Abstract: The emergence of large-scale phenotypic, genetic, and other multi-model biochemical data has offered unprecedented opportunities for drug discovery including drug repurposing. Various knowledge graph-based methods have been developed to integrate and analyze complex and heterogeneous data sources to find new therapeutic applications for existing drugs. However, existing methods have limitations in modeling and capturing context-sensitive inter-relationships among tens of thousands of biomedical entities. In this paper, we developed KG-Predict: a knowledge graph computational framework for drug repurposing. We first integrated multiple types of entities and relations from various genotypic and phenotypic databases to construct a knowledge graph termed GP-KG. GP-KG was composed of 1,246,726 associations between 61,146 entities. KG-Predict then aggregated the heterogeneous topological and semantic information from GP-KG to learn low-dimensional representations of entities and relations, and further utilized these representations to infer new drug-disease interactions. In cross-validation experiments, KG-Predict achieved high performances [AUROC (the area under receiver operating characteristic) = 0.981, AUPR (the area under precision-recall) = 0.409 and MRR (the mean reciprocal rank) = 0.261], outperforming other state-of-art graph embedding methods. We applied KG-Predict in identifying novel repositioned candidate drugs for Alzheimer's disease (AD) and showed that KG-Predict prioritized both FDA-approved and active clinical trial anti-AD drugs among the top (AUROC = 0.868 and AUPR = 0.364).

Abstract: Doxorubicin is a common treatment for breast cancer. However, not all patients respond to this drug, which sometimes causes life-threatening side effects. Accurately anticipating doxorubicin-resistant patients would therefore permit to spare them this risk while considering alternative treatments without delay. Stratifying patients based on molecular markers in their pretreatment tumors is a promising approach to advance toward this ambitious goal, but single-gene gene markers such as HER2 expression have not shown to be sufficiently predictive. The recent availability of matched doxorubicin-response and diverse molecular profiles across breast cancer patients permits now analysis at a much larger scale. 16 machine learning algorithms and 8 molecular profiles are systematically evaluated on the same cohort of patients. Only 2 of the 128 resulting models are substantially predictive, showing that they can be easily missed by a standard-scale analysis. The best model is classification and regression tree (CART) nonlinearly combining 4 selected miRNA isoforms to predict doxorubicin response (median Matthew correlation coefficient (MCC) and area under the curve (AUC) of 0.56 and 0.80, respectively). By contrast, HER2 expression is significantly less predictive (median MCC and AUC of 0.14 and 0.57, respectively). As the predictive accuracy of this CART model increases with larger training sets, its update with future data should result in even better accuracy.

Abstract: Cancer is a complex disease with usually multiple disease mechanisms. Target combination is a better strategy than a single target in developing cancer therapies. However, target combinations are generally more difficult to be predicted. Current CRISPR-cas9 technology enables genome-wide screening for potential targets, but only a handful of genes have been screend as target combinations. Thus, an effective computational approach for selecting candidate target combinations is highly desirable. Selected target combinations also need to be translational between cell lines and cancer patients. We have therefore developed DSCN (double-target selection guided by CRISPR screening and network), a method that matches expression levels in patients and gene essentialities in cell lines through spectral-clustered protein-protein interaction (PPI) network. In DSCN, a sub-sampling approach is developed to model first-target knockdown and its impact on the PPI network, and it also facilitates the selection of a second target. Our analysis first demonstrated a high correlation of the DSCN sub-sampling-based gene knockdown model and its predicted differential gene expressions using observed gene expression in 22 pancreatic cell lines before and after MAP2K1 and MAP2K2 inhibition (R2 = 0.75). In DSCN algorithm, various scoring schemes were evaluated. The 'diffusion-path' method showed the most significant statistical power of differentialting known synthetic lethal (SL) versus non-SL gene pairs (P = 0.001) in pancreatic cancer. The superior performance of DSCN over existing network-based algorithms, such as OptiCon and VIPER, in the selection of target combinations is attributable to its ability to calculate combinations for any gene pairs, whereas other approaches focus on the combinations among optimized regulators in the network. DSCN's computational speed is also at least ten times fast than that of other methods. Finally, in applying DSCN to predict target combinations and drug combinations for individual samples (DSCNi), DSCNi showed high correlation between target combinations predicted and real synergistic combinations (P = 1e-5) in pancreatic cell lines. In summary, DSCN is a highly effective computational method for the selection of target combinations.

Abstract: Drug repurposing is an approach to identify new medical indications of approved drugs. This work presents a graph neural network drug repurposing model, which we refer to as GDRnet, to efficiently screen a large database of approved drugs and predict the possible treatment for novel diseases. We pose drug repurposing as a link prediction problem in a multi-layered heterogeneous network with about 1.4 million edges capturing complex interactions between nearly 42,000 nodes representing drugs, diseases, genes, and human anatomies. GDRnet has an encoder-decoder architecture, which is trained in an end-to-end manner to generate scores for drug-disease pairs under test. We demonstrate the efficacy of the proposed model on real datasets as compared to other state-of-the-art baseline methods. For a majority of the diseases, GDRnet ranks the actual treatment drug in the top 15. Furthermore, we apply GDRnet on a coronavirus disease (COVID-19) dataset and show that many drugs from the predicted list are being studied for their efficacy against the disease.

Abstract: In response to fighting COVID-19 pandemic, researchers in machine learning and artificial intelligence have constructed some medical knowledge graphs (KG) based on existing COVID-19 datasets, however, these KGs contain a considerable amount of semantic relations which are incomplete or missing. In this paper, we focus on the task of knowledge graph embedding (KGE), which serves an important solution to infer the missing relations. In the past, there have been a collection of knowledge graph embedding models with different scoring functions to learn entity and relation embeddings published. However, these models share the same problems of rarely taking important features of KG like attribute features, other than relation triples, into account, while dealing with the heterogeneous, complex and incomplete COVID-19 medical data. To address the above issue, we propose a graph feature collection network (GFCNet) for COVID-19 KGE task, which considers both neighbor and attribute features in KGs. The extensive experiments conducted on the COVID-19 drug KG dataset show promising results and prove the effectiveness and efficiency of our proposed model. In addition, we also explain the future directions of deepening the study on COVID-19 KGE task.

Abstract: The coronavirus disease 2019 (COVID-19) pandemic has been severely impacting global society since December 2019. The related findings such as vaccine and drug development have been reported in biomedical literature-at a rate of about 10 000 articles on COVID-19 per month. Such rapid growth significantly challenges manual curation and interpretation. For instance, LitCovid is a literature database of COVID-19-related articles in PubMed, which has accumulated more than 200 000 articles with millions of accesses each month by users worldwide. One primary curation task is to assign up to eight topics (e.g. Diagnosis and Treatment) to the articles in LitCovid. The annotated topics have been widely used for navigating the COVID literature, rapidly locating articles of interest and other downstream studies. However, annotating the topics has been the bottleneck of manual curation. Despite the continuing advances in biomedical text-mining methods, few have been dedicated to topic annotations in COVID-19 literature. To close the gap, we organized the BioCreative LitCovid track to call for a community effort to tackle automated topic annotation for COVID-19 literature. The BioCreative LitCovid dataset-consisting of over 30 000 articles with manually reviewed topics-was created for training and testing. It is one of the largest multi-label classification datasets in biomedical scientific literature. Nineteen teams worldwide participated and made 80 submissions in total. Most teams used hybrid systems based on transformers. The highest performing submissions achieved 0.8875, 0.9181 and 0.9394 for macro-F1-score, micro-F1-score and instance-based F1-score, respectively. Notably, these scores are substantially higher (e.g. 12%, higher for macro F1-score) than the corresponding scores of the state-of-art multi-label classification method. The level of participation and results demonstrate a successful track and help close the gap between dataset curation and method development. The dataset is publicly available via https://ftp.ncbi.nlm.nih.gov/pub/lu/LitCovid/biocreative/ for benchmarking and further development. Database URL https://ftp.ncbi.nlm.nih.gov/pub/lu/LitCovid/biocreative/.

Abstract: Predictive modeling of drug-induced gene expressions is a powerful tool for phenotype-based compound screening and drug repurposing. State-of-the-art machine learning methods use a small number of fixed cell lines as a surrogate for predicting actual expressions in a new cell type or tissue, although it is well known that drug responses depend on a cellular context. Thus, the existing approach has limitations when applied to personalized medicine, especially for many understudied diseases whose molecular profiles are dramatically different from those characterized in the training data. Besides the gene expression, dose-dependent cell viability is another important phenotype readout and is more informative than conventional summary statistics (e.g., IC50) for characterizing clinical drug efficacy and toxicity. However, few computational methods can reliably predict the dose-dependent cell viability. To address the challenges mentioned above, we designed a new deep learning model, MultiDCP, to predict cellular context-dependent gene expressions and cell viability on a specific dosage. The novelties of MultiDCP include a knowledge-driven gene expression profile transformer that enables context-specific phenotypic response predictions of novel cells or tissues, integration of multiple diverse labeled and unlabeled omics data, the joint training of the multiple prediction tasks, and a teacher-student training procedure that allows us to utilize unreliable data effectively. Comprehensive benchmark studies suggest that MultiDCP outperforms state-of-the-art methods with unseen cell lines that are dissimilar from the cell lines in the supervised training in terms of gene expressions. The predicted drug-induced gene expressions demonstrate a stronger predictive power than noisy experimental data for downstream tasks. Thus, MultiDCP is a useful tool for transcriptomics-based drug repurposing and compound screening that currently rely on noisy high-throughput experimental data. We applied MultiDCP to repurpose individualized drugs for Alzheimer's disease in terms of efficacy and toxicity, suggesting that MultiDCP is a potentially powerful tool for personalized drug discovery.

Abstract: Severe acute respiratory syndrome coronavirus 2 was originally identified in Wuhan city of China in December 2019 and it spread rapidly throughout the globe, causing a threat to human life. Since targeted therapies are deficient, scientists all over the world have an opportunity to develop novel drug therapies to combat COVID-19. After the declaration of a global medical emergency, it was established that the Food and Drug Administration (FDA) could permit the use of emergency testing, treatments, and vaccines to decrease suffering, and loss of life, and restore the nation's health and security. The FDA has approved the use of remdesivir and its analogs as an antiviral medication, to treat COVID-19. The primary protease of SARS-CoV-2, which has the potential to regulate coronavirus proliferation, has been a viable target for the discovery of medicines against SARS-CoV-2. The present research deals with the in silico technique to screen phytocompounds from a traditional medicinal plant, Bauhinia variegata for potential inhibitors of the SARS-CoV-2 main protease. Dried leaves of the plant B. variegata were used to prepare aqueous and methanol extract and the constituents were analyzed using the GC-MS technique. A total of 57 compounds were retrieved from the aqueous and methanol extract analysis. Among these, three lead compounds (2,5 dimethyl 1-H Pyrrole, 2,3 diphenyl cyclopropyl methyl phenyl sulphoxide, and Benzonitrile m phenethyl) were shown to have the highest binding affinity (-5.719 to -5.580 kcal/mol) towards SARS-CoV-2 M(pro). The post MD simulation results also revealed the favorable confirmation and stability of the selected lead compounds with M(pro) as per trajectory analysis. The Prime MM/GBSA binding free energy supports this finding, the top lead compound 2,3 diphenyl cyclopropyl methyl phenyl sulphoxide showed high binding free energy (-64.377 +/- 5.24 kcal/mol) towards M(pro) which reflects the binding stability of the molecule with M(pro). The binding free energy of the complexes was strongly influenced by His, Gln, and Glu residues. All of the molecules chosen are found to have strong pharmacokinetic characteristics and show drug-likeness properties. The lead compounds present acute toxicity (LD50) values ranging from 670 mg/kg to 2500 mg/kg; with toxicity classifications of 4 and 5 classes. Thus, these compounds could behave as probable lead candidates for treatment against SARS-CoV-2. However further in vitro and in vivo studies are required for the development of medication against SARS-CoV-2.

Abstract: We characterized the impact of chronic kidney disease (CKD) on the cytochrome P450 (CYP) 3A4-mediated metabolism of saxagliptin to its metabolite, 5-hydroxysaxagliptin, using a physiologically based pharmacokinetic (PBPK) model. A PBPK model of saxagliptin and its CYP3A4 metabolite, 5-hydroxysaxagliptin, was constructed and validated for oral doses ranging from 5 to 100 mg. The observed ratios of area under the plasma concentration-time curve (AUC) and maximum plasma concentration (Cmax ) between healthy subjects and subjects with CKD were compared with those predicted using PBPK model simulations. Simulations were performed with virtual CKD populations having decreased CYP3A4 activity (ie, 64%-75% of the healthy subjects' CYP3A4 abundance) and preserved CYP3A4 activity (ie, 100% of the healthy subjects' CYP3A4 abundance). We found that simulations using decreased CYP3A4 activity generally overpredicted the ratios of saxagliptin AUC and Cmax in CKD compared with those using preserved CYP3A4 activity. Similarly, simulations using decreased CYP3A4 activity underpredicted the ratio of 5-hydroxysaxagliptin AUC in moderate and severe CKD compared with simulations using preserved CYP3A4 activity. These findings suggest that decreased CYP3A4 activity in CKD underpredicts saxagliptin clearance compared with that observed clinically. Preserving CYP3A4 activity in CKD more closely estimates saxagliptin clearance and 5-hydroxysaxagliptin exposure changes observed in vivo. Our findings suggest that there is no clinically meaningful impact of CKD on the metabolism of saxagliptin by CYP3A4. Since saxagliptin is not a highly sensitive substrate and validated probe for CYP3A4, this work represents a case study of a CYP3A4 substrate-metabolite pair and is not a generalization for all CYP3A4 substrates.

Abstract: Legionella pneumophila is the causative agent of a severe type of pneumonia (lung infection) called Legionnaires' disease. It is emerging as an antibiotic-resistant strain day by day. Hence, identifying novel drug targets and vaccine candidates is essential to fight against this pathogen. Here, attempts were taken through a subtractive genomics approach on the complete proteome of L. pneumophila to address the challenges of multidrug resistance. A total of 2,930 proteins from L. pneumophila proteome were investigated through diverse subtractive proteomics approaches, e.g., identification of human nonhomologous and pathogen-specific essential proteins, druggability and "anti-target" analysis, subcellular localization prediction, human microbiome nonhomology screening, and protein-protein interaction studies to find out effective drug and vaccine targets. Only three fulfilled these criteria and were proposed as novel drug targets against L. pneumophila. Furthermore, outer membrane protein TolB was identified as a potential vaccine target with a better antigenicity score. Antigenicity and transmembrane topology screening, allergenicity and toxicity assessment, population coverage analysis, and a molecular docking approach were adopted to generate the most potent epitopes. The final vaccine was constructed by the combination of highly immunogenic epitopes, along with suitable adjuvant and linkers. The designed vaccine construct showed higher binding interaction with different major histocompatibility complex (MHC) molecules and human immune TLR-2 receptors with minimum deformability at the molecular level. The present study aids the development of novel therapeutics and vaccine candidates for efficient treatment and prevention of L. pneumophila infections. However, further wet-lab-based phenotypic and genomic investigations and in vivo trials are highly recommended to validate our prediction experimentally. IMPORTANCE Legionella pneumophila is a human pathogen distributed worldwide, causing Legionnaires' disease (LD), a severe form of pneumonia and respiratory tract infection. L. pneumophila is emerging as an antibiotic-resistant strain, and controlling LD is now difficult. Hence, developing novel drugs and vaccines against L. pneumophila is a major research priority. Here, the complete proteome of L. pneumophila was considered for subtractive genomics approaches to address the challenge of antimicrobial resistance. Our subtractive proteomics approach identified three potential drug targets that are promising for future application. Furthermore, a possible vaccine candidate, "outer membrane protein TolB," was proposed using reverse vaccinology analysis. The constructed vaccine candidate showed higher binding interaction with MHC molecules and human immune TLR-2 receptors at the molecular level. Overall, the present study aids in developing novel therapeutics and vaccine candidates for efficient treatment of the infections caused by L. pneumophila.