these methods do not allow any analysis or interpretation due to their black-box characteristic. This work is an attempt to overcome these limitations by taking advantage of the use of attention mechanisms with convolution models. In this paper, we define a new mutual attention based model for DTBA prediction. We represent both compounds and targets by sequences. Our model starts by aligning the drug-target pairs, then a learned masking is performed to retain the most promising regions, of both sequences, and amplify them with a learned factor in such a way to make the learning focus more on them. We evaluate the performance of our method on two benchmark datasets, KIBA and Davis. The results show that our mutual attention approach is very effective. Compared to other well-known approaches, it achieved excellent results regarding the considered performance metrics.

Combining DELs and machine learning for toxicology prediction

DNA-encoded libraries (DELs) allow starting chemical matter to be identified in drug discovery. The volume of experimental data generated also makes DELs an attractive resource for machine learning (ML). ML allows modeling complex relationships between compounds and numerical endpoints, such as the binding to a target measured by DELs. DELs could also empower other areas of drug discovery. Here, we propose that DELs and ML could be combined to model binding to off-targets, enabling better predictive toxicology. With enough data, ML models can make accurate predictions across a vast chemical space, and they can be reused and expanded across projects. Although there are limitations, more general toxicology models could be applied earlier during drug discovery, illuminating safety liabilities at a lower cost.

Drug reprofiling history and potential therapies against Parkinson's disease

Given the high whittling down rates, high costs, and moderate pace of new medication, revelation, and improvement, repurposing "old" drugs to treat typical and uncommon illnesses is progressively becoming an appealing proposition. Drug repurposing is the way toward utilizing existing medications in treating diseases other than the purposes they were initially designed for. Faced with scientific and economic challenges, the prospect of discovering new medication indications is enticing to the pharmaceutical sector. Medication repurposing can be used at various stages of drug development, although it has shown to be most promising when the drug has previously been tested for safety. We describe strategies of drug repurposing for Parkinson's disease, which is a neurodegenerative condition that primarily affects dopaminergic neurons in the substantia nigra. We also discuss the obstacles faced by the repurposing community and suggest new approaches to solve these challenges so that medicine repurposing can reach its full potential.

Drug Treatment Effect Model Based on MODWT and Hawkes Self-Exciting Point Process

In precision medicine, especially in the pharmacodynamic area, the lack of an adequate long-term drug effect monitoring model leads to a quite low robustness to the instant drug treatment. Modelling the effect of drug based on the monitoring variables is essential to measure the drug benefit and its side effect preciously. In order to model the complex drug behavior in the context of time series, a sin function is selected to describe the basic trend of heart rate variable that is medically monitored. A Hawkes self-exciting point process model is chosen to describe the effect caused by multiple and sequential drug usage at different time points. The model considers the time lag between the drug given time and the drug effect during the whole drug emission period. A cumulative Gamma distribution is employed to describe the time lag effect. Simulation results demonstrate the established model effectively when describing the baseline trend and the drug effect with low noise levels, where the maximal overlap discrete wavelet transformation is utilized for the information decomposition in the frequency zone. The real data of the variables heart rate and drug liquemin from a medical database is analyzed. Instead of the original time series, scale variable s4 is selected according to the Granger cointegration test. The results show that the model accurately characterizes the cumulative drug effect with the Pearson correlation test value as 0.22, which is more significant for the value under 0.1. In the future, the model can be extended to more complicated scenarios through taking into account multiple monitoring variables and different kinds of drugs.

AI-based chest CT semantic segmentation algorithm enables semi-automated lung cancer surgery planning by recognizing anatomical variants of pulmonary vessels

Background

The recognition of anatomical variants is essential in preoperative planning for lung cancer surgery. Although three-dimensional (3-D) reconstruction provided an intuitive demonstration of the anatomical structure, the recognition process remains fully manual. To render a semiautomated approach for surgery planning, we developed an artificial intelligence (AI)–based chest CT semantic segmentation algorithm that recognizes pulmonary vessels on lobular or segmental levels. Hereby, we present a retrospective validation of the algorithm comparing surgeons’ performance.

Methods

The semantic segmentation algorithm to be validated was trained on non-contrast CT scans from a single center. A retrospective pilot study was performed. An independent validation dataset was constituted by an arbitrary selection from patients who underwent lobectomy or segmentectomy in three institutions during Apr. 2020 to Jun. 2021. The golden standard of anatomical variants of each enrolled case was obtained via expert surgeons’ judgments based on chest CT, 3-D reconstruction, and surgical observation. The performance of the algorithm is compared against the performance of two junior thoracic surgery attendings based on chest CT.

Results

A total of 27 cases were included in this study. The overall case-wise accuracy of the AI model was 82.8% in pulmonary vessels compared to 78.8% and 77.0% for the two surgeons, respectively. Segmental artery accuracy was 79.7%, 73.6%, and 72.7%; lobular vein accuracy was 96.3%, 96.3%, and 92.6% by the AI model and two surgeons, respectively. No statistical significance was found. In subgroup analysis, the anatomic structure-wise analysis of the AI algorithm showed a significant difference in accuracies between different lobes (p = 0.012). Higher AI accuracy in the right-upper lobe (RUL) and left-lower lobe (LLL) arteries was shown. A trend of better performance in non-contrast CT was also detected. Most recognition errors by the algorithm were the misclassification of LA¹⁺² and LA³. Radiological parameters did not exhibit a significant impact on the performance of both AI and surgeons.

Conclusion

The semantic segmentation algorithm achieves the recognition of the segmental pulmonary artery and the lobular pulmonary vein. The performance of the model approximates that of junior thoracic surgery attendings. Our work provides a novel semiautomated surgery planning approach that is potentially beneficial to lung cancer patients.

Emerging trends in the nanomedicine applications of functionalized magnetic nanoparticles as novel therapies for acute and chronic diseases

High-quality point-of-care is critical for timely decision of disease diagnosis and healthcare management. In this regard, biosensors have revolutionized the field of rapid testing and screening, however, are confounded by several technical challenges including material cost, half-life, stability, site-specific targeting, analytes specificity, and detection sensitivity that affect the overall diagnostic potential and therapeutic profile. Despite their advances in point-of-care testing, very few classical biosensors have proven effective and commercially viable in situations of healthcare emergency including the recent COVID-19 pandemic. To overcome these challenges functionalized magnetic nanoparticles (MNPs) have emerged as key players in advancing the biomedical and healthcare sector with promising applications during the ongoing healthcare crises. This critical review focus on understanding recent developments in theranostic applications of functionalized magnetic nanoparticles (MNPs). Given the profound global economic and health burden, we discuss the therapeutic impact of functionalized MNPs in acute and chronic diseases like small RNA therapeutics, vascular diseases, neurological disorders, and cancer, as well as for COVID-19 testing. Lastly, we culminate with a futuristic perspective on the scope of this field and provide an insight into the emerging opportunities whose impact is anticipated to disrupt the healthcare industry.

Feasibility and application of machine learning enabled fast screening of poly-beta-amino-esters for cartilage therapies

Despite the large prevalence of diseases affecting cartilage (e.g. knee osteoarthritis affecting 16% of population globally), no curative treatments are available because of the limited capacity of drugs to localise in such tissue caused by low vascularisation and electrostatic repulsion. While an effective delivery system is sought, the only option is using high drug doses that can lead to systemic side effects. We introduced poly-beta-amino-esters (PBAEs) to effectively deliver drugs into cartilage tissues. PBAEs are copolymer of amines and di-acrylates further end-capped with other amine; therefore encompassing a very large research space for the identification of optimal candidates. In order to accelerate the screening of all possible PBAEs, the results of a small pool of polymers (n = 90) were used to train a variety of machine learning (ML) methods using only polymers properties available in public libraries or estimated from the chemical structure. Bagged multivariate adaptive regression splines (MARS) returned the best predictive performance and was used on the remaining (n = 3915) possible PBAEs resulting in the recognition of pivotal features; a further round of screening was carried out on PBAEs (n = 150) with small variations of structure of the main candidates from the first round. The refinements of such characteristics enabled the identification of a leading candidate predicted to improve drug uptake > 20 folds over conventional clinical treatment; this uptake improvement was also experimentally confirmed. This work highlights the potential of ML to accelerate biomaterials development by efficiently extracting information from a limited experimental dataset thus allowing patients to benefit earlier from a new technology and at a lower price. Such roadmap could also be applied for other drug/materials development where optimisation would normally be approached through combinatorial chemistry.

Machine Learning and Pharmacometrics for Prediction of Pharmacokinetic Data: Differences, Similarities and Challenges Illustrated with Rifampicin

Pharmacometrics (PM) and machine learning (ML) are both valuable for drug development to characterize pharmacokinetics (PK) and pharmacodynamics (PD). Pharmacokinetic/pharmacodynamic (PKPD) analysis using PM provides mechanistic insight into biological processes but is time- and labor-intensive. In contrast, ML models are much quicker trained, but offer less mechanistic insights. The opportunity of using ML predictions of drug PK as input for a PKPD model could strongly accelerate analysis efforts. Here exemplified by rifampicin, a widely used antibiotic, we explore the ability of different ML algorithms to predict drug PK. Based on simulated data, we trained linear regressions (LASSO), Gradient Boosting Machines, XGBoost and Random Forest to predict the plasma concentration-time series and rifampicin area under the concentration-versus-time curve from 0–24 h (AUC_0–24h) after repeated dosing. XGBoost performed best for prediction of the entire PK series (R²: 0.84, root mean square error (RMSE): 6.9 mg/L, mean absolute error (MAE): 4.0 mg/L) for the scenario with the largest data size. For AUC_0–24h prediction, LASSO showed the highest performance (R²: 0.97, RMSE: 29.1 h·mg/L, MAE: 18.8 h·mg/L). Increasing the number of plasma concentrations per patient (0, 2 or 6 concentrations per occasion) improved model performance. For example, for AUC_0–24h prediction using LASSO, the R² was 0.41, 0.69 and 0.97 when using predictors only (no plasma concentrations), 2 or 6 plasma concentrations per occasion as input, respectively. Run times for the ML models ranged from 1.0 s to 8 min, while the run time for the PM model was more than 3 h. Furthermore, building a PM model is more time- and labor-intensive compared with ML. ML predictions of drug PK could thus be used as input into a PKPD model, enabling time-efficient analysis.

A Collaborative Study on the Classification of Silicone Oil Droplets and Protein Particles Using Flow Imaging Method

In this study, we conducted a collaborative study on the classification between silicone oil droplets and protein particles detected using the flow imaging (FI) method toward proposing a standardized classifier/model. We compared four approaches, including a classification filter composed of particle characteristic parameters, principal component analysis, decision tree, and convolutional neural network in the performance of the developed classifier/model. Finally, the points to be considered were summarized for measurement using the FI method, and for establishing the classifier/model using machine learning to differentiate silicone oil droplets and protein particles.

Industrializing AI-powered drug discovery: lessons learned from the Patrimony computing platform

INTRODUCTION

As a mid-size international pharmaceutical company, we initiated four years ago the launch of a dedicated high-throughput computing platform supporting drug discovery. The platform named "Patrimony" was built-up on the initial predicate to capitalize on our proprietary data while leveraging public data sources in order to foster a Computational Precision Medicine approach with the power of Artificial Intelligence.

AREAS COVERED

Specifically, Patrimony is designed to identify novel therapeutic target candidates. With several successful use cases in Immuno-inflammatory diseases, and current ongoing extension to applications to Oncology and Neurology, we document how this industrial computational platform has had a transformational impact on our R&D, making it more competitive, as well time and cost effective through a model-based educated selection of therapeutic targets and drug candidates.

EXPERT OPINION

We report our achievements, but also our challenges in implementing data access and governance processes, building-up hardware and user interfaces, and acculturing scientists to use predictive models to inform decisions.

Machine Learning for Discovery of New ADORA Modulators

Adenosine (ADO) is an extracellular signaling molecule generated locally under conditions that produce ischemia, hypoxia, or inflammation. It is involved in modulating a range of physiological functions throughout the brain and periphery through the membrane-bound G protein-coupled receptors, called adenosine receptors (ARs) A₁AR, A_2AAR, A_2BAR, and A₃AR. These are therefore important targets for neurological, cardiovascular, inflammatory, and autoimmune diseases and are the subject of drug development directed toward the cyclic adenosine monophosphate and other signaling pathways. Initially using public data for A₁AR agonists we generated and validated a Bayesian machine learning model (Receiver Operator Characteristic of 0.87) that we used to identify molecules for testing. Three selected molecules, crisaborole, febuxostat and paroxetine, showed initial activity in vitro using the HEK293 A₁AR Nomad cell line. However, radioligand binding, β-arrestin assay and calcium influx assay did not confirm this A₁AR activity. Nevertheless, several other AR activities were identified. Febuxostat and paroxetine both inhibited orthosteric radioligand binding in the µM range for A_2AAR and A₃AR. In HEK293 cells expressing the human A_2AAR, stimulation of cAMP was observed for crisaborole (EC₅₀ 2.8 µM) and paroxetine (EC₅₀ 14 µM), but not for febuxostat. Crisaborole also increased cAMP accumulation in A_2BAR-expressing HEK293 cells, but it was weaker than at the A_2AAR. At the human A₃AR, paroxetine did not show any agonist activity at 100 µM, although it displayed binding with a K_i value of 14.5 µM, suggesting antagonist activity. We have now identified novel modulators of A_2AAR, A_2BAR and A₃AR subtypes that are clinically used for other therapeutic indications, and which are structurally distinct from previously reported tool compounds or drugs.

Artificial intelligence and machine learning in precision and genomic medicine

The advancement of precision medicine in medical care has led behind the conventional symptom-driven treatment process by allowing early risk prediction of disease through improved diagnostics and customization of more effective treatments. It is necessary to scrutinize overall patient data alongside broad factors to observe and differentiate between ill and relatively healthy people to take the most appropriate path toward precision medicine, resulting in an improved vision of biological indicators that can signal health changes. Precision and genomic medicine combined with artificial intelligence have the potential to improve patient healthcare. Patients with less common therapeutic responses or unique healthcare demands are using genomic medicine technologies. AI provides insights through advanced computation and inference, enabling the system to reason and learn while enhancing physician decision making. Many cell characteristics, including gene up-regulation, proteins binding to nucleic acids, and splicing, can be measured at high throughput and used as training objectives for predictive models. Researchers can create a new era of effective genomic medicine with the improved availability of a broad range of datasets and modern computer techniques such as machine learning. This review article has elucidated the contributions of ML algorithms in precision and genome medicine.

Artificial intelligence and machine learning in precision and genomic medicine

The advancement of precision medicine in medical care has led behind the conventional symptom-driven treatment process by allowing early risk prediction of disease through improved diagnostics and customization of more effective treatments. It is necessary to scrutinize overall patient data alongside broad factors to observe and differentiate between ill and relatively healthy people to take the most appropriate path toward precision medicine, resulting in an improved vision of biological indicators that can signal health changes. Precision and genomic medicine combined with artificial intelligence have the potential to improve patient healthcare. Patients with less common therapeutic responses or unique healthcare demands are using genomic medicine technologies. AI provides insights through advanced computation and inference, enabling the system to reason and learn while enhancing physician decision making. Many cell characteristics, including gene up-regulation, proteins binding to nucleic acids, and splicing, can be measured at high throughput and used as training objectives for predictive models. Researchers can create a new era of effective genomic medicine with the improved availability of a broad range of datasets and modern computer techniques such as machine learning. This review article has elucidated the contributions of ML algorithms in precision and genome medicine.

Global Analysis of Deep Learning Prediction Using Large-Scale In-House Kinome-Wide Profiling Data

In drug discovery, the prediction of activity and absorption, distribution, metabolism, excretion, and toxicity parameters is one of the most important approaches in determining which compound to synthesize next. In recent years, prediction methods based on deep learning as well as non-deep learning approaches have been established, and a number of applications to drug discovery have been reported by various companies and organizations. In this research, we performed activity prediction using deep learning and non-deep learning methods on in-house assay data for several hundred kinases and compared and discussed the prediction results. We found that the prediction accuracy of the single-task graph neural network (GNN) model was generally lower than that of the non-deep learning model (LightGBM), but the multitask GNN model, which combined data from other kinases, comprehensively outperformed LightGBM. In addition, the extrapolative validity of the multitask model was verified by using it for prediction on known kinase ligands. We observed an overlap between characteristic protein-ligand interaction sites and the atoms that are important for prediction. By building appropriate models based on the conditions of the data set and analyzing the feature importance of the prediction results, a ligand-based prediction method may be used not only for activity prediction but also for drug design.

A generalized image analytical algorithm for investigating tablet disintegration

Commonly used methods for analyzing tablet disintegration are based on visual observations and can thus be user-dependent. To address this, a generally applicable image analytical algorithm has been developed for machine vision-based quantification of tablet disintegration. The algorithm has been tested with a conventional immediate release tablet, as well as model compacts disintegrating mainly through erosion, and finally, with a polymeric slow-release system. Despite differences in disintegration mechanisms between these compacts, the developed image analytical algorithm demonstrated its general applicability through quantifying the extent of disintegration without adaptation of image analytical parameters. The reproducibility of the approach was estimated with commercial tablets, and further, it could differentiate a range of different model compacts. The developed image analytical algorithm mimics the human decision-making processes and the current experience-based visual evaluation of disintegration time. In doing so the algorithmic method allows a user-independent approach for development of the optimal tablet formulation as well as gaining an understanding on how the selection of excipients and manufacturing processes ultimately influences tablet disintegration.

Review on Artificial Intelligence Based Ophthalmic Application

Artificial intelligence is the leading branch of technology and innovation. The utility of artificial intelligence in the field of medicine is also remarkable. From drug discovery and development till the introduction of products in the market, artificial intelligence can play its role. As people age, they are more prone to be affected by eye diseases around the globe. Early diagnosis and detection help in minimizing the risk of vision loss and providing a quality life. With the help of artificial intelligence, the workload of humans and manmade errors can be reduced to an extent. The need for artificial intelligence in the area of ophthalmic is also found to be significant. As people age, they are more prone to be affected by eye diseases around the globe. Early diagnosis and detection help in minimizing the risk of vision loss and providing a quality life. In this review, we elaborated on the use of artificial intelligence in the field of pharmaceutical product development mainly with its application in ophthalmic care. AI in the future has a high potential to increase the success rate in the drug discovery phase has already been established. The application of artificial intelligence for drug development, diagnosis, and treatment is also reported with the scientific evidence in this paper.

A Classification Model with Cognitive Reasoning Ability

In this paper, we study the classification problem of large data with many features and strong feature dependencies. This type of problem has shortcomings when handled by machine learning models. Therefore, a classification model with cognitive reasoning ability is proposed. The core idea is to use cognitive reasoning mechanism proposed in this paper to solve the classification problem of large structured data with multiple features and strong correlation between features, and then implements cognitive reasoning for features. The model has three parts. The first part proposes a Feature-to-Image algorithm for converting structured data into image data. The algorithm quantifies the dependencies between features, so as to take into account the impact of individual independent features and correlations between features on the prediction results. The second part designs and implements low-level feature extraction of the quantified features using convolutional neural networks. With the relative symmetry of the capsule network, the third part proposes a cognitive reasoning mechanism to implement high-level feature extraction, feature cognitive reasoning, and classification tasks of the data. At the same time, this paper provides the derivation process and algorithm description of cognitive reasoning mechanism. Experiments show that our model is efficient and outperforms comparable models on the category prediction experiment of ADMET properties of five compounds.This work will provide a new way for cognitive computing of intelligent data analysis.

Natural Compound ZINC12899676 Reduces Porcine Epidemic Diarrhea Virus Replication by Inhibiting the Viral NTPase Activity

Porcine epidemic diarrhea virus (PEDV) is an alphacoronavirus (α-CoV) that causes high mortality in suckling piglets, leading to severe economic losses worldwide. No effective vaccine or commercial antiviral drug is readily available. Several replicative enzymes are responsible for coronavirus replication. In this study, the potential candidates targeting replicative enzymes (PLP2, 3CLpro, RdRp, NTPase, and NendoU) were screened from 187,119 compounds in ZINC natural products library, and seven compounds had high binding potential to NTPase and showed drug-like property. Among them, ZINC12899676 was identified to significantly inhibit the NTPase activity of PEDV by targeting its active pocket and causing its conformational change, and ZINC12899676 significantly inhibited PEDV replication in IPEC-J2 cells. It first demonstrated that ZINC12899676 inhibits PEDV replication by targeting NTPase, and then, NTPase may serve as a novel target for anti-PEDV.

Machine Learning in Antibacterial Drug Design

Advances in computer hardware and the availability of high-performance supercomputing platforms and parallel computing, along with artificial intelligence methods are successfully complementing traditional approaches in medicinal chemistry. In particular, machine learning is gaining importance with the growth of the available data collections. One of the critical areas where this methodology can be successfully applied is in the development of new antibacterial agents. The latter is essential because of the high attrition rates in new drug discovery, both in industry and in academic research programs. Scientific involvement in this area is even more urgent as antibacterial drug resistance becomes a public health concern worldwide and pushes us increasingly into the post-antibiotic era. In this review, we focus on the latest machine learning approaches used in the discovery of new antibacterial agents and targets, covering both small molecules and antibacterial peptides. For the benefit of the reader, we summarize all applied machine learning approaches and available databases useful for the design of new antibacterial agents and address the current shortcomings.

Computational prediction of potential inhibitors for SARS-COV-2 main protease based on machine learning, docking, MM-PBSA calculations, and metadynamics

The development of new drugs is a very complex and time-consuming process, and for this reason, researchers have been resorting heavily to drug repurposing techniques as an alternative for the treatment of various diseases. This approach is especially interesting when it comes to emerging diseases with high rates of infection, because the lack of a quickly cure brings many human losses until the mitigation of the epidemic, as is the case of COVID-19. In this work, we combine an in-house developed machine learning strategy with docking, MM-PBSA calculations, and metadynamics to detect potential inhibitors for SARS-COV-2 main protease among FDA approved compounds. To assess the ability of our machine learning strategy to retrieve potential compounds we calculated the Enrichment Factor of compound datasets for three well known protein targets: HIV-1 reverse transcriptase (PDB 4B3P), 5-HT2A serotonin receptor (PDB 6A94), and H1 histamine receptor (PDB 3RZE). The Enrichment Factor for each target was, respectively, 102.5, 12.4, 10.6, which are considered significant values. Regarding the identification of molecules that can potentially inhibit the main protease of SARS-COV-2, compounds output by the machine learning step went through a docking experiment against SARS-COV-2 Mpro. The best scored poses were the input for MM-PBSA calculations and metadynamics using CHARMM and AMBER force fields to predict the binding energy for each complex. Our work points out six molecules, highlighting the strong interaction obtained for Mpro-mirabegron complex. Among these six, to the best of our knowledge, ambenonium has not yet been described in the literature as a candidate inhibitor for the SARS-COV-2 main protease in its active pocket.

Machine Learning in Chemoinformatics and Medicinal Chemistry

In chemoinformatics and medicinal chemistry, machine learning has evolved into an important approach. In recent years, increasing computational resources and new deep learning algorithms have put machine learning onto a new level, addressing previously unmet challenges in pharmaceutical research. In silico approaches for compound activity predictions, de novo design, and reaction modeling have been further advanced by new algorithmic developments and the emergence of big data in the field. Herein, novel applications of machine learning and deep learning in chemoinformatics and medicinal chemistry are reviewed. Opportunities and challenges for new methods and applications are discussed, placing emphasis on proper baseline comparisons, robust validation methodologies, and new applicability domains. Expected final online publication date for the Annual Review of Biomedical Data Science, Volume 5 is August 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Towards a robust out-of-the-box neural network model for genomic data

Background

The accurate prediction of biological features from genomic data is paramount for precision medicine and sustainable agriculture. For decades, neural network models have been widely popular in fields like computer vision, astrophysics and targeted marketing given their prediction accuracy and their robust performance under big data settings. Yet neural network models have not made a successful transition into the medical and biological world due to the ubiquitous characteristics of biological data such as modest sample sizes, sparsity, and extreme heterogeneity.

Results

Here, we investigate the robustness, generalization potential and prediction accuracy of widely used convolutional neural network and natural language processing models with a variety of heterogeneous genomic datasets. Mainly, recurrent neural network models outperform convolutional neural network models in terms of prediction accuracy, overfitting and transferability across the datasets under study.

Conclusions

While the perspective of a robust out-of-the-box neural network model is out of reach, we identify certain model characteristics that translate well across datasets and could serve as a baseline model for translational researchers.

The Role of in silico Research in Developing Nanoparticle-Based Therapeutics

Nanoparticles (NPs) hold great potential as therapeutics, particularly in the realm of drug delivery. They are effective at functional cargo delivery and offer a great degree of amenability that can be used to offset toxic side effects or to target drugs to specific regions in the body. However, there are many challenges associated with the development of NP-based drug formulations that hamper their successful clinical translation. Arguably, the most significant barrier in the way of efficacious NP-based drug delivery systems is the tedious and time-consuming nature of NP formulation—a process that needs to account for downstream effects, such as the onset of potential toxicity or immunogenicity, in vivo biodistribution and overall pharmacokinetic profiles, all while maintaining desirable therapeutic outcomes. Computational and AI-based approaches have shown promise in alleviating some of these restrictions. Via predictive modeling and deep learning, in silico approaches have shown the ability to accurately model NP-membrane interactions and cellular uptake based on minimal data, such as the physicochemical characteristics of a given NP. More importantly, machine learning allows computational models to predict how specific changes could be made to the physicochemical characteristics of a NP to improve functional aspects, such as drug retention or endocytosis. On a larger scale, they are also able to predict the in vivo pharmacokinetics of NP-encapsulated drugs, predicting aspects such as circulatory half-life, toxicity, and biodistribution. However, the convergence of nanomedicine and computational approaches is still in its infancy and limited in its applicability. The interactions between NPs, the encapsulated drug and the body form an intricate network of interactions that cannot be modeled with absolute certainty. Despite this, rapid advancements in the area promise to deliver increasingly powerful tools capable of accelerating the development of advanced nanoscale therapeutics. Here, we describe computational approaches that have been utilized in the field of nanomedicine, focusing on approaches for NP design and engineering.

Efficient Identification of Anti-SARS-CoV-2 Compounds Using Chemical Structure- and Biological Activity-Based Modeling

Identification of anti-SARS-CoV-2 compounds through traditional high-throughput screening (HTS) assays is limited by high costs and low hit rates. To address these challenges, we developed machine learning models to identify compounds acting via inhibition of the entry of SARS-CoV-2 into human host cells or the SARS-CoV-2 3-chymotrypsin-like (3CL) protease. The optimal classification models achieved good performance with area under the receiver operating characteristic curve (AUC-ROC) values of >0.78. Experimental validation showed that the best performing models increased the assay hit rate by 2.1-fold for viral entry inhibitors and 10.4-fold for 3CL protease inhibitors compared to those of the original drug repurposing screens. Twenty-two compounds showed potent (<5 μM) antiviral activities in a SARS-CoV-2 live virus assay. In conclusion, machine learning models can be developed and used as a complementary approach to HTS to expand compound screening capacities and improve the speed and efficiency of anti-SARS-CoV-2 drug discovery.

Extended DeepILST for Various Thermodynamic States and Applications in Coarse-Graining

Molecular dynamics (MD) simulations are widely used to obtain the microscopic properties of atomistic systems when the interatomic potential or the coarse-grained potential is known. In many practical situations, however, it is necessary to predict the interatomic or coarse-grained potential, which is a tremendous challenge. Many approaches have been developed to predict the potential parameters based on various techniques, including the relative entropy method, integral equation theory, etc., but these methods lack transferability and are limited to a specific range of thermodynamic states. Recently, data-driven and machine learning approaches have been developed to overcome such limitations. In this study, we expand the range of thermodynamic states used to train deep inverse liquid-state theory (DeepILST)¹, a deep learning framework for solving the inverse problem of liquid-state theory. We also assess the performance of DeepILST in coarse-graining various multiatom molecules and identify the molecular characteristics that affect the coarse-graining performance of DeepILST.

Advancing pharmacy and healthcare with virtual digital technologies

Digitalisation of the healthcare sector promises to revolutionise patient healthcare globally. From the different technologies, virtual tools including artificial intelligence, blockchain, virtual, and augmented reality, to name but a few, are providing significant benefits to patients and the pharmaceutical sector alike, ranging from improving access to clinicians and medicines, as well as improving real-time diagnoses and treatments. Indeed, it is envisioned that such technologies will communicate together in real-time, as well as with their physical counterparts, to create a large-scale, cyber healthcare system. Despite the significant benefits that virtual-based digital health technologies can bring to patient care, a number of challenges still remain, ranging from data security to acceptance within the healthcare sector. This review provides a timely account of the benefits and challenges of virtual health interventions, as well an outlook on how such technologies can be transitioned from research-focused towards real-world healthcare and pharmaceutical applications to transform treatment pathways for patients worldwide.

Drug Properties Prediction Based on Deep Learning

In recent research on the formulation prediction of oral dissolving drugs, deep learning models with significantly improved performance compared to machine learning models were proposed. However, the performance degradation due to limitations of an imbalanced dataset with a small size and inefficient neural network structure has still not been resolved. Therefore, we propose new deep learning-based prediction models that maximize the prediction performance for disintegration time of oral fast disintegrating films (OFDF) and cumulative dissolution profiles of sustained-release matrix tablets (SRMT). In the case of OFDF, we use principal component analysis (PCA) to reduce the dimensionality of the dataset, thereby improving the prediction performance and reducing the training time. In the case of SRMT, the Wasserstein generative adversarial network (WGAN), a neural network-based generative model, is used to overcome the limitation of performance improvement due to the lack of experimental data. To the best of our knowledge, this is the first work that utilizes WGAN for pharmaceutical formulation prediction. Experimental results show that the proposed methods have superior performance than the existing methods for all the performance metrics considered.

Screening of antibacterial compounds with novel structure from the FDA approved drugs using machine learning methods

Bacterial infection is one of the most important factors affecting the human life span. Elderly people are more harmed by bacterial infections due to their deficits in immunity. Because of the lack of new antibiotics in recent years, bacterial resistance has increasingly become a serious problem globally. In this study, an antibacterial compound predictor was constructed using the support vector machines and random forest methods and the data of the active and inactive antibacterial compounds from the ChEMBL database. The results showed that both models have excellent prediction performance (mean accuracy >0.9 and mean AUC >0.9 for the two models). We used the predictor to screen potential antibacterial compounds from FDA-approved drugs in the DrugBank database. The screening results showed that 1087 small-molecule drugs have potential antibacterial activity and 154 of them are FDA-approved antibacterial drugs, which accounts for 76.2% of the approved antibacterial drugs collected in this study. Through molecular fingerprint similarity analysis and common substructure analysis, we screened 8 predicted antibacterial small-molecule compounds with novel structures compared with known antibacterial drugs, and 5 of them are widely used in the treatment of various tumors. This study provides a new insight for predicting antibacterial compounds by using approved drugs, the predicted compounds might be used to treat bacterial infections and extend lifespan.

Knowledge-based approaches to drug discovery for rare diseases

The conventional drug discovery pipeline has proven to be unsustainable for rare diseases. Herein, we discuss recent advances in biomedical knowledge mining applied to discovering therapeutics for rare diseases. We summarize current chemogenomics data of relevance to rare diseases and provide a perspective on the effectiveness of machine learning (ML) and biomedical knowledge graph mining in rare disease drug discovery. We illustrate the power of these methodologies using a chordoma case study. We expect that a broader application of knowledge graph mining and artificial intelligence (AI) approaches will expedite the discovery of viable drug candidates against both rare and common diseases.

Mycobacterium abscessus drug discovery using machine learning

The prevalence of infections by nontuberculous mycobacteria is increasing, having surpassed tuberculosis in the United States and much of the developed world. Nontuberculous mycobacteria occur naturally in the environment and are a significant problem for patients with underlying lung diseases such as bronchiectasis, chronic obstructive pulmonary disease, and cystic fibrosis. Current treatment regimens are lengthy, complicated, toxic and they are often unsuccessful as seen by disease recurrence. Mycobacterium abscessus is one of the most commonly encountered organisms in nontuberculous mycobacteria disease and it is the most difficult to eradicate. There is currently no systematically proven regimen that is effective for treating M. abscessus infections. Our approach to drug discovery integrates machine learning, medicinal chemistry and in vitro testing and has been previously applied to Mycobacterium tuberculosis. We have now identified several novel 1-(phenylsulfonyl)-1H-benzimidazol-2-amines that have weak activity on M. abscessus in vitro but may represent a starting point for future further medicinal chemistry optimization. We also address limitations still to be overcome with the machine learning approach for M. abscessus.

Explainable Machine Learning for Property Predictions in Compound Optimization

The prediction of compound properties from chemical structure is a main task for machine learning (ML) in medicinal chemistry. ML is often applied to large data sets in applications such as compound screening, virtual library enumeration, or generative chemistry. Albeit desirable, a detailed understanding of ML model decisions is typically not required in these cases. By contrast, compound optimization efforts rely on small data sets to identify structural modifications leading to desired property profiles. In this situation, if ML is applied, one usually is reluctant to make decisions based on predictions that cannot be rationalized. Only few ML methods are interpretable. However, to yield insights into complex ML model decisions, explanatory approaches can be applied. Herein, methodologies for better understanding of ML models or explaining individual predictions are reviewed and current challenges in integrating ML into medicinal chemistry programs as well as future opportunities are discussed.

Green Chemistry in the Synthesis of Pharmaceuticals

The principles of green chemistry (GC) can be comprehensively implemented in green synthesis of pharmaceuticals by choosing no solvents or green solvents (preferably water), alternative reaction media, and consideration of one-pot synthesis, multicomponent reactions (MCRs), continuous processing, and process intensification approaches for atom economy and final waste reduction. The GC's execution in green synthesis can be performed using a holistic design of the active pharmaceutical ingredient's (API) life cycle, minimizing hazards and pollution, and capitalizing the resource efficiency in the synthesis technique. Thus, the presented review accounts for the comprehensive exploration of GC's principles and metrics, an appropriate implication of those ideas in each step of the reaction schemes, from raw material to an intermediate to the final product's synthesis, and the final execution of the synthesis into scalable industry-based production. For real-life examples, we have discussed the synthesis of a series of established generic pharmaceuticals, starting with the raw materials, and the intermediates of the corresponding pharmaceuticals. Researchers and industries have thoughtfully instigated a green synthesis process to control the atom economy and waste reduction to protect the environment. We have extensively discussed significant reactions relevant for green synthesis, one-pot cascade synthesis, MCRs, continuous processing, and process intensification, which may contribute to the future of green and sustainable synthesis of APIs.

Remdesivir and EIDD-1931 Interact with Human Equilibrative Nucleoside Transporters 1 and 2: Implications for Reaching SARS-CoV-2 Viral Sanctuary Sites

Equilibrative nucleoside transporters (ENTs) are present at the blood-testis barrier (BTB), where they can facilitate antiviral drug disposition to eliminate a sanctuary site for viruses detectable in semen. The purpose of this study was to investigate ENT-drug interactions with three nucleoside analogs, remdesivir, molnupiravir, and molnupiravir's active metabolite, β-d-N4-hydroxycytidine (EIDD-1931), and four non-nucleoside molecules repurposed as antivirals for coronavirus disease 2019 (COVID-19). The study used three-dimensional pharmacophores for ENT1 and ENT2 substrates and inhibitors and Bayesian machine learning models to identify potential interactions with these transporters. In vitro transport experiments demonstrated that remdesivir was the most potent inhibitor of ENT-mediated [3H]uridine uptake (ENT1 IC50: 39 μM; ENT2 IC50: 77 μM), followed by EIDD-1931 (ENT1 IC50: 259 μM; ENT2 IC50: 467 μM), whereas molnupiravir was a modest inhibitor (ENT1 IC50: 701 μM; ENT2 IC50: 851 μM). Other proposed antivirals failed to inhibit ENT-mediated [3H]uridine uptake below 1 mM. Remdesivir accumulation decreased in the presence of 6-S-[(4-nitrophenyl)methyl]-6-thioinosine (NBMPR) by 30% in ENT1 cells (P = 0.0248) and 27% in ENT2 cells (P = 0.0054). EIDD-1931 accumulation decreased in the presence of NBMPR by 77% in ENT1 cells (P = 0.0463) and by 64% in ENT2 cells (P = 0.0132), which supported computational predictions that both are ENT substrates that may be important for efficacy against COVID-19. NBMPR failed to decrease molnupiravir uptake, suggesting that ENT interaction is likely inhibitory. Our combined computational and in vitro data can be used to identify additional ENT-drug interactions to improve our understanding of drugs that can circumvent the BTB. SIGNIFICANCE STATEMENT: This study identified remdesivir and EIDD-1931 as substrates of equilibrative nucleoside transporters 1 and 2. This provides a potential mechanism for uptake of these drugs into cells and may be important for antiviral potential in the testes and other tissues expressing these transporters.

Enhancing preclinical drug discovery with artificial intelligence

Artificial intelligence (AI) is becoming an integral part of drug discovery. It has the potential to deliver across the drug discovery and development value chain, starting from target identification and reaching through clinical development. In this review, we provide an overview of current AI technologies and a glimpse of how AI is reimagining preclinical drug discovery by highlighting examples where AI has made a real impact. Considering the excitement and hyperbole surrounding AI in drug discovery, we aim to present a realistic view by discussing both opportunities and challenges in adopting AI in drug discovery.